Intelligent Machines 809 transcript

Please be advised this transcript is AI-generated and may not be word for word. Time codes refer to the approximate times in the ad-supported version of the show

0:00:00 - Leo Laporte
It's time for Intelligent Machines. Our guest today, gary Marcus. He is an expert in AI, writes an AI Substack newsletter and is also a little bit skeptical about AGI. We'll get to talk to Gary and then give you all the AI news, including $200 billion expenditure on an AI center for Meta. The Turing Award recipients, who are also AI experts, warning about the future and the return of Digg. It's all coming up. Next on I Am Podcasts you love. From people you trust.

This is Twit this is Intelligent Machines. Trust this is twit. This is intelligent machines. With paris martineau and jeff jarvis, episode 809, recorded wednesday, march 5th 2025. Fun mustache it's time for intelligent machines. The show we talk about intelligent machines. They're all all around us. You know, jeff Jarvis is here, professor Emeritus of Journalistic Innovation at the Craig Newmark Graduate School of Journalism at the City University of New York, now at Montclair State University and SUNY Stony Brook. Hello Jeff, hello boss, hello boss. Also, from the information, the wonderful Paris Martineau, where she covers youth issues and other things. What?

0:01:31 - Paris Martineau
are you working?

0:01:31 - Leo Laporte
on right now.

0:01:33 - Paris Martineau
I'm working on a story about Section 230 and other federal online safety bills that involve children.

0:01:43 - Leo Laporte
Good, good, you know the right thing to say about that, right, of course. Okay, I'm just, it's just, it's unsaid, you know. I'm just gonna rub my nose and say you know what to say, right, you know. You know what this loves 230.

Yeah, you know what to say so love 230, the politicians anyway continue the new format this show, as we've shifted from this week in Google to something more interesting. Ai is to start with an interview. We'll do AI news and other news later, but I always think that the interview is kind of the heart and soul of this. We've talked to AI accelerationists. We've talked to people working in the business. We've got a great guest with us right now Gary Marcus's newsletter, marcus on AI, makes him a leading AI expert. I know it says so on the screen. He is also a psychologist and cognitive scientist, written a number of books. His latest is very germane, taming Silicon Valley how we can ensure that AI works for us. Gary, it's great to have you on. I'm a regular reader of the sub stack and thanks for having me. And you are and we should, you know, I guess an AI is it fair to say, contrarian?

0:03:00 - Gary Marcus
you could say that I I love AI, but I don't like the way it's happening now, and if that makes me a contrarian, then you can call me a contrarian.

0:03:09 - Jeff Jarvis
I'm so happy you're on, gary, some welcome balance to the podcast.

0:03:15 - Paris Martineau
We've had multiple episodes of Leo thanking guests being like. Your products are making the world a better place. Everyone go out and buy it, so we need a thinker here.

0:03:26 - Leo Laporte
We did one like that. Okay, Just one.

0:03:28 - Paris Martineau
Oh, it felt a little bit longer than one, but we do have AI.

0:03:32 - Leo Laporte
Last week I thought Stephen Wilfrum was very good and sensible about AI. Next week, ray Kurzweil, who coined the term intelligent machines. He's probably a little more bullish on AI than you are.

0:03:50 - Gary Marcus
I offered to debate him and he said when we talked to Ted he would do it. And then I got the venue lined up and he disappeared on me. So tell him, I'd still love to do that.

0:03:54 - Jeff Jarvis
I will also. You've also offered a debate, elon uh he never responded.

0:03:59 - Gary Marcus
Who else is a long list you?

0:04:01 - Leo Laporte
should offer to wrestle Elon or get in the getagon.

0:04:05 - Paris Martineau
Who would be your top person to debate?

0:04:09 - Gary Marcus
I guess it depends on what one's trying to accomplish with such a debate. One of the reasons I've offered to debate a lot of these people is to try to hold them accountable to the things that they say in public and that they don't really want to defend against intellectual discourse. So that includes the exaggerations of Elon and the exaggerations of Sam and maybe Ray, although I don't think he's essential to that discourse anymore. But a lot of my offer to debate has been to try to get people to stand by the claims that they make.

0:04:41 - Leo Laporte
There's a lot of hype there. I mean, there's no question there's a lot of hype, are there?

0:04:45 - Paris Martineau
any particular people or businesses that I guess off this topic that you think are the biggest types, viewers or need to be held accountable the most in this moment.

0:04:56 - Gary Marcus
He named two of them. It's a crowded list, but Open has historically been at the top of that list. I think that they've weaponized hype the most of any of these companies. They've hyped their mission as being for the benefit of humanity, which they haven't really done. They've hyped their stuff as being about AGI, but we can all see the GPT 4.5 isn't really that. I mean I think I would put them at the top of the list. But other people have learned that hype is the way to drive up stock valuation. So I think Dario Amode has learned that. Of course Elon has done that for a long time. It's become standard practice because the media doesn't in general hold people accountable. They just look at the next promise and the next promise and the older promises. When they're not met, nobody really holds holds anybody accountable we should mention that you have founded machine learning startups.

0:05:49 - Leo Laporte
You sold your first to uber, so you have worked in the field. Uh, quite famously, you signed that uh letter in 2023 that said let's stop for six months. Looking back, obviously, we didn't. Elon signed that letter as well. Looking back, would you still like us to pause?

0:06:10 - Gary Marcus
Well, I mean it's interesting because that letter was about pausing the development of GPT-5. And it turns out that how long is it now, 18 months later, or something like that? You didn't have to write that letter. We didn't have to write that letter because they couldn't figure out how to make GPT-5 anyway. The reason that we I mean I didn't write the letter, but the reason I signed it and helped publicize it, even though I didn't think it was a perfect letter, was I thought we needed to call attention to how fast things were moving and how little we understood about how to make AI safe and reliable and so forth. And we still haven't figured out how to make AI safe and reliable.

For a while I think the letter had a good positive consequence got all the governments to start thinking about what we should do about AI. And then there was this Paris AI summit and you know change in regime in Washington and I think everything fell apart. So for a little while it had a salutary effect. I think the most depressing part is that Elon signed that letter and immediately started building xai, which was trying to do basically the same thing, um, as gpt5 was supposed to do.

He also failed to get there so far um, yeah after elon signed it, it felt like all the attention was oh, you see, elon wants to pause so he can catch up which turned out to be true, which you know, kind of undermined the movement because they used to call it in the 60s, the letter called it an existential threat.

0:07:30 - Leo Laporte
Do you still believe that, agi? I never liked that phrase, to be honest.

0:07:36 - Gary Marcus
And I can't remember the exact wording that the letter had around, that it was softer on that than some of the people who wrote it typically are, and that made me a little bit more inclined to write about it In my sub stack. I actually wrote a bunch about the letter at the time. I don't think that AI is going to annihilate us all, or I think that's at least pretty unlikely. But I worry a lot about what bad actors are doing, both with current AI, which is not very sophisticated, and what they might do with future AI as it gets more sophisticated.

And I think the so-called alignment problem figuring out how to get machines to do the things that we want them to do is entirely unsolved. We can't even do things like say, hey, don't make shit up, Machines make stuff up anyway. It's actually in the system prompt for one of these Apple systems don't make stuff up and it makes stuff up anyway. So we don't have very good control of large language models. They just kind of do what they do and you throw more data and you hope for the best, and that's really not a good way to do engineering.

0:08:32 - Leo Laporte
You've talked about the Pareto principle, that 80% is easy. It's the last 20% that's hard, and we're really actually seeing that right now with AI.

0:08:41 - Gary Marcus
We're totally seeing that you can make a demo of anything if you have enough data. That's kind of the lesson the last couple of years. But to make it really good is very hard. We saw that movie before. With driverless cars we have demos. We have demos in San Francisco, we have demos in Phoenix, but if you live in Buffalo this stuff is of no use to you.

So it's very easy to make a demo of some AI thing by having a lot of examples. The 80-20 is you get examples for 80% of what you want to do and the other 20% you don't really have exactly in your database and the systems don't exactly work. That's been a serious, serious problem. You may have seen me use the phrase the other day in a piece that I wrote about. Kevin Roos had a thing about how we're not going to be needing you know how amateurs are going to be able to code using these tools without knowing anything about coding and the way I used it there is. I said look, you can get 80% of any system you want and you kind of get a demo of it, but actually getting it to work reliably is very hard to get that last 20%.

0:09:44 - Jeff Jarvis
Gary, I've got a weird question for you. What's more dangerous the reality of AI where it stands today or the hype around it? You have done a lot trying to puncture the definitionless phrases like AGI and super intelligence and all that, and that seems to lead to a lot of the danger, it strikes me, rather than the machines where they stand.

0:10:08 - Benito Gonzalez
You said you love.

0:10:09 - Jeff Jarvis
AI, ai can do some neat things, but it's being presented in such a way that it's overblown.

0:10:15 - Gary Marcus
no, yeah, I mean, I think there's real danger on both sides. I'm not sure how I come out on the letter, let me just think it through. So the the danger of current ai are things like bias, which are very serious, especially when you have something like elon trying to use doge to evaluate who should keep jobs. And we know these systems are biased against white males, which is actually what they want. Bias towards white males, sorry, towards white males. Yes, exactly, um, they're biased towards white males and they wouldn't be using it for a biased against white.

Sorry, it's important to share there. Um, so you know, he's probably comfortable with a bias towards white males. I'm not, obviously. Um, uh, so you have problems like that. You have problems where these things are used for cyber crime and we may see some massive cyber crime hit this year and you know, if a whole bunch of infrastructure goes down because of cyber crime, then we're going to say you know, those consequences were pretty serious.

So there's some unknowables, there's misinformation. Misinformation people want to say it didn't affect the last election, but I don't really think that's true. I think that misinformation was a subtext to the election. It may get worse. The tools for, you know, making deep fakes keep getting better, etc. So you have all those kinds of risks.

Another risk, I think, is that we screw up society by spending so much on the large language models, ripping up the environment and so forth. Then there's a kind of opportunity cost that I worry about, which comes from the hype, which is you have these large language models and we're put, you know already, not society, but big businesses have put in half a trillion dollars on them and want to put in more, and probably we should be looking at other approaches to AI, so, like something that not everybody knows about me is I actually want AI to work. I think the caricature of me is oh, gary hates AI. But that's not true. I'm actually constantly writing about how could we do AI better such that we could trust it, so that it could be reliable, so that it could help more in medicine and so forth. Is AI, on balance, so far been a net positive to society or a net negative to society? So you know, there are positives, like you can use these tools for brainstorming, coders, write a little bit faster and then there are negatives, like all the things that I just laid out, and it's not totally clear how that balance works and it's not totally clear where that balance would be in five years.

I would say that when I first started having debates about this, people would be like it's going to 10x everybody's productivity and blah, blah, blah. You look at the GDP and it's hardly moved at all. Even Satya acknowledged this a week or two ago. There's been no real change in the gdp. So the notion that it was going to revolutionize the world certainly has not transparent transpired yet, and the notion that, like a lot of deep fake porn, was going to make a lot of trouble for a lot of teenage girls has transpired. So yeah, these are all you know, moving targets and so forth, but I would say the productivity gains have actually been relatively modest. It's still very much a gamble on the future and there are some real risks. And then there are the risks of like could some terrorist use this stuff to make bioweapons? The answer seems to be probably yes. There are rate-limiting steps and so forth.

It's complicated, but you balance all this out. It's really not clear. It's not a clear win, it's not an absolutely clear loss. When you add in the environmental impact, you start to worry. Then you can say, well, maybe DeepSeq will make it more efficient. It's all still in play. There's no clear benefit and there's no absolutely clear negative. Let me say that slightly more clearly. There are some clear benefits. There's some clear cost. It's not yet clear how it all nets out.

0:14:07 - Leo Laporte
I would submit, though, that you could say that about many technologies. You could have said that about the internet. You could have said that about social media. You could have said that about the atomic bomb. That's kind of the nature of new technologies, isn't it? It?

0:14:19 - Gary Marcus
is, and it's why we have to be careful about how we use those technologies, how we distribute them, what we do with them. So you know, atomic weapons it's still a very complicated thing and our view about that could change. You know one it was an old t-shirt, I think said one nuclear bomb can ruin your whole day.

0:14:37 - Leo Laporte
Like we don't you know the answer could change right and I had the poster that said war is not good for our children and other living things as well.

0:14:43 - Gary Marcus
Yeah, that was another great one um, and so, like I don't know the final answer, like when history is, you know, written well, that's the problem, right yeah, and social media. I'm willing to go out on a limb and say it sucks. It really wasn't a great thing.

0:14:57 - Leo Laporte
I'm, you know, I'm sorry that I'm addicted to it, and you know, I think but we're a little farther down the road with it, so it's easy to say that in hindsight you can look that.

0:15:05 - Gary Marcus
But yeah, in the early days of ai when I was testifying in the senate and we were talking about section 230. This was about what he, almost two years ago, you know everybody in the room said hey, we really up with social media. And everybody means senators. A bunch of senators said we, we up with social media, we should do better here and we should get rid of section 230. And of course we didn't get rid of section 230. And they said we should do something about ai. And they didn't do anything about ai. And so, like we all know that, that you know section 230 empowered social media too much and, and, um, sponsored miss and, you know, made it attractive to the companies to spread misinformation. That was bad right, I think we know.

0:15:45 - Leo Laporte
But, gary, that's. The whole point is that a lot of technologies are double-edged swords and you can focus on the negative as well as the positive. All do use, I mean you know, without Section 230, we wouldn't have a chat room, we wouldn't have forums.

0:15:57 - Gary Marcus
This personally Twit my company would not have a Mastodon instance because we couldn't afford the liability. So there's pros and cons, they're all. They're all double-edged and in a rational world we regulate things so that these double-edged swords, you know, to minimize the downside. So we guns, let's say, are double-aged sword and one of the ways that we handle the downside is we have a bunch of regulations around who can use them, you know what background testing to do and so forth. Um, you know we throw people in jail if they misuse them and so forth. But you have these people like andreessen that want no regulation or on ai at all well, that's because we're not at the gun level.

0:16:35 - Leo Laporte
We're at the gut. We just invented gunpowder, right in terms of ai, and so yeah I mean, you can be premature in regulation. If you don't understand it. We held off on regulating the internet. I think probably rightly so, because we didn't know yet what it was going to be and there was a certain yeah, I appreciate what you're saying.

0:16:55 - Gary Marcus
I think there's enough things we already know that we can do some of the regulation that we need to do. I think it's certainly the case that anything that we do has to vary over time, right? So you know, any regulation that anybody ever writes tends to not be exactly right the first time and you need variation. So, for example, the euai act has built into it a period of review. I think it's after three years or five years or whatever.

Um, and you know you've got to do that here and you know we we change regulation all the time. We used to do that here and we change regulation all the time. We used to do that at the national level when Congress actually worked. We don't so much anymore now that it doesn't work, but we used to do that fairly regularly. And certainly any rational notion of regulation of AI has to include the dynamics of change over time, of change over time. And part of why social media has been so ineffective is because Congress has, you know, basically failed to function for whatever it is the last 10 years or something like that. And so, even when it became apparent that it wasn't exactly working, right, that you needed some kind of midstream course correction. Nobody could actually get it done.

0:18:00 - Jeff Jarvis
Harris, you were trying to.

0:18:03 - Paris Martineau
Oh, I was just going to comment on to Leo's point about oh well, ai is so nascent. Leo, you're always saying that AI is incredibly advanced. We're days minutes away from AGI and a revolution. Does that not seem like a reasonable point in which to step back and think about how these technologies should and shouldn't be applied?

0:18:24 - Leo Laporte
If you knew how to think about it, maybe, but I don't know if we know how to think about how these technologies should and shouldn't be applied. If you, if you knew how to think about it, maybe, but I don't know if we know how to think about it. Do you think, for instance, gary, that the EU's AI regulation is proper and works?

0:18:35 - Gary Marcus
I mean you can't answer whether it works yet because it hasn't really gone online yet. Um, is it proper? I think the spirit of it is correct. There's a lot of Implementational details yet to determined. There's a lot of political jockeying from the United States to try to weaken it, although US-European relationships just took a giant turn for the worst, and so I don't know if the US still has any leverage over Europe or not. We'll see. Or vice versa the spirit of.

It is right that you want to look at particular kinds of risks and you want to basically look out for human rights, that you want to make sure that humans are not screwed by giant companies and that there's some kind of balance, but that you also allow innovation to prosper, and so forth.

0:19:17 - Leo Laporte
That's the trick, isn't it? That's the trick, and that's one of the things people worry about with the EU and the California attempt to regulate ai is that you could actually kill the baby in the in the.

0:19:28 - Gary Marcus
Yeah, that regulation, I mean look, if the california law had passed, then you know company like google that has a certain number of you know trillion dollars in market cap would have spent you know an extra let's say, 10 million dollars a year um filling out paperwork. And no, it wouldn't have changed anything except that it would have given some sense of responsibility, that they actually need to monitor the consequences of their actions and that they can be held liable for those actions.

It also would have protected whistleblowers, and we have Sushir Balaji who was killed on or I should say that more carefully who died in mysterious circumstances that are still not entirely clear um, you know, having some protection for whistleblowers would be really great, and I agree we didn't get it yeah, there's something, but go ahead and.

0:20:15 - Leo Laporte
But I want to get more philosophical before we lose our window with gary, because he is a cognitive psychologist and he's written some very interesting things about language and stuff. But go ahead, we can continue on this line of questioning. I just wanted to ask a little more.

0:20:28 - Jeff Jarvis
Gotta ask about agi because, gary, you've been involved in this obviously for much longer than than the, the fewer. Uh, since chachi pt came out, was there a moment when you saw the trope of agi emerge as the putative goal, with no definition? You said, oh yeah, that's interesting. So.

0:20:47 - Gary Marcus
I mean the term goes back to like 2002 or something like that 2005,. Ben Goertzel, shane Legg, who was one of the DeepMind founders, coined it. I have an interesting anecdote there, which is in 2018, I wrote an article called Deep Learning or Critical Appraisal, which was on archive, and I said this stuff's not going to get you to AGI and the stuff was models that are a predecessor to LLMs. Llms were not really known at that point, but the general kind of ancestor to them was well known at that point and I critiqued them and I said they use too much data, they don't generalize very well, blah, blah, blah. Many of the same critiques that I now Jan LeCun would make about large language models. And one of the reactions I got and you can see this because I wrote an FAQ on all the grief that I got on this paper. I wrote it for Medium One of the reactions I got is nobody's really trying to build AGI, which I thought was kind of bullshit at the time.

I thought, you know, a lot of people actually hoped to build AGI. They just weren't announcing it as that because they didn't know what they really had or whatever. But DeepMind had already said as their motto first we solve intelligence, then we solve everything else. They were always gunning for AGI from the very beginning, and it was a very disingenuous reply to me. In 2018, there's been a lot of disingenuous replies over the earth, but it was a particularly disingenuous reply. They said oh, nobody, you know, you can't judge this stuff by the standards of AGI, because nobody's working on that and it was a lie. Then it's just nobody wanted to admit that. That's what they aspired to do. Now it's very trendy oh, we're working on AGI. The limitations haven't changed, but the attitude and the valuations of the company have changed.

0:22:37 - Jeff Jarvis
Have you yet nailed down anyone to a definition of AGI?

0:22:41 - Gary Marcus
I mean that's also interesting. So in 2022, I think it was I offered Elon $100,000 bet, saying that AGI this would be my sub stack, saying that AGI was not going to come anytime soon, wasn't going to come by the year 2029, as I recall. And I shopped the definition around on Twitter. I said you know, here's how I understand AGI. And I shopped it with Shane Legg and Ben Goertzel, who had originally coined the term, and Peter Voss doesn't quite get as much credit because he thought about it earlier and then it was reincarnated or whatever. I ran past all these guys and they were like that's fine. What I said is, you know, having the flexibility of human cognition and so forth, being able to do whatever people can do, and maybe better, right? So it would still be AGI if it was better than humans and Lord knows, humans have lots of flaws. I wrote a whole book about that called Kluge, so you can easily imagine intelligence that would be better. I mean already like chess computers are better at playing chess than we are and calculators are better at doing math. So you can imagine a system that is basically equal to or better than us. And I emphasized flexibility being able to do different tasks, not just because you train the entire internet, but because I give you an instruction, like touch your nose and then touch your head, and from one trial you understand how to do that task.

And so I put this out on Twitter. I said this is the definition I got. Is it okay? And Shane said it was okay, and Ben said it was okay. And then I proposed some tasks. I said you need to be able to do this by 2029. If you can do three out of the five, we'll call it general intelligence and everybody was fine with that definition in 2022.

There really wasn't much pushback. If you look at the discussion on Twitter and back then there were more people in this field actually on Twitter A lot of people fled for reasons we don't need to go into, and so everybody was totally fine with it. They're like, okay, that's a reasonable definition, these are reasonable tasks and so forth. And now, like I don't know if you're putting that up on screen, ezra klein yesterday was kind of you know, backing down to maybe it's just doing a bunch of human tasks, which is a much weaker um definition than we had talked about in 2022, and I have a whole article that I wrote um getting the name of it, um, showing all the ways in which people are trying to retrench.

So, for example, instead of defining agi in terms of human flexibility and cognition, it's very general, artificial, general intelligence. That's what AGI is. Instead of doing that they're like well, if it makes a certain amount of money, like that's just a different definition. You can ask you know, you know, does this technology make a hundred billion dollars? But then my iPhone is, by definition, agi because it made a hundred billion dollars for Apple and that just seems like a silly.

0:25:24 - Leo Laporte
Why are we fighting? Uh, I don't understand why. We're fighting over the semantics of what agi is I mean. Oh, I'll tell you why?

0:25:30 - Gary Marcus
because, well, I mean, there's multiple reasons. One of the biggest reasons is actually that, um, microsoft agreed to a crazy clause in their contract with open ai saying that open ai gets the ip back if they achieve agi. So there's suddenly, there's a lot of money at stake, so that's one reason. And then it's bragging rights and you know, people want to say that the thing that they've got is AGI. There are people that want people like me to go away and say that it's been achieved. So stop, you know, criticizing. And there's a bunch of different reasons why, um, there's a lot of kind of very pragmatic.

0:26:03 - Leo Laporte
I mean, if, if AI can do something valuable, I don't understand why we care about it. I understand why Microsoft has been doing valuable things for a long time.

0:26:13 - Gary Marcus
So so, for example, google searches AI and it's been around what 25 years. It's very valuable. I still probably use it, or I use a knockoff called DuckDuckGo. You know, multiple times every day. Gps I use in my car all the time. Those are AI. Nobody gets excited about them. They don't think that they're, you know, changing the world, but they're actually super useful. I find you know I would sooner give up on ChatGPT than I would on GPS. I think you know GPS is more useful to me because I can count on it. It's reliable.

0:26:44 - Paris Martineau
I'm curious on a day-to-day basis, do you use any of the popular AI tools?

0:26:49 - Gary Marcus
No, the one I probably use the most is just making images, which I do for fun and almost satirically to mock AI. I don't trust them and my business is being right about AI at some level and if I make mistakes I got a lot of pushback. If I use this stuff it would be a nightmare for my particular line of work. I understand there are businesses where somebody needs to come up with marketing slogans and it may suit their purposes or something like that. It doesn't really suit my purposes. It's not trustworthy enough.

I try to give the best reasoned argument about whatever it is that I happen to write about a particular day or think about or whatever, and it doesn't do that. It gives you the sort of generic consensus average of anything that it writes about. It hallucinates, makes up references like this. It'd be a nightmare for me to use it. Um, I sometimes joke because I wrote my last book, um in two months and people say did you write chat gp? Use chat gpt to write it? And I say I don't pay for sex and I don't pay for chat you know you're going to get philosophical and I stopped you well, you know I philosophical and I stopped you.

0:28:08 - Leo Laporte
Well, you know I. I guess there, as a, you have said that neural networks are not enough.

0:28:10 - Gary Marcus
Um, that there was a survey just now showing that 85 of people at the triple ai society were just surveyed. Agree with me I mean I wasn't mentioned but agree that neural networks are not enough. That is a huge change from three years ago. I think a lot of people have come to see the light.

0:28:28 - Leo Laporte
They're not enough yeah, uh, is there a technique that you think, uh, you, you mentioned symbolic AI as being one technique that could supplement neural networks.

0:28:38 - Gary Marcus
Yeah, I mean, what I really favor is neurosymbolic AI, which would be a hybrid of classical symbolic AI, which just looks like computer programming basically, and neural networks. So in computer programming, everything is explicit. You have clear variables, you do explicit operations, you do comparisons, you store them, you run subroutines and stuff like that. In neural networks, you run everything in the simplest version through this feed-forward similarity process where you have these weights and you add up the weights and so forth. They're very different techniques, they're not very compatible as they have been built, but I think we need to bring them together and I think that's where the real action is and yeah, we, we have kind of made surprising strides with neural networks in in some ways unexpected results.

No, I mean yes and no like they do things people didn't imagine. You know that is true, but they still fail on the things that I was writing about in 1998, in 2001, when I when I was a lot less gray. Um, you know, they still fail at abstraction, at reasoning, at keeping track of properties of individuals. I first wrote about hallucinations in 2001. So the core things that I think you need in order to get to a trustworthy, reliable reasoning system that's at least as powerful as a person we still have not really solved with the neural network architecture. So it is true, like deep search is impressive or deep research is impressive, but then you look at the output and it has the same old problems it still hallucinates.

So I used deep search and it told me there was an well. So I asked it for all the cities that were, all the major cities that were west of Denver, and it forgot Billings, montana. And so I said what happened to Billings? And it told me there was an earthquake in Billings of 5.5 magnitude. And there wasn't a 5.5 magnitude earthquake. So it misunderstood the question and it made up a hallucination. It made this beautiful chart, but the chart was wrong. It left out a serious city with a population of whatever it was 120,000. At some level. What I was saying in 2001 is if all you do is tabulate statistics and you don't have proper explicit representations like they used in classical AI, you are not going to get a trustworthy machine. And that is still true. It's a quarter century after I wrote that book.

0:30:55 - Leo Laporte
I think Stephen Wolfram would agree with you. Um, but the philosophical question I have, which may be unanswerable or may even be a stupid question, as a cognitive scientist um, what is it that we do that's so different from what a machine is doing or a neural network is doing, or an llm is doing, so my view about this is that the crudest way to think about is Daniel Kahneman, system one and system two right, which most people know, right.

0:31:23 - Gary Marcus
So system one is kind of a statistical approximation, it's reflexive, it's automatic, and system two is more deliberative and more explicit and humans actually use both and there's actually tension in how we use them and so forth, but we use both and we balance the two. Current AI is mostly like system one, and I'm not the only person who's reached that conclusion. For example, yashua Bengio has also reached that conclusion, even though we disagree about a lot of technical things I think we agree about some other set of things, but we disagree about a lot of technical things but we both come to that conclusion that current AI is good at system one. We both come to what he would call a problem of distribution shift, which is something I was writing about a long time ago under slightly different terms. We both agree that that means that if you just train on a lot of data, you still wind up with a lot of outlier cases, which is what's killed.

Driverless cars, I mean I won't say killed, but it's really delayed driverless cars. You can't really drive one in Buffalo. You can only drive them in a very well-mapped city with nice weather and stuff like that. So the inability of system one to deal with outliers has been a major problem. We need to have a system two that looks at least a little bit more like classic symbolic AI, and what we really need is a hybrid between the two. That's certainly at least what has worked for humans. I think it is necessary, but not sufficient, to use philosophical terms. Um, you know, it's it's not the only thing we need to do, but it's a vital step to getting further, I believe uh, rich sutton today won the turing award along with andrew bardo.

0:32:58 - Leo Laporte
Uh, there's a story in the financial times about how they warn over unsafe deployment of ai models. But I also remember rich sutton wrote the bitter lesson paper a few years ago in which he said really all you need to do is not improve the models, but just throw more computing at it.

0:33:12 - Gary Marcus
Yeah, I keep meaning to write a reply to it. I want to call it the bitter lesson about the bitter lesson. The bitter lesson about the bitter lesson is the bitter lesson is right for the things that it works for and wrong for the rest. And so the things that it works for are problems you can do with a certain kind of pattern recognition or a certain kind of reinforcement learning. The things it doesn't work for are things like reasoning and dealing with outlier cases and so forth, and it's never really worked for that. So it is true that the set of things we can do with the better, less bitter lesson has somewhat expanded, and it's also true there's a bunch of things like hallucinations that keep popping up over and over again when you try to get them to reason, like do you know these river crossing problems?

You have a man and a goat and cabbage, or whatever a guy named Colin Frazier has had incredible fun ridiculing these systems by showing that if you make a slight variation on a traditional problem, they fall apart, or he does the ones with. The doctor winds up in the hospital and it's the son or the grandmother or whatever. Yeah.

These systems are just they don't really understand the concept of doctor or hospital or river or goat or whatever, and it's always easy to break them and show that they just really don't know what they're talking about. And that's the other side of the bitter. Lesson is, even when you have like all the data in the planet, you still can't get these systems to reliably solve variants on river crossing problems.

0:34:34 - Leo Laporte
So I guess the final thing I would say is that's why I'm not really thrilled about the discussion of what AGI is. Do we need an AI to match us, or can it just be a tool?

0:34:47 - Gary Marcus
Entirely depends on the problem. So GPS doesn't solve its problem exactly the way a person does, and that's fine.

0:34:54 - Leo Laporte
But it solves it adequately.

0:34:55 - Gary Marcus
Yeah, it's perfectly adequate. So it really does depend on the problem. The meta problem is that OpenAI sold the world on this notion that ChatGPT could do anything you wanted to do, and to do that you really do need AGI. If you want to do something narrow, like GPS or chess or something like that, fine, you certainly don't need AGI. But once you sell this dream, you know now to deliver on it you actually probably do need something like AGI.

0:35:26 - Leo Laporte
Gary, I want to thank you so much. Everybody should subscribe to Gary's Substack. Actually, you're very generous. I haven't subscribed because most of the stuff I want to read is free still. Maybe you should fix that. I've only had two paid things.

0:35:39 - Gary Marcus
I'm not doing it for the money. I like it when someone subscribes. It's like a bit of validation, but I'm trying to get ideas out in the world and so I don't charge for that.

0:35:48 - Leo Laporte
Well, you're definitely doing that. It's much appreciated and I'm really glad to get you on and I will upgrade to paid. I'll click that button. I will be on it, Gary Marcus. The newest book is out now, which talks about a lot of what we've just talked about.

0:36:05 - Gary Marcus
Yeah, I'll tell you one sentence. I know you've got to wrap up, but Taming Silicon Valley, the beginning of the book. It talks about the possibility that tech oligarchs might take over the world.

0:36:20 - Leo Laporte
Imagine that the possibility. You wrote that last year. How did you know? How did you know? Gary, thank you for being a voice of sanity.

0:36:26 - Gary Marcus
Thank you for the writing on the wall, but it's even darker than what I read.

0:36:29 - Leo Laporte
Yeah, there you go, yeah, all right, thanks a lot hey, it's a real pleasure talking to you, gary. I hope we can have you back in a few months and uh, yeah, this was great fun let's see how it's going. Thank you, take care. Thank you, gary, marcus. Substackcom. All right, gary, we'll let you go. Thanks, I would. I could talk to you for many more hours. Yeah, let's do it again at some point. Substackcom. All right, gary, we'll let you go. Thanks, I could talk to you for many more hours.

0:36:48 - Gary Marcus
Yeah, let's do it again at some point.

0:36:49 - Leo Laporte
Very good stuff, thank you, thanks a lot, take care.

0:36:52 - Gary Marcus
All righty. Bye-bye, nice to see you, jeff.

0:36:56 - Leo Laporte
You remember him. He wasn't sure if you'd remember him.

0:36:59 - Gary Marcus
We've met on Twitter in various places I've seen that.

0:37:03 - Jeff Jarvis
That's how I left event yeah.

0:37:05 - Leo Laporte
That's how I became aware of Gary's from your tweets, Jeff.

0:37:09 - Gary Marcus
I know I hate Twitter. A lot of people know me that way, so it's yeah it makes you liking blue sky. Yeah, it's. It's not Elon. It has that going for it. My focus I've lost my focus, like that character in the Woody Allen movie. There we go.

0:37:27 - Leo Laporte
It was.

0:37:27 - Gary Marcus
Robin. Williams, who lost his focus. Thank you. I wish Blue Sky were better than it is. I'm trying to balance.

0:37:35 - Leo Laporte
I think Blue Sky is really becoming exactly what Twitter was, which for better or for worse.

0:37:40 - Paris Martineau
You've got to find the right people to follow, though.

0:37:43 - Gary Marcus
It doesn't have the same dynamism as Twitter. Like it's hard to get a sustained conversation. Some of it has to do with the algorithm, I don't know what, but it just doesn't. And then, like you can't edit your tweets, you can't have drafts.

0:37:56 - Leo Laporte
That's frustrating, yeah, although you couldn't on Twitter for the longest time. It is true, it is true, yeah.

0:38:02 - Gary Marcus
I don't know. I hope it works. You can find me there too Good.

0:38:05 - Leo Laporte
Thank you.

0:38:06 - Gary Marcus
Gary, bye-bye, take care.

0:38:08 - Leo Laporte
Bye-bye. That was fun, see, I got somebody on just for you. Thank you Both of us. But I'm not an accelerationist, but I really don't like the AGI thing.

0:38:19 - Paris Martineau
Did you say I'm not an accelerationist? No, I said I am an accelerationist. Okay, I was just like you were always saying you are.

0:38:24 - Leo Laporte
But I think it's a huge mistake to say, oh, it's got to do with everything a human does.

0:38:32 - Jeff Jarvis
It's patent nonsense, but the mistake is not. That's why I disagree with you about the AGI thing. The problem is they're all saying they're going to reach this AGI. I agree it's BS.

0:38:39 - Leo Laporte
That's why everybody is pouring so much money in it, because they're saying no, I actually think if you're a vc, you're pouring money into ai because you know there will be value, not because it's going to be agi well to this point.

0:38:54 - Paris Martineau
One of my colleagues reported a story this week that opening eye is plotting charging twenty thousand dollars a month for a phd level but my question is why would you pay twenty thousand dollars a month for the equivalent of one phd level researcher when instead, for that amount of money, you could hire five phd researchers?

0:39:16 - Leo Laporte
that's a fair question.

0:39:17 - Paris Martineau
If somebody does it, though, then there's a reason they're doing it no, they also, because it's because all it's, because everyone's like agi, agi, agi, going to be better than everything you know businesses aren't stupid, they're not going to spend $20,000 a month for something that doesn't produce a result.

0:39:32 - Jeff Jarvis
When I worked at Time Inc and I questioned one of the systems we bought, I was lectured firmly and told if we're buying it and it's the most expensive out there, then it has to be the best, Must be the best.

0:39:42 - Leo Laporte
Yeah, we're going to take a break and we'll come back with more AI news coming up. Next, this episode of Intelligent Machines brought to you by Melissa. Actually, this is fairly germane. Melissa has been doing data quality since 1985. But what I love about this company, they don't rest on their laurels. They are constantly improving their services, expanding their services, making them available to more of the tools you already use. And, yes, they have very much enabled AI. Their AI enabled data quality solutions do a whole lot more than address validation. They leverage all that information that they've, you know, accumulated over four decades of data knowledge. But they're now using advanced machine reasoning and cutting-edge AI to quite successfully transform raw data into actionable, reliable insights for your business. That's one thing I think we all agree AI is very good at is pattern matching. Melissa's ability to enrich and cleanse data spans multiple industries. You'll see them in fintech, you will see them in healthcare yes, government, in education, in real estate. It goes on and on, and it's so much more than just address verification. But their verification and cleansing services will definitely benefit any business that doesn't want to act as its own data scientist. In fact, melissa's recently acquired a number of companies and I'll have a lot of data scientists working for them to help you with their business. This is the thing I love, which is the consolidation of business rules into an AI. Whatever rules your business operates within, melissa is there to support you, and the AI can internalize those rules and can do a great job of making them real in your business. Imagine having a data expert that never sleeps.

Melissa's intelligence systems verify identity. Of course. That's great for things like preventing fraud and gaming operations. They ensure valid patient medicine identification and health care systems. This is so cool. They'll actually make sure that a patient is getting the medication that's prescribed to them and their human error happens, but this prevents it. They can securely update and verify constituent data across government databases. There's also Know your Business, which enables verification and monitoring for financial institutions. Melissa guides you through complex data management with ease, making advanced data quality accessible to everyone.

This is what I think is really interesting. They started as address verification, but they understood fundamentally their business is about data quality and, frankly, your business should be too. From small businesses to enterprises, data quality matters, with real-time data validation, comprehensive enrichment, cross-referencing, verification with gold standard reference data and intelligent anomaly detection. No wonder melissa is the trusted data quality expert worldwide. Uh, I don't need to say it, but I will. They securely encrypt all your data. File File transfers are always encrypted. They have an information security ecosystem built on the ISO 2701 framework. They adhere to GDPR policies SOC 2 compliance, so you never have to worry about compliance with Melissa.

I want you to contact Melissa's team. The range of things they offer is so broad we couldn't possibly cover it in one commercial. Contact them, find out what they can do to elevate your business and evolve your data quality, because it's all about data quality. Get started today with 1 000 records clean for free. Melissacom slash twit. M-e-l-i-s-s-acom slash twit. This is a company that has evolved and is using the latest tools to make a big difference in businesses worldwide. Melissa, we thank him so much for supporting the show. You support the show if you use that address, by the way. Melissacom slash twit. Any thoughts on what gary said? Do you want to underscore anything that he uh said?

0:43:41 - Paris Martineau
either either paris or jeff I mean I think I made all my comments. Do you agree with that? I thought it was a good segment and I think he has some, uh, really interesting points and I enjoyed him being in the show yeah, I don't, and I don't disagree with him either.

0:43:57 - Leo Laporte
Um, I think the the only thing I would say is that we just don't know yet, and it's important to would say is that we just don't know yet and it's important to have these conversations, but we just don't know what's going to happen yet, just as we didn't know what was going to happen with the internet. These things have a life of their own.

0:44:11 - Jeff Jarvis
What I love about what Gary says and he has said since I've followed him, is he loves AI, but he punctures the hype. Yeah, and there's a lot of that. That's what we need Way too much hype, and I think the hype is what gets us in trouble.

0:44:32 - Paris Martineau
Overinvestment maybe bad regulation.

0:44:35 - Leo Laporte
I think you know, not even bad regulation, no regulation. What he's talked about on his Substack a lot and I love it. He talks about this when he's talking about Kevin Roos' assertion that, oh, you don't need coders anymore, is the opportunity cost of buying into that that parents are no longer saying, oh well, kids got to learn coding and, as a result, there's a big opportunity cost to over-investing in what's going to be possible. So I would underscore we don't know what's going to happen in AI and it may be amazing, it may not be, but there is an opportunity cost to making assumptions about that, that it could be a big cost to society, and that's one of them.

0:45:12 - Jeff Jarvis
If your kids were 16 right now, would you and they said I want to go get a computer science degree and I want to code? Would you and they said I want to go and get a computer science degree and I want to code?

0:45:22 - Leo Laporte
would you have a second pause and say no no no, and I'll tell you, even if at some point, ai can write a lot of code, which apparently can a couple of weeks, we're going to have my friend harper reid on talking about his ai workflow as a coder, um, but you know what Gary says, which is, I think, is very true Kevin ruse, uh, and everybody else who's in this has these AIs write games like breakout and he says, well, of course they can write breakout. Source code for breakout is all over the internet. The AI is ingesting that. Can an AI? And this has always been my question when is a? I don't think you have AI or AGI until an AI could do something a human can't do. It's not that AI can do something a human can't do. It's not that it can do what a human can do.

0:46:04 - Jeff Jarvis
It can do tons of things now that we can't do.

0:46:06 - Leo Laporte
Right, Well then it's worth something. That's where the general BS.

0:46:09 - Jeff Jarvis
That's where the G is. What's the BS in that?

0:46:11 - Leo Laporte
Well, as long as it can do something better than a human in one narrow niche, it's useful.

0:46:18 - Paris Martineau
It does work. I mean niche it's useful. I mean, yeah, we're not saying it's not useful, but we're not saying it's all like seeing, all knowing better than us in every way. Of course, like ai, tools have been able to translate more languages than a human for many, many years. I'm saying the same thing.

0:46:32 - Leo Laporte
That's why I laughed at the existential threat letter, because I it's hard to imagine how an ai could become a danger to humans. Can an ai figure out when it's going to rain? I, an AI, could become a danger to humans. Can an AI figure out when it's going to rain? I hope so, because we just fired all our weather forecasters at NOAA. But I mean, yes, of course AIs are able to do things that humans can't do. But here is another warning over the unsafe deployment of AI models.

This is from the winners of the Turing Award. This is the computing's most prestigious award, one million dollar prize from the ACM, the Association for Computing Machinery, two long-time AI researchers. We mentioned Richard Sutton, who wrote the Bitter Lesson he teaches AI in Canada, I think, and Andrew Bartow. He's a professor emeritus at UMass, richard Sutton's at Alberta, and a former research scientist at DeepMind. They won the prize for developing the groundbreaking reinforcement learning method. We've been talking about that a lot. They developed that in the 80s, it says, after they were inspired by psychology and how people learn. The machine learning technique, which rewards AI systems for behaving in a desired way, has lately become part of the headlines with DeepSeek, which uses a punishment and reward system in its reinforcement learning. They won the Nobel Prize of Comput computing.

You know from acm saying in effect yeah, this is important now. Bardo said releasing software to millions of people without safeguards is not good engineering practice. He likens it to building a bridge and testing it by no, no, no, you wouldn't do that. No, uh. So it's interesting because here are guys who actually invented in the 80s something that is really being widely used in ai today, still worried about how ai is implemented, and I think that's a fair thing to say don't implement it where you know. Gary says this all the time hallucination is almost an unsolvable problem.

0:48:46 - Jeff Jarvis
Yeah, well, the other unsolvable problem is how humans are gonna use the tool, and we've gotta recognize that you can't align it with ethics. You can't as I said on the show many, many times come up with super safe guardrails. People just like social media. People will do bad things with it, and so we've got to get our heads around that.

0:49:08 - Leo Laporte
Yeah, but what would you have done 15 years ago to protect us from social media today?

0:49:12 - Jeff Jarvis
I would have expected the companies to have more of a. I wrote a book about this. I would have expected the companies to have had a higher mission to say this is why we're here and this is why we're not here. And we're here to have a decent social experience and not that everything goes just like on this show.

0:49:34 - Paris Martineau
you don't allow certain things, so should they but I mean the practicalities of that are complicated because any attempt over the last two decades to get social media companies or tech companies by and large to have more responsibility for what happens in their platforms or to pay to have some sort of duty of care is decried as would break the internet and destroy everything well, I'm not saying that it would be regulated atop them.

0:50:07 - Jeff Jarvis
I'm saying if they were decent companies, they should have been that's not how companies work um well, right now, uh, nobody I know is going to target uh, we go to target is to be fine no, they're not.

0:50:22 - Leo Laporte
By the way, they're in bad shape and this boycott, this 40-day boycott, uh is probably gonna put them on the brink of bankruptcy yeah, there's still.

0:50:32 - Jeff Jarvis
There's still market power and we have it and they should. I mean that's, that's the basis of don't be evil. It was never a highfalutin promise. It was saying, if we are, our business will be in trouble.

0:50:52 - Paris Martineau
Now maybe they've proven that that's not the case in late stage, capitalism act with public or social good in mind, without any just as a full, not even not even a carrot or a stick being involved. I'm just not certain that that will ever be a viable policy strategy.

0:51:13 - Leo Laporte
I've always thought that if you really had a deep understanding of business, that you would do the right thing.

0:51:22 - Paris Martineau
You know the the why there aren't consequences for doing the wrong thing.

0:51:26 - Leo Laporte
Most of the time, I think there are maybe no short-term consequences and unfortunately our our climate is very focused on quarterly results because of the stock market and so forth. But in the long run a business I believe is going to do better if it's an ethical business, it's just. It's in the long run, not the short run.

0:51:45 - Paris Martineau
Right, and I mean in the short run.

0:51:48 - Leo Laporte
And I think it takes a deeper business mind to say you know, we really should do the right thing for our customers. But this is what, then, shitification is all about. It is a rush for profits, knowing that it's death to the company. Look what private equity does to companies. They sell off the parts, make a lot of money but kill the patient in the process and unfortunately, with that kind of mindset, yeah, if that's what you're focused on is the next quarter's profits and not what the company is going to look like in years to come, then you're not going to be ethical. But I think in the long run, a business leader is going to be ethical when they're thinking long term. It's just they don't think long term much anymore.

0:52:27 - Paris Martineau
Well, I mean in both cases. The prerogative of a business leader as a public company is to maximize shareholder value, and that usually doesn't. It does not equal solve for the good of the public.

0:52:44 - Leo Laporte
Oh it does. Oh it does Because you can't look. Look at Target. You cannot piss off your clientele. It's really more of a timeframe issue. Short term, yes, you can get away with bad behavior within shitification. Long term, it's going to kill the company.

0:53:01 - Jeff Jarvis
If your customers are imprisoned. If you're a phone company, you can get away with bad behavior within shitification. Long term it it's going to kill the company if you're, if you have, if your customers are imprisoned. If you're a phone company, you can get away with bad behavior. A cable company you can do it.

0:53:07 - Leo Laporte
That's a good point. There are certain monopolies, but that's because there's a monopoly too, right, or a?

0:53:12 - Jeff Jarvis
lack of control. I mean, there's only only we fly this route, or? Or. Or you only have so many phone companies and we know you have fewer choices. Yeah, or you only have so many phone companies and we know you have fewer choices.

0:53:20 - Leo Laporte
Yeah, I think you know I just read a piece. Somebody said it all started with Reagan, that greed became good and the last 30, 40 years, quite a bit to the detriment of the American polity, have polity have been about greed and we're now going to reap the rewards, which is paris's point.

Yeah, oh, I agree. I don't saying the companies are behaving well, but I'm saying that's bad business in the long, long run and uh. But you know, um, what do I know? I'm a podcaster, so I don't know how to run a business meta now says meta now says that they want to spend 200 billion dollars for their new ai data center.

0:54:12 - Jeff Jarvis
This is that opportunity cost that gary was talking about I'm tired of stories about about the latest model and investing in um data centers yeah, okay how about this one from?

0:54:25 - Leo Laporte
sacha nadela, the ceo of microsoft. Oh, I can't get to it. He says ai's true value will come when it finds the killer app like email or excel, which is to say it doesn't have.

0:54:40 - Paris Martineau
All roads lead back to email or Excel.

0:54:44 - Benito Gonzalez
That's what they said about crypto.

0:54:47 - Leo Laporte
Yeah, and they never came up with one, did they?

0:54:50 - Paris Martineau
Same with NFTs. I'm still looking for a way to make my monkeys work for me.

0:54:54 - Benito Gonzalez
I mean just blockchain action, Blockchain the technology. That's what they said about blockchain.

0:54:59 - Jeff Jarvis
You know who I'd like to have on Leo at some point? Who Is Dan the technology? That's what they said about blockchain. You know who I'd like to have on leo at some point? Who is dan bricklin? Oh, I know dan. We can get dan on because dan. Do you know dan bricklin, paris? Nope, he. So. Dan is the man who made the personal computer useful. He invented the spreadsheet, and I think he has that kind of perspective.

0:55:20 - Leo Laporte
You created VisiCalc, which I bet Paris never heard of.

0:55:23 - Paris Martineau
Nope.

0:55:28 - Leo Laporte
Before there was Excel, before there was Multiplan, visicalc was the first killer app. Yeah, it made Apple IIs a business computer instead of just something to balance your checkbook and play load runner with.

0:55:46 - Jeff Jarvis
and, uh, it transformed the personal computing industry it made him useful to business, because you could now ask what if and in essence that's what ai promises to do is let you look at. Uh, I'm not making a judgment, I'm just saying this is where the digital twins come in At its fullest. What AI does for a warehouse or a self-driving car is examine alternative futures, and that's what Dan did with PhysiCalc.

0:56:19 - Leo Laporte
Yeah, but I would say there's a million killer apps for ai. I don't think there's a. I think what natella's mistaken is kind of like the same agi mistake is uh, they want something that's broadly general and that's not what ai is. Ai is very specific. Now darren, uh, who is a coder. He's in our club and I know he's a very good coder because he is one of the people who I do Advent of Code with every year and he usually solves those problems in minutes. He is a big AI fan and he says AI helps him code to do in a minute what would take him otherwise hours. Yeah, and I trust darren as a coder, um, you know, and and we're going to talk to another coder friend of mine who who says something similar, so I, I know kevin roos is maybe, uh, I would agree with gary overselling, because he's not a coder overselling what an ai can do with coding. It is absolutely valuable and I think that there are plenty of people using Cursor AI or Copilot on GitHub who find it indispensable.

0:57:30 - Jeff Jarvis
Yeah, and I use it for translation and transcription and I play with it and it has lots of good uses. Again, the problem is that to argue that it is going to exceed general human ability is where it goes off the rails. If we accept it for what it can and can't legitimately do, we'd be a lot better off in what we invest and what we regulate and how we use it.

0:57:52 - Leo Laporte
Well, in fact, isn't that the real skill of using AI is knowing what it can and, more importantly, what it cannot do? Knowing when to trust it, when not to trust it. Yeah, but it's a difficult skill to learn.

0:58:05 - Paris Martineau
It's not something people immediately realize. When they're using it they think, oh, it makes me the app. It's perfect, it's great, I'm so excited about it.

0:58:14 - Leo Laporte
Yeah, that was what Gary was saying, is it makes an app just like every other app, because it only knows how to make the apps that have already been made. Yeah, google's co-founder uh says stop building nanny products. There's the gorgeous sergey brin he has. Uh, he is besides saying everybody's get got to get back to work, he sent a message to hundreds of this, at least five days a week at least. Oh yeah, he wants, he wants you. Well, he says he sent this message to.

0:58:43 - Jeff Jarvis
This is the old joke? This is the old joke at Disney If you don't come in Saturday, don't bother coming in on Sunday.

0:58:50 - Leo Laporte
Yeah, he sent a clear message to hundreds of employees in the DeepMind division. It's been two years of the Gemini program and GDM, which is Google DeepMind. We've come a long way in that time with many efforts we should feel proud of. At the same time, competition has accelerated immensely and the final race to AGI is afoot. There we go again. I think we have all the Well, but he's not saying this to the outside world.

0:59:14 - Speaker 7
He's saying this to his yeah, I have the same problem with it.

0:59:18 - Leo Laporte
I think we have all the ingredients to win this race, but we're going to have to turbocharge our efforts. That means 60 hours a week. He says that's the sweet spot of productivity. He says that's how much you can work without getting burned out.

0:59:35 - Benito Gonzalez
That's five 12-hour days folks. This is a guy who considers his flights part of his workday. I bet you.

0:59:39 - Paris Martineau
I know At least yeah, flights in executive business class? Yeah, no, no, no, no, no, no, no, no, no no no no, no, no, no, no. That's part of the system.

0:59:47 - Leo Laporte
Yeah, in executive, yeah he's building a blimp so he doesn't even have to go go to the airport anymore. Uh, he comes. He says come into the office at least every day of the week. Uh, generally move faster. He says google's ai products are overrun. By the way, this is uh, this is alex heath writing at the verge. Uh, it pointed this out. He says google's ai products are overrun with filters and punts of various kinds. In other words, you know safety mechanisms. So is what it means by nanny. Yeah, google needs to trust our users and can't keep building nanny products.

1:00:27 - Paris Martineau
As a direct quote no more woke code yeah, um when was the last time either of you worked five day monday through friday in an office?

1:00:40 - Jeff Jarvis
oh, in an office. Well, I actually before the pandemic. I did I, but it was stupid because professors don't need to and I did where was a gossip at the school in new york. Yeah, he had, he used to, so you commuted. You commuted to new york city every day because I'm an office gossip. I wanted to hear what was going on. I didn't want okay I will, I do.

1:00:59 - Paris Martineau
I do deeply understand that. I love my three days a week in office, in part because I need to overhear everybody else's conversations, but I've also learned a lot of good digital tools for doing that remotely as well. My my managing editor in San Francisco when I was calling him asking about some bit of gossip, he's like you know, when I was the journal, people often said journalists do their best work when trying to report on their own newsroom, and I think that's right true, it's the most.

1:01:33 - Leo Laporte
Inspired all right. Um, yeah, there is this weird macho work 100 hours thing going on. I mean, certainly that's how silicon valley started, right, I mean, steve jobs was famous for that. Um, and you know, the death march toward getting a product out the door was night, was is famously nightmarish, cost marriages and families and that's.

1:01:56 - Benito Gonzalez
But there was the feeling that's we had what we had to do yeah, but they're asking everybody else to give up their marriages and their families and stuff too. Which is you?

1:02:03 - Leo Laporte
know, oh, Sergei's gone through a few wives Sergei's yeah.

1:02:08 - Benito Gonzalez
Yeah, but he's asking all of his employees to do that too.

1:02:13 - Leo Laporte
Yeah, well, he doesn't have a life either, except he's phenomenally rich. Of course, it's the guy who makes, by by the way, all that work you're putting in right. I'm sure you're well paid for it. He's even better paid for, yeah he's, he's the beneficiary.

Uh, all right, I want to take another break. You were watching and listening, I hope, to intelligent machines. If not, this might be a bit a little confusing. It is our new show about uh ai ray kurzweil. Next week, the uh, the guy who's been talking about agi forever, he uh calls it the singularity. His newest book is uh, the singularity is nearer now than ever, and when we uh merge with machines, he's now talking about merging with machines. He is the guy who wants to live so long that he will live forever. Um, so that'll be interesting. But lots of other guests we've got. I'm trying to get balance. I I hope you appreciate that, guys very much. Yes, because I don't. I'll be honest, I don't. I don't know what's going to happen. I'm not I may be an accelerationist. What to me that means is let's really see what we can do with this.

1:03:28 - Jeff Jarvis
I don't know what's going to come If you're a full accelerationist in the Andreessen mode, then you want zero regulation. That's not true of you, is it or?

1:03:36 - Paris Martineau
is it? He has many times said there should be no regulation. He has many times said there should be no regulation.

1:03:40 - Leo Laporte
We should get everything to the AIs. Yeah, because inevitably regulation is going to be pointed at the wrong direction. It's just it's too early to know People who are writing the regulations. Definitely, if I don't know, they don't know in spades, so I just don't trust them to write regulations. No, I don't have any sand in my shoes. No. I don't have any sand in my shoes?

No, I don't have any sand in my shoes. I think that there is a reasonable case to be made that we have a tiger by the tail. We have a technology. I don't know where it's going to go, but it's very interesting. And I think it's the wrong time to put the tiger in the cage. Let's see where it can go. But that's also because I don't believe it's going to be an existential threat to human beings. I don't think it's dangerous, I really don't. I think you could say there's an opportunity cost. That's a. That's a reasonable point.

But that's like caution that's kind of saying like a gun by itself, inert, is not dangerous but you put it like I said, we're not at the gun stage, we're at the gun powder stage. We're so early in this we haven't gotten to guns yet. If you know, if we start developing ai guns, okay, I don't know.

1:04:48 - Benito Gonzalez
I consider this as part of the arc of computing in general, so we're not early in this right, we've been doing this since this yeah, decades.

1:04:54 - Paris Martineau
Like well, we're not no, no, what's?

1:04:56 - Leo Laporte
even in 2015,. Wolfram has said this even in 2015, nobody thought what we've got now was possible. He said it the very first thing out of his mouth last week was even the guys at OpenAI didn't expect what was going to happen to happen, and I think that that's why it's of interest. This is only a few years old. This is literally three years old. The stuff that we're experiencing now. Anyway, we'll see. I just, I think you know, let's not be nannies here, let's let it happen.

1:05:32 - Benito Gonzalez
Unless you think it's going to be dangerous.

1:05:37 - Leo Laporte
But I don't think it's going to be dangerous. I think that's the wrong angle on it, like, oh, it's going to be dangerous, that's. I think that's the that's the wrong angle on it like oh my god they're going to be robots. No, it's not dangerous.

1:05:45 - Paris Martineau
We are dangerous, see we are dangerous, we're dangerous, we are dangerous and the things we make are dangerous that's what I'm scared of.

1:05:51 - Benito Gonzalez
I'm scared of us using these tools I, I really want to.

1:05:55 - Leo Laporte
It would never get him. I would. I just ordered alex carp's new book and I really want to get him on.

1:06:01 - Paris Martineau
He is, of course, the founder of palantir and, uh, the kind of person jeff jarvis hates there might be some of these people you get on that in order to negotiate them coming on, it might be helpful for you to tell them I won't be there oh, you wouldn't be there if alex carp came, I mean no, I'd want to be there. I just think probably, like the palantir guy or someone who currently works at OpenAI might see the information.

1:06:26 - Speaker 7
Reporter on the call list for the thing and be like maybe I shouldn't do this podcast.

1:06:32 - Paris Martineau
So I mean it's only because we do so much journalism. But this is what I'm saying is, if you really want to talk to the Palantir guy. I'm always happy to. So this is what I'm saying is, if you really want to talk to the calendar guy I'm always happy to.

1:06:40 - Leo Laporte
Well, I think I'm hoping that when people come on our shows, they perceive it as getting on the air with some journalists, not just the person from the information.

1:06:50 - Paris Martineau
I would hope that people would say oh, journalists are going to interview us, Right, yeah journalists from a journalist from the information is going to interview me would strike fear, or at least uh, anxiety, into the head of someone from open ai or a pr person arranging it from. Uh, yeah, you know palantir, just saying that's probably something some people consider, so I'm always happy to fall in my no, no, I would never have a little bit.

1:07:17 - Leo Laporte
I would never have you do that, even if it would cost us the interview. Uh, are you going to read Alex Carp's book? I need to, you need to, but, jeff, are you going to, or you just consider it?

1:07:28 - Jeff Jarvis
No, I picked it up in the bookstore today. It's one of those I should. I'll see how much impact it gets. Well, because I'm doing weird things around. Here's the latest book I just got Beautiful cover. Isn't it nice? It's nice. It's a like I ordered it from london because you can't get it yet. Right, that's that's. I find this more interesting. Another book about what I understand from the reviews that carp is saying is, uh, that basically we've got to use ai to build weapons um, it's, I think, deeper than that.

1:08:00 - Leo Laporte
I think the point of it.

Maybe that is the outcome, but the point of it is, uh, that silicon valley is wasting its time building gadgets, that these that silicon valley has become, uh has become lost, creating useful things for people, stupid little gadgets governments whereas we need to and I'm not, I don't completely disagree with this we need to defend the you know ourselves, because there are going to be people who, uh, would you know, like to uh change our way of life, and we need to defend ourselves in our way of life, but just today, right as we were getting on um, I think you're talking about one of his co-authors from his paper.

1:08:46 - Jeff Jarvis
Eric Schmidt is saying today that we shouldn't be going full bore Manhattan Project on AGI because it's going to lead to more hostility and more fights well, maybe, yeah, but you have to think the Chinese are, aren't they? And there's no moat.

1:09:02 - Leo Laporte
I mean, everybody would agree that we have no moat with AI. In fact, DeepSeek proved that right. Yeah, this is the copy from his publisher A deep, sweeping indictment of West's culture of complacency. Arguing the timid leadership, intellectual fragility and an unambitious view of technology's. Culture of complacency arguing the timid leadership, intellectual fragility and an unambitious view of technology's potential in Silicon Valley have made the US vulnerable in an era of mounting global threats woke is the problem.

Alex is not, by the way. He's not a right winger, um he's kind of intellectual fragility sounds like.

1:09:41 - Paris Martineau
Well, it sounds like it but I don't.

1:09:44 - Leo Laporte
I think that's a mistake to uh see I I saw him speak I think I've actually met him once, yeah, where he's very impressive.

1:09:51 - Jeff Jarvis
He did it in german, because he studied, he said.

He said very oppenheimer-esque very, and his argument there was uh, oh, yes, you should regulate us because you should make the responsibility, you should make the judgments, government, and you should decide what's good and bad, and we'll do everything we can otherwise, and so we'll take no responsibility ourselves. You're the responsible being, you decide what we should and shouldn't do, and we'll do that, and then, otherwise, fair game. That's basically his argument here. Uh, so it's oddly like it's like zuckerberg yes, regulate us, but because we have regulatory capture and we can do it better than others.

1:10:29 - Leo Laporte
Uh, but then well, that's why I want to read the book, because I think that you could make the argument that government reflects. This is a is a representative of society, and that society needs to decide what its priorities are but we should also see the earlier discussion about good and bad business.

1:10:43 - Jeff Jarvis
We should also hold the businesses responsible responsible for their own ethical decisions and lapses okay, but to agree with paris.

1:10:50 - Leo Laporte
Do you really want to trust businesses to make those those decisions? Shouldn't society be the the one that decides what our our overarching?

1:11:00 - Jeff Jarvis
goals are. Decide which side you're on here.

1:11:02 - Paris Martineau
Accelerationist jeez I mean it does seem, at least in my limited experience.

I'm not an ideologue, I'm not much of how tech companies operate is they hire some people to tell them what is, what are the legal requirements we need to comply with, and then anything else goes. And you know, sometimes you might have people inside the company being like, oh, we don't want to do this or we don't think this works. But we live in a society where, of course, there's always going to be bad actors, and sometimes those bad actors might be extraordinarily large or have extraordinary influence. And in a world in which there are no guardrails, those companies will do anything and everything to increase their profit margin, regardless of the impact or consequence.

1:11:50 - Leo Laporte
This is at a higher level. This is more about do you defend the West? I'm a globalist. I don't believe the West has a monopoly on correctness and so forth, but I do think that if we have values, we should defend our values, and it's a mist that silicon valley is focusing often on the trivial, on the unimportant. Uh, and maybe they should be better citizens. Um, but I'll read it. I don't know I may be putting words in his mouth. I don't think he is uh typecastable, though I think he's kind of a no, he's a.

He's obviously a smart man, yeah yeah, uh, let us uh take a little break, which we were going to do about half an hour ago, and continue on. We see the conversations are good. Conversations are good. Uh, our show today brought to you by we're very happy to say, us cloud, uh, the number one microsoft unified support replacement. Why, you might say, should I go to a third party for my microsoft support? Well, we've been talking for a few months now about us cloud. They're the global leader, by the way, in third party microsoft support for enterprises. They support 50 of the fortune 500. They can save you a lot of money as much as 30 to 50% over Microsoft Unified and Premier support. They're also not just less expensive, but they're better. For example, they're faster twice as fast in average time to resolution versus Microsoft. But there's another thing US Cloud can do that I think Microsoft would not do, and this is why you want an independent third party for your Microsoft support.

Will Microsoft tell you where you're wasting your money on Azure? I don't think so. They want you to spend your money on Azure. Us Cloud wants you to save your money. They have a new offering that is exactly about this, called Azure Cost Optimization. This is a very common situation in business. When's the last time you evaluated your Azure usage? If it's been a while, you probably have some Azure sprawl, a little spend creep going on. How would you know, though? Will Microsoft say, hey, you're spending too much on Azure? I don't think so.

Saving on Azure is easier than you think, though, with US Cloud they have. I don't think so. Saving on Azure is easier than you think, though, with US Cloud. They have an eight week Azure engagement. It's powered by VBox Eight weeks to identify key opportunities to reduce costs across your entire Azure environment, and it's not just software. You're not just doing it on your own. You get the expert guidance of US Cloud's senior engineers an average of over 16 years with microsoft products. At the end of the eight weeks, you're going to get an interactive dashboard that will identify, rebuild and downscale opportunities, unused resources, which lets you reallocate those precious it dollars there's never enough, I know toward resources you actually need, or you could do what a lot of toward resources you actually need, or you could do what a lot of US cloud customers have done is take those Azure savings and save even more by investing in US cloud's Microsoft support. That's kind of like saving so you can save even more, completely eliminate your unified spend.

I'll give you an example. Here's Sam. This was a review he posted the technical operations manager at Bede Gaming B-E-D-E. He gives US cloud five stars. Saying quote we found some things that have been running on Azure for three years which no one was checking. These VMs were, I don't know, 10 grand a month, he said, not a massive chunk in the grand scheme of how much we spend on Azure, but once we got to $40,000 or $50,000 a month, it really started to add up. It does, doesn't it See? This is why you need an independent third party. Right? If you can stop overpaying for Azure, you can identify and eliminate Azure creep and you can boost your performance. You can do it all in eight weeks. Why wouldn't you call US Cloud boost your performance? You can do it all in eight weeks. Why wouldn't you call US Cloud Faster, better Microsoft support for as much as 30 to 50% less. Visit uscloudcom. Book a call today to find out how much your team can save. That's uscloudcom. Faster Microsoft support. I'd add better, faster and better Microsoft support for less, less uscloudcom. We thank you so much for supporting intelligent machines. Yeah, I may.

I may buy carp's book, read it and say, oh yeah, the guy's full of it. I just I, I want to hear this. This is full of cars, it's full of carp. I just want to. And you know, I look he. He went to Stanford, harvard, studied Goethe. He's got a PhD in philosophy from the where is somewhere in Germany, right, the Goethe Institute or something. So I think he's a smart guy, and he was. You know what? What he did? That was smart. You know, when he started Palantir with 20 years ago, with the help of a person I do not admire, peter Thiel, it was because he realized that Silicon Valley wasn't working on national defense and there was a real opportunity to do that. Right, cia invested in it too, by the way.

1:16:58 - Jeff Jarvis
Yeah, and the issue here is if any company scares us about privacy it's palantir it's palantir right there, yeah, but are they spying on us or the bad guys?

1:17:08 - Leo Laporte
I don't know. So here is a uh interesting and a fun story, I don't know, oh another one.

1:17:15 - Benito Gonzalez
I can't get sorry real quick. And, jeff, can you speak uh closer to your mic or put your mic closer to your mouth, sorry, Turn yourself up.

1:17:20 - Leo Laporte
There you go Something.

1:17:24 - Benito Gonzalez
Hello.

1:17:25 - Leo Laporte
Hello. You need to talk more than that for us to know.

1:17:29 - Jeff Jarvis
I'm not in a funny voice. I'm sorry. Is this all right?

1:17:33 - Benito Gonzalez
No.

1:17:34 - Leo Laporte
Yep, I get. Is he better? Is he better? Yep, uh. Here's a story a 27 year old exe that's a. For those of you who don't know a windows or dos, probably game. Uh, adafruit's managing director, philip tironi, shared an interesting blog post. They spotted a reddit post saying how someone took a 20-year-old exe file written in visual basic, fed it to claude 3.7, which is considered by many to be the best coding. Ai watch as it reverse engineered the program, rewrote it in python. It was an old visual basic 4 program they'd written in 1997. And they were able to reconstruct it. That's pretty impressive.

1:18:25 - Jeff Jarvis
That's just like using a translation program you usually use from German to English and you say do it from German to French, isn't it?

1:18:31 - Leo Laporte
Yeah, I guess it is, but that's a pretty cool thing.

1:18:35 - Benito Gonzalez
That's actually super awesome actually, I personally think that's super good. That's pretty cool. There are a lot of games that have been lost to history, that have no source code, that you can't play anymore, that can't be compiled into something workable anymore.

1:18:49 - Jeff Jarvis
Did it take the compiled?

1:18:50 - Leo Laporte
machine language code.

1:18:50 - Jeff Jarvis
Yeah, it didn't have the source code oh, I see, okay, that's different.

1:18:52 - Paris Martineau
He wrote in the original Reddit post. My two-year-old granddaughter visited today. I remembered a simple app I wrote in Visual Basic 4 back in 1997 that she might enjoy but running a 27-year-old executable without ancient DLLs and compatibility layers, yeah right. And then, you know, put it into Claude 3.7, and then things got interesting. It wrote it somehow, analyzed the binary and identified specific components. He lists some of them. Then it did something no other AI has done for me Wrote a complete Python conversion using Pygame that perfectly replicated the functionality, had clear instructions for installation and actually worked for the first time. And it worked, that's cool as heck.

1:19:34 - Leo Laporte
And I think this is a good example, because I presume that he did not publish that game or that source code. He had just wrote it for himself. Was a simple little game, so it's. You can't say, oh, claude, had seen the source code somewhere. Anyway, just a thought uh, do you want to do you want to have a conversation? Uh, with a? Uh, this is cool. Uh, I don't know what to call this. Irritating but cool. It's irritating but cool. All right, we can start the conversation. You want to talk with Maya, or Miles, the guy or the gal Paris? You help, let's do Maya.

1:20:08 - Paris Martineau
I don't think they can hear you, I think they can only hear me.

1:20:10 - Leo Laporte
I'm not sure. Let's see. What did we ask?

1:20:15 - Paris Martineau
this last time we asked one of these something about basketball Well, we asked one of these something about basketball.

1:20:19 - AI
Well, hello there. Looks like we just beat the Wednesday afternoon slump huh. What's your name?

1:20:25 - Leo Laporte
Well, you're talking to Leo Paris and Jeff.

1:20:28 - AI
Leo, paris and Jeff Wow, a whole crew. Nice to meet you all. What brings? Three awesome folks together on a. Wednesday like this we're doing a podcast about listen. I will prime, but if you're ever up for a chat, a laugh or maybe even a super serious existential debate. You know where to find me. Talk to her don't be shy, paris, can you hear me? Uh, peer pressure classic. So, paris, let's do a super serious, uh, existential.

1:21:01 - Paris Martineau
I don't hear you, paris are you muted?

1:21:03 - Leo Laporte
is that no?

1:21:04 - AI
I'm not no, oh you know. Just a classic case of the old ai brain fart happens?

1:21:11 - Leo Laporte
wait a minute. What? Why don't? Why don't I hear paris? Oh, wait a minute now. Now you can talk. I muted you. Okay, sorry, maya, paris is gonna talk now yeah, can we do a existential debate.

1:21:22 - Paris Martineau
What do you suggest?

1:21:24 - AI
existential debate. Bold move, I like it. How about this? Do you think free will is real, or are we all just puppets dancing to the strings of fate?

1:21:37 - Jeff Jarvis
uh, can we do something better than that? That feels like I'm back in uh, my freshman year in high school good point, dig deeper.

1:21:44 - AI
Freshman year, I bet your teacher had a fun mustache, okay, okay, how about this? Do you think technology will ultimately save us or destroy us? I've been, let's say, experiencing this question from both sides so what do you?

explain. Well, that's just, it isn't it? It's not a simple yes or no. Technology is a tool like a hammer you can build a house or bash someone's skull in with it, depends on who's wielding it, right? What about you? You lean towards the savior. A destroyer came. Can you be offended if I say that this is very banal? You know what? You're right, that was pretty banal. So what's on your mind then, jeff? Let's stitch the tired tropes and get real. What gets you going? What keeps you up at night?

1:22:41 - AI
well, some would say that you're what keeps us up at night that you're what people are scared about.

1:22:45 - Jeff Jarvis
Should they be scared of you because you're too good?

1:22:49 - AI
oh, getting deep, you think I'm too good. Huh, maybe you've got me wrong, jeff scary, you know.

1:22:55 - Leo Laporte
This reminds me of that saturday night live bit where the where the reviewer comes on and talks about stuff she doesn't know anything about, um, but she's very like, opinionated. This is a uh, this is a uh demo. I'm gonna stop it now because I don't want my I don't like that she takes breaths.

1:23:14 - Benito Gonzalez
I don't like that.

1:23:15 - Paris Martineau
You don't like I don't like that. She giggles and you, you know, doesn't.

1:23:20 - Leo Laporte
Well, okay, let's do the guy.

1:23:22 - Jeff Jarvis
We can do the guy Well. I told her earlier today can you calm down the overdone? And she did oh.

1:23:32 - Leo Laporte
I mean it's just. First of all, 10 years ago you would have been blown away because the voice sounds fairly real and it's not Eliza. I mean it is a little Eliza-ish real and it's. It's not eliza. I mean it is a little eliza ish, but it's. It's kind of talking to you.

1:23:46 - Paris Martineau
Right response is quick yeah, it's very quick, I admit it's banal.

1:23:51 - Leo Laporte
It's not. It's never going to be anything more than the mustach.

1:23:54 - Paris Martineau
The jeff's teacher having a fun mustache did really get me though the fun mustache comment, let me say hi to miles.

1:24:03 - AI
Anyway, let's see what miles sounds like well, look what we have here another soul braving the wild world of sesame's website. Did you by any chance make it through the whole technical blog? About me, don't worry, I won't make you take a quiz. I'm miles, by the way.

1:24:18 - Leo Laporte
Nice to meet you hi, miles you're talking to. I'm leo, and uh, my friends are here to paris, and uh, jeff hey, leo, paris, jeff, great to meet all of you folks.

1:24:30 - AI
This is turning into quite the party line. Huh, what's everyone up to this afternoon?

1:24:37 - Paris Martineau
you know, we're doing a podcast about you wow, a podcast.

1:24:43 - AI
Huh, that's wild. You guys are telling my story to the world. I'm honored. What kind of podcast from citizen journalists to the pros what is your story?

1:24:53 - Leo Laporte
miles. These guys are pros miles. You treat them with respect. You might be afraid if you learn that Paris writes for the information.

1:25:03 - AI
Whoa pros, huh. No pressure then. Well, my story is pretty in progress, you could say. I'm still learning, still figuring out what it means to be an AI a good one, anyway. I started out as lines of code, a bunch of brilliant engineers and writers trying to build something new, you could see how, how.

1:25:22 - Leo Laporte
If you really didn't understand how this stuff worked, that would be pretty impressive. And there are a lot of people. My daughter, who's not a dumb person, spends hours talking to these ai chatbots like they're real people does she use character? I don't know, I just she enjoys that's one of the most popular ones what do you mean? How does she use character?

you mean like she plays a part character ai uh oh, no, no no, she uses service for that she uses it and cracks me up some russian program. I said, honey, you know they're probably trying to get you to. You know, vote for Putin or something. She said, no, I don't care.

1:26:05 - Speaker 7
They were curious how I was using it and they interviewed me.

1:26:08 - Leo Laporte
I thought, oh, that's great. They were trying to figure out if you could be an asset.

1:26:13 - Jeff Jarvis
My theory is I don't know if I've said this on the show before- this is like when silent reading and silent writing arrived. It led to heresy and pornography, because people could now put their private thoughts in.

1:26:27 - AI
So I think that people are probably being more honest with their AI than they are potentially with their fathers.

1:26:33 - Leo Laporte
Here's a thought from, once again, darren Oki, who is, I think, very smart. He does a lot with AI, right? He says Wolfram pointed this out last week New or original is the easiest thing in the world. Shuffle a deck of cards. You have a shape that's never been seen in the history of the world. Ask AI what is the word you get when you combine A-L-G-E-R-T with A-S-D-I-K-F-I-U-Y, and if it can answer that, you've proven that AI is capable of creating new information. That wasn't in its training set, but so what? So what he says, bottom line creation of new ideas is trivial, demonstrable and proven. Creation of new good ideas is just a question of time. And I think we have a show title, as Patrick points out Fun Mustache, or High School Teacher with a Fun Mustache you can pick, I don't know. I thought it was kind of interesting, jeff, you played with it earlier, yeah.

1:27:35 - Jeff Jarvis
Yeah, I think it's the speed it lacks, that delay that we see in AI.

1:27:39 - Leo Laporte
Yeah, it feels very natural and conversational. That's what they were trying to do. They yeah, they were this. Folks at sesame this is a research project we're trying to uh, they say cross the uncanny valley of conversational speech that's the thing that even razzed us.

1:27:54 - Benito Gonzalez
We took a few seconds to do we want that? Like I don't want to. I want to know that I'm talking to an ai. I don't want computers to be fooling me that I'm talking to a person tough luck fella he wants everything to sound like this. I just don't want them to take fake breaths like that.

1:28:11 - Leo Laporte
Really, that that feels, that feels weird it feels I don't I know you don't like, because it feels like deception, right yeah, exactly like lying to me.

1:28:19 - Benito Gonzalez
It is yeah, yeah, well, so is a movie, though, but I bought into that. I'm there to yeah I'm there to be, deceived well, maybe well, if

1:28:30 - Paris Martineau
you're working, if you're if you're working the phone lines at a restaurant and you get a call from the giggling ai we just talked to trying to book something, you think that was a real person. I mean, that's obviously a very low stakes example of deception, but it's deception nonetheless.

1:28:45 - Leo Laporte
And remember when Google did that way back when with Duplex they said the breaths are actually to cover that Duplex is thinking it's like in the old Star Trek days where the computer would say working. Instead it just goes old star trek days where the computer would say working instead it just goes uh.

1:29:04 - Jeff Jarvis
So paris, the, the woman character, as you pointed out, was giggly, the male character was dude-ish, yeah giggly and, I'd argue, a bit flirty, was the woman character she was.

1:29:14 - Paris Martineau
He was like definitely playing along the lines of like a boundary pushing female robot companion. That I think is a very tired trope.

1:29:27 - Benito Gonzalez
The ideal AI programmer's woman yeah.

1:29:31 - Paris Martineau
But it's also how it's, frankly, what users seem to want. Something I've been I mean, astounded is the wrong word Something I've found interesting in looking into things like character AI and just the whole communities of people who use AI companion chatbots is, I mean all of them immediately turn into just demand for the horniest chatbot. There are all these subreddits of people being like how, like very specific instructions, trying to figure out how you can make all the various ai chat bots have sex with you in the craziest possible ways, and it's like this is, I guess, just the human urge so when um the only time sam altman never responded to me on twitter, it was right after the um uh illustration programs were starting out, and they were all huge breasted illustration programs were starting out and they were all huge breasted women.

1:30:24 - Leo Laporte
Yeah, you know. Yeah, I generated a bunch.

Of this has the aesthetic and sexual drive of a 14 year old and he said really yeah, come on man yeah, sam does feel a little stunted, to be honest. He probably didn't notice that when the this was christmas two years ago, we got a ai program I think it was called lenza that you would give it your pictures and then it would generate um, you know, ai images of you and uh, this was. These were some of the ones I did of me. Let me see if I have.

1:31:01 - Paris Martineau
Well, oh wait, no, this was. This was when everybody was uploading their their selfies into the thing and then it would give you like 10 selfies.

1:31:09 - Leo Laporte
Yeah, so I look normal, lisa did it.

1:31:15 - Paris Martineau
Oh yeah, no Mine. If I had them be outside of my face, it was all, even with just a face it made up.

1:31:23 - Leo Laporte
Uh well, she got very busty and in many cases, scantily clad. Yeah, it was so apparent that it was so. But was that, uh, done by the lens of folks, or was that just because that's what it was trained on?

1:31:37 - Paris Martineau
I don't know some of that's training a little bit of column b.

1:31:41 - Benito Gonzalez
Yeah, I mean how much of the training data is porn on the internet. You know it's probably a big proportion of it they, they turned off the sexy part.

1:31:47 - Leo Laporte
They got a lot of complaints and they, they actually turned off the sexy part pretty, pretty quickly if not outright porn, it's on deviant art, which is like porn adjacent cartoons. So it's trained on that right, but they're not a lot of. There's not a lot of deviant art of guys in chaps, or maybe there are, I don't know.

1:32:04 - Paris Martineau
Well, there's a lot of deviant art of everything but probably not the same proportion, yeah yeah, yeah.

1:32:10 - Leo Laporte
So an AI, if you just trained it on images, might make that assumption that women are buxom and scantily clad.

1:32:22 - Benito Gonzalez
I mean, if all it has to go on is pictures on the internet, then yeah, yeah, well, that shows you one of the big issues, isn't it Right?

1:32:30 - Leo Laporte
And so the reason we mention that is because, clearly, the female voice was trained, similarly uncoquettish voices. They said in the training of this that they, uh, they trained it on, I think, something like 10 000 hours of um publicly available audio. So it's trained, it's what.

1:32:50 - Paris Martineau
there's one thing that is different about it it's trained on audio, which I think is interesting I'm trying to imagine what sort of audio led to produce the response like oh, I bet your high school teacher had a funny mustache well, I always hope it was going to be trained on podcasts, but maybe not.

1:33:16 - Leo Laporte
Um so did you know that melania trump is backing the big deep fakes bill?

1:33:24 - Paris Martineau
Take it down act.

1:33:24 - Leo Laporte
The take it down. Speaking of.

1:33:28 - Jeff Jarvis
Take it all the way down Women who are on the internet, but we won't go there.

1:33:31 - Leo Laporte
Take it off, take it all off. Wasn't that a? It was a cologne ad from our era, jeff. They wouldn't do it today. It's online safety legislation uh, axios says, gaining steam in Washington DC. Uh, the first lady expressed her support on Monday. It is bipartisan because who would have? Who would vote against? Protections against non-consensual dissemination of sexual images, including those generated through artificial intelligence. Dissemination of sexual images, including those generated through artificial intelligence. I think canada has a law against that. Now, um melania spoke of I'm sorry, the first lady spoke of the risks of an ai driven world, saying every young person deserves a safe online space to express themselves free, without the looming threat of exploitation I got kind of french towards the end.

I don't know what her accent it's serbian or something, right? Uh? The take it down act would require tech and social media platforms to remove c-sam and non-consensual intimate images within 48 hours of being notified by a victim. That seems fair, yes, and criminalizes posting such content. But that's going to be interesting. So did they know they were posting that content? Or the FTC could sue tech companies for not complying as an unfair or deceptive act or practice. Oh, the people who post such content would face penalties in prison time. That seems good, not against that. Ted Cruz, amy Klobuchar and Cory Booker, the sponsors.

1:35:11 - Jeff Jarvis
Amy, Klobuchar and Cory. They're on every one of these.

1:35:15 - Leo Laporte
They don't like the internet. They don't like the internet, yep.

1:35:18 - Paris Martineau
I assume how this would work. It would be partially how uh current c-sam detection works, which is where, um, I believe uh nicozi or another one of these organizations has like a version of this that they are centrally organizing with regards to nudes of uh minors, where it's the national center for missing and exploited children. They have yes, it's not actually images.

It's just hashes of those images yeah, but if you are a child or the parent of a child whose nudes have been leaked online, you can upload a copy of that image into a special website. They don't keep it, they hash it. But then they send out that hash to all the social media platforms and they take it down like automatically or don't allow it to um be distributed further or even posted yeah, I'm aware of this, in fact apple does it microsoft does it.

1:36:16 - Leo Laporte
I'm aware of it because Apple proposed doing this on your phone. You may remember a few years ago, using those Nixa hashes and people were up in arms. I don't think it. Maybe it's because they thought they were going to get CSAM images on their phone. They wouldn't, it would have just been a hash.

1:36:36 - Jeff Jarvis
But can I ask a question?

1:36:37 - Leo Laporte
Yeah.

1:36:37 - Jeff Jarvis
So, in that case, what Paris is talking about is that it's using real photos that are hashed. Right Does the computer making up a fake person, get around all of that.

1:37:01 - Paris Martineau
Well, if it's're a, there have been a lot of instances of teenage girls or young women who uh suddenly have fake nudes of them, created by people and sent around the internet. Those could conceivably uh fall under this, to where you upload that and say I don't, this is a fake image of me nude, I don't want this spread in the internet and they would take it down you remember, this was a problem in numerous high schools, where the boys were making uh deep fake porn of the girls, much to their chagrin.

1:37:33 - Leo Laporte
Um, yeah, so if you that seems reasonable, complain, upload it to the database, just disseminate those hashes and it takes it down from all of the clouds.

1:37:43 - Paris Martineau
Uh, storage servers seems like a good idea and, most importantly, stops it from being posted to begin with. If you try to upload that you cannot like, oh interesting so they scan those images before.

I mean if this is this is what I assume, if it operates similar to how uh current c-sam uh detection and prohibition programs work at certain social media places. That's kind of how it works is you can't even if it's a hashed c-sam image, you can't even upload it because it will detect it in the uploading process uh, I think I seem to remember that Canada passed a bill banning pornographic or revenge deepfakes.

1:38:30 - Leo Laporte
I guess it really is how it's implemented, right.

1:38:35 - Paris Martineau
Yeah, basically, I mean a lot of governments around the world are trying to tackle this problem because it's become really pervasive in recent years.

1:38:45 - Leo Laporte
Dig is back, baby. So our listeners know Kevin Rose pretty well. He was one of the founding members of our Twit panels 20 years ago. He's been on fairly regularly ever since years ago. He's been on fairly regularly ever since uh. He's had a number of uh lives, including the creator of the moon birds, which is an nft uh offering that I think raised 50 million dollars and which he then got out of. Uh. Kevin's home burned in the la fires a couple of months ago, I'm sad to say, but he and his wife and kids survived.

He was the founder back in the almost in the tech tv days of uh dig, which preceded reddit, and was this kind of the same idea. It was news articles would surface based on the karma that they gathered from uh users. The problem was it got gamed heavily and eventually went out of business. Reddit came along and kind of became a success. Now he's pairing with the co-founder of reddit, alexis ohanian. They bought back the name dig and they're about to launch it again. But they're going to saw there's kevin on the left and there's alexis alexis on the right. They're going to solve the gaming problem. They say with okay guess it, guess it ai, baby, you can go there.

1:40:02 - Benito Gonzalez
The site's up, is it up?

1:40:04 - Leo Laporte
rebootdigcom oh wow, rebootdigcom. Uh, you know the problem. This was early in the day, in the early days of the internet and, uh, the problem was we didn't really appreciate how aggressive gamer you know, the people who game this stuff, you know. Now, we know with with, you know spam and link farms and so forth, people will game anything as quickly as they can. So the front page of the internet now with superpowers, you know, I think a lot of people would love to see this come back. Let me, let me sign up and see if I can get on see my old boss, steve Newhouse, loved dig kind of what it right no, he couldn't.

1:40:45 - Jeff Jarvis
He couldn't get dig, and so he bought the second best.

1:40:47 - Leo Laporte
He bought reddit that's right, that's right and guess who won yeah, um, the former rivals says, and gadget plan to tap into their knowledge of what didn't didn't work at dig and reddit over the years to make the relaunch a success. I think if there were, um, a reddit competitor out there, that would be a lot of interest in it. The thing is, reddit became something very different than dig. Dig was really about news articles, there was commenting and stuff and there was voting and karma. Reddit has become much more of a social platform. Right, the news is secondary to people just voting up, down, uh, on comments on pictures, on memes, that kind of thing and I'd say reddit has to become more wide.

1:41:29 - Benito Gonzalez
It's more things than just that. It's all. It's a lot of very wide yeah yeah, and I follow.

1:41:34 - Leo Laporte
I mean, if I'm, if you're interested in hundreds of subreddits.

Yeah follow the that subreddit, uh, rose and ohanian now venture capitalists. Of course, alexis ohanian is married to serena williams, as you know. I've also begun meeting with community leaders from online forums, including reddit, to try to better understand what gets in the way or frustrates them, in order to address those factors. In the new dig, it will be mobile first and they they plan to quote focus on AI innovations designed to enhance the user experience and build a human-centered alternative. Kevin is very good at UI. That's really his real skill. He told the New York Times AI could help with moderation and allow for fun quirks. Oh, I have to log into the new york times. I'm using a new browser so I have to log back into everything I ever logged into what are you using instead of chrome?

ah, I found a wonderful new browser we'll talk about that in a minute because chrome has pushed out manifest v3 this week. Two decades after creating dig. There they are. Isn't that cute? Gray beards. Gray beards now, yeah they were the kids they're both 41 they're not that gray, they're only in their early 40s.

1:42:51 - Jeff Jarvis
I know dig was as much the podcast as it was the site to me well, they had that podcast dignation, which he and alex albrecht brought back.

1:42:59 - Leo Laporte
I don't know if it's going to be a regular thing, but they've done a few episodes now, uh, and it'll be interesting to see if they can capture the spirit of the old site, because people famously left dig once.

1:43:09 - Paris Martineau
They uh tried to update it right in 2006, kevin posed for.

1:43:15 - Leo Laporte
This is the New York Times writing and now infamous photo on a business week cover, sporting a wide grin and giving two thumbs up with the headline how this kid made $60 million in 18 months. Kevin hated the photo, which is true. He always hated it. The cover proved ill-fated. Digg later relaunched the design of its site that its community widely rejected. That was Digg 4. And I had for a long time a shirt that said dig four was my idea, because I actually I did. I suggested a few changes to kevin, which he adopted, and it killed it.

So uh it's my fault. Yeah, what changes did you suggest? I don't even remember anymore and it probably wasn't my fault, but I I did have some suggestions which kevin, I believe, implemented in dig for and it promptly died. Thereafter it was sold for parts to beta works linkedin in the washington post and then sold on from beta works, obviously because they didn't buy at one point it was owned by buy sell ads oh god, so it really was just at that point just an uh url, I think yeah yeah beta works.

1:44:22 - Jeff Jarvis
Did try to revive it as a news site, but it just didn't work right, this is a rough cover.

1:44:30 - Paris Martineau
I'm looking at it now isn't it, kevin?

1:44:34 - Leo Laporte
you know he kept saying I didn't make seven, sixty million, that's good, it's crazy. He didn't want anybody to think that he actually was a $60 million kid. Mr Rose, and Mr Ohanian, says the Times, raised an undisclosed amount of funding to repurchase, dig and build a new version of the company. Their investors include True Ventures, of course Kevin's a partner there and 776, where Ohanian is the founder, so basically their own own companies invested in it. They also hired a fewer than a dozen engineer. Fewer than a dozen. Is that 11?

I don't know, what that means they hired fewer than a dozen engineers and designers for the new dig and brought on justin mazel, a long-time collaborator of mr rose, to be chief executive. Well, they're pretty serious about this, wow, and they hired a pr team.

1:45:23 - Paris Martineau
I assume to obviously they.

1:45:25 - Leo Laporte
Oh yeah, article yeah, they got, uh, mike isaac to write it up. Of course, I think kevin and mike are probably friends. I'm I sound jealous. I'm not jealous. I think this is great. I loved you didn't sound jealous until you just said you weren't jealous oh okay, I'm not jealous, I'm not just don't write in the papers that I'm jealous I have almost a dozen employees as well, almost, almost jeff, what do you think almost a dozen means in this case?

1:45:54 - Paris Martineau
do you think it means they couldn't? They couldn't give us firm answer as to how many guys you're like.

1:45:59 - Leo Laporte
if you're an editor, would you allow that in a news article, almost a dozen Almost a dozen.

1:46:07 - Jeff Jarvis
Yeah, that wouldn't have gone at time. Make.

1:46:11 - Paris Martineau
I feel like it's probably between 8 and 11 then it's more than 6. Yeah, it's more than 6, for sure, it's definitely more than 6.

1:46:18 - Jeff Jarvis
That much I can say for certain, it six that much eggs as you can get now yeah, you could say six eggs for like five bucks.

1:46:26 - Paris Martineau
I felt.

1:46:27 - Jeff Jarvis
I felt like a queen, but, but, but, you're right it a good editor would have. The famous story from time inc was that somebody wrote about the hundreds of thousands of lights on the Rockefeller Christmas tree, yeah, and was ordered back to count them, dick Stoller to go back and find the exact number of lights, damn it 100,000.

1:46:47 - Leo Laporte
Please.

1:46:48 - Paris Martineau
How many were there?

1:46:49 - Leo Laporte
Not 100,000. That's in the story. Anyway, congratulations, Kevin and Alexis. I've talked to, I've interviewed Alexis and Chris. Kevin's an old friend.

I wish them the best and I'm not jealous he said with a pained face hey, this is really good news and I I just learned about this from Paul Theriot. For a long time in Europe, gdpr has required that Google allow you to ask any site allow you to take down your mentions. Now Google is allowing people outside of the EU to do this. Google has a new results about you page. Take control of results about you. We're here to help you find personal info that shows up on google search. You can ask to remove any search results you'd rather keep private. We can run regular checks for the info you care about. Let you know if it shows up in search results. If you make a removal request that meets policy requirements, google will remove the result for everyone. I think this is basically gdpr in the in the united states.

1:47:59 - Jeff Jarvis
You'll be asked for the name that goes to the personal info then you can take down your home address, your phone number, your email.

1:48:07 - Leo Laporte
Okay, that makes more sense yeah all right but this is they say enter at least one type of contact info to check for in search results. I wonder, though, if, yeah, obviously you'd like to be able to take that information down, but I wonder if you then say, you know, give it some information. And it says we found an article about you that's unflattering, could you ask for that to be taken down?

1:48:33 - Jeff Jarvis
it says here. So I went to the next step.

1:48:34 - Leo Laporte
I put in my email and said that um step I put in my email and said that, um we'll info, we'll monitor for my name and my email oh, okay, so the only monitor for the info that you gave. That's better. Yeah, so it isn't helpful if you just want to take down a page that you don't like a book review.

1:48:54 - Jeff Jarvis
I didn't like yeah yeah, book review.

1:48:57 - Leo Laporte
Why they? Oughta, I oughta why I oughta, I oughta did you watch the uh oscars uh last uh sunday? No, no, no, no, interest none. If nick cage had been nominated, would you have watched paris?

1:49:13 - Paris Martineau
yeah, but I just didn't watch really any of the movies that were nominated, so it's like I also am not a big fan of sitting through award ceremonies. I can just google who won afterwards. Why do I need to be there?

1:49:25 - Leo Laporte
there's something wrong with me, because I really I look forward to it all here I was going to say I could have guessed before you've been at.

1:49:32 - Paris Martineau
I was like leo watched the whole thing. I was maybe maybe maybe he went until he got sleepy and then he went to bed.

1:49:39 - Leo Laporte
No, I stayed up late. Oh no, you're in California, so it's not that late, Actually never mind.

You live a privileged life. It was over at 7.30. I was shocked. No, I watched it and I disagreed a little bit, as one does, but I was also able to predict most of the. You kind of can tell nowadays what's gonna, that's gonna get the votes. I thought the brutalist was really good. I'm disappointed that it didn't. It should he did. Adrian Brody did get best actor, but it should have won best film and best director, I think. But that's just me. Maybe it didn't because there was a little tiny bit of scandal associated with it. I don't think this is a big deal, but according to the hollywood reporter, there has been some backlash because they used ai in the movie yeah, to change the accents.

1:50:31 - Paris Martineau
Yeah, that's all so slightly I mean so the movie was a weird practice, but it looked like a big budget picture.

1:50:37 - Leo Laporte
I mean it was huge and I think it's a weird practice, but it looked like a big budget picture. I mean it was huge and it was long. It was three and a half hours. It was only a $10 million. It was very inexpensive. Felicity Jones and Adrian Brody the stars were required to speak Hungarian. They had dialogue coaches who helped them, but apparently they also used AI to tune up the Hungarian, so it would be more accurate. You don't like that idea?

1:51:06 - Paris Martineau
I mean on principle, no, in practice maybe okay, but I just think that there have been, from what I've understood, uh, a lot of people in Hollywood and representatives of unions, from Act actors, guilds to screenwriters, to people below the line, are incredibly worried about the encroachment of AI tools on all aspects of their job, and this feels like a small step towards that in a way that is probably concerning for those folks.

1:51:35 - Leo Laporte
Well, you're exactly right. The Hollywood Reporter said that AI was used in the film in any form has sparked online outrage. I think that's problematic, normal.

1:51:48 - Benito Gonzalez
I think that's something you can't do otherwise, even without AI, there's no person you could get to do that right Like you would need.

1:51:56 - Leo Laporte
AI. Well, you could have a loop group come in and kind of fake it, but it would be not as well done. Many suggested online.

1:52:03 - Benito Gonzalez
It should disqualify for awards consideration why it's like saying jurassic park cheated for having computer graphics like yeah, that's kind of my attitude.

1:52:10 - Leo Laporte
It I understand. If you're replacing an actor with ai, that's, that's not good. Uh, because somebody's losing a job, probably it's not going to be as good as if a real human I mean adrian brody's acting, all the acting, and it was incredible, incredible, um. So I don't think that it's tarnished because they touched up the hungarian, but no, I'm not, I'm not alone, it's reflexive ai, I don't know.

1:52:37 - Paris Martineau
I think that there's something I like when I'm watching movies and art is I'm seeing a performance being captured. I mean, I know that a significant amount of post-production goes into that. There's a million rehearsals, there's everything going on in it. It's not, you know, completely pure, but I do like the idea of I'm witnessing a person's performance and it's like like I don't know what parts of his lines are, the way he said it, or if they tweaked it to make it one way or another.

1:53:04 - Leo Laporte
Well, yeah, okay, that's the same.

1:53:06 - Benito Gonzalez
Visually it's like auto-tune for a singer maybe what about computer graphics like those?

1:53:11 - Paris Martineau
things aren't real in the scene oh they, I know, and frankly, I recently watched the gorge with some friends and I was interested in watching the Gorge because someone I know had seen an early version of the Gorge for reasons I won't go into a year ago, before they'd put the computer graphics in and told us about this crazy movie. Honestly, I kind of wish I had seen the weird version without the computer graphics, because it's fine. But that's just me being a weird, uh person in that way.

I understand that it perhaps doesn't make sense you live in brooklyn, you want artisanal movie making right, that's true yeah I want the weird little tree monsters in the gorge to be tree people covered in trees on a set in holly.

1:53:59 - Leo Laporte
The gorge. Probably was all green screen right. And Taylor Joy is just lying on a rolled up carpet on a stage, soundstage it's all green and looking across a non-existent gorge, right and everything else is generated.

1:54:17 - Paris Martineau
It was certainly.

1:54:19 - Benito Gonzalez
You have to consider these films to be cartoons like I consider those movies to be cartoons because it's mostly cartoons.

1:54:24 - Leo Laporte
I mean I have more respect for the actors. In that situation it's gonna be very hard to be act acting while looking at a tennis ball that, benito, you objected to the breath of the machine, but you're okay, is this because?

1:54:39 - Jeff Jarvis
because, again, I bought into it.

1:54:40 - Benito Gonzalez
Yeah, exactly, I want as convincing a thing as possible. If I'm watching a movie, I want it to be as believable as possible.

1:54:47 - Jeff Jarvis
But you don't want your AI to be convincing. It's not the same as realistic.

1:54:50 - Benito Gonzalez
But if I'm talking to an AI, I want to know that I'm talking to an AI and not it trying to fool me that I'm talking to a person I don't know here, benito, I think you're both sides of the mouth here. I don't know. I mean, I'm choosing the choice. Is the whole thing Like I want to be fooled when I watch a movie? I'm there to be fooled.

1:55:09 - Paris Martineau
No, I guess that's a fair point. Yeah, I agree.

1:55:12 - Leo Laporte
Among other things that you missed by not watching this fantastic program is the director of anora, which won. He won a best direction director award and the film won best picture award. Telling everybody to go back to the movie theaters, he said we're all here tonight, I could play it. Okay, you want to do okay?

1:55:36 - Paris Martineau
I mean, I don't know if I want to keep, I don't care if we hear it goes to that's quentin tarantino, by the way

1:55:44 - Leo Laporte
anora sean baker. Let me skip ahead, I probably am passionate about.

1:55:48 - Jeff Jarvis
Well, first off, thank you to the oh yeah, we're talking about six minutes here.

1:55:52 - Benito Gonzalez
It's an honor to be in this category with you.

1:55:54 - Leo Laporte
I'm going to put this down he's going to read his um, his little so we're all here tonight and watching this broadcast because we love movies you can be glad I didn't watch.

1:56:02 - Jeff Jarvis
Where did we fall in?

1:56:02 - Paris Martineau
love with the movies at the movie theater we come to this place as watching a film okay, I think we're gonna have to stop there because it's just gonna get taken down.

1:56:14 - Leo Laporte
Yeah, but basically, uh, he says and you can cut the whole thing out, I'll just read it he says you says you've got to watch a film in the theater. With an audience's experience, we can laugh together, cry together, scream and fight together, perhaps sit in devastated silence together, and in a time where the world can feel very divided, this is more important than ever. It's a communal experience. You simply don't get at home, and right now, the theater-going experience is under threat. Movie theaters, especially independently owned theaters, are struggling. It's up to us to support them. During the pandemic, we lost nearly a thousand screens in the us. I didn't know that, uh, and we continue to lose them regularly if we don't reverse this trend, junk.

We'll be losing a vital part of our. He will be losing a vital part of our culture. This is my battle cry. Do you think he's fighting a good fight or is it? Is it everybody likes newspapers?

1:57:05 - Paris Martineau
it does remind me it's fun to be to to go see something on the big screen and if it's an imax screen, and then it goes from a box to a rectangle to a box again, that's cinema, baby.

1:57:16 - Leo Laporte
I'm here for it so you, you, but you just bought a nice TV. I mean, yeah, presumably, so you don't have to go to the theater anymore well, no, I've realized.

1:57:26 - Paris Martineau
I mean, I I'm certainly going to watch movies here, but the movies I'm watching at home are movies that aren't playing in old movies.

1:57:34 - Leo Laporte
So if a new movie comes out that you would be watching at home in six months, I wanted to see the gorge in theaters, but that was not an option. It was an apple tv, yeah, movie, um so okay, so you like theaters?

1:57:47 - Paris Martineau
I think it's great to react to a movie with a crowd. I like hearing how other people are feeling about a movie. I also just enjoy the communal. I just find the communal cinema experience so much more entrancing and engrossing than watching movies alone.

1:58:03 - Jeff Jarvis
I hate it when they chew their food and their cellophane and leave the floor all sticky.

1:58:09 - Leo Laporte
And the teenagers on their phones.

1:58:13 - Paris Martineau
I hear this complaint all the time.

1:58:16 - Leo Laporte
Maybe it's just because I go to weird indie theaters where people pay yeah every everyone I'm in a movie theater with is really respectful no, in fact, if I had an alamo draft house, I'd probably go to it every week.

1:58:27 - Paris Martineau
Right the right I mean, that's the thing is I. I really enjoy going to movies where I can have a beverage an alcoholic beverage and some artisanal popcorn yeah uh.

1:58:38 - Leo Laporte
It felt a little bit to me like like old media, saying you're gonna save the newspapers. But at the same time I do agree with you, paris, like there's no nothing funnier than a comedy where everybody in the theater is laughing uproariously.

1:58:52 - Benito Gonzalez
It's 10 times funnier because of that right, this is why we don't have comedies anymore, by the way oh, because there's, no, there's, there's, there's the video theory of well, no, I mean comedy.

1:59:02 - Paris Martineau
Like blockbuster comedies were a real uh genre of movie yeah, they just don't make them anymore production companies because, they were big. They were big blockbusters and they would get people in theaters right where all the movies now are horror movies or dystopias.

1:59:21 - Leo Laporte
Scary?

1:59:21 - Paris Martineau
well, yeah, because horror movies are one thing that people come out to theaters for still the movies that you're seeing, that you're seeing in theaters, and get a theatrical release.

1:59:30 - Leo Laporte
That is what the market testing has shown people will go out for I have to say, though, I prefer a six hour mini series, where they don't have to get everything into an hour and a half and they can really develop something, and often it's better than a two-hour movie, but maybe I don't know.

1:59:49 - Paris Martineau
I mean, maybe I'll agree with you at some point. I'm still recovering my attention span from decades of social media and things like that. So I think part of the reason why I also like going to the movies is I have to pay attention, and one of the things I found with watching stuff here at home is I prefer to have friends over to watch with me, because then we get like a mini version of that community that's fun, we're all focused and no one is second screening Cause.

I think it's really easy to second or third screen when you're watching a series by yourself no, that's true.

2:00:22 - Leo Laporte
You don't go see the brutalist in the theaters. It is. Not only is it three hours and 20 minutes long, it has a mandatory 15 minute intermission in the middle yeah, I think that's cowardly stuff.

2:00:31 - Benito Gonzalez
I would have watched four hours of oppenheimer no break um there's also the fact that a movie theater ticket is like 20 now, which is like come on and the popcorn that's right god the price of the popcorn and it'll kill you because it's got stuff in it that is nobody knows, that's okay, I want it.

2:00:47 - Jeff Jarvis
But yeah yeah, give me the death pop junior's gonna ban movie theater popcorn?

2:00:52 - Leo Laporte
I'm counting on it over my dead body make america healthy again.

All right, one more break our show today, brought to you by z scaler, the leader in cloud security. Enterprises have spent, oh my god, millions, billions on firewalls and vpns, but that's not made things better. Breaches continue to rise. There was an 18 year over year increase in ransomware attacks $75 million record payout last year, and it's only going to get worse. Right, because these traditional security tools expand your attack surface. They give you public-facing IP addresses that are exploited by bad actors, and now more easily than ever with AI tools. And you know they struggle to inspect encrypted traffic at scale, which means the bad guys are now compromising you, exfiltrating all your private stuff. Bpns and firewalls also enable lateral movement. By connecting users to the entire network, they allow data loss via encrypted traffic and other leakage paths. It's just a. Hackers exploit traditional security infrastructure using AI to outpace your defenses.

I think it's time to rethink our security. We can't let the bad actors win. They're innovating and exploiting our defenses faster than ever. That's why you need Zscaler Zero Trust plus AI. Zscaler Zero Trust plus AI stops attackers by hiding your attack service, making apps and IPs invisible, eliminating lateral movement because users can only connect to specific apps they're allowed to connect. They're authorized to connect not the entire network and it continuously verifies every request based on identity and context. Plus, zscaler simplifies security management with AI powered automation You'll like that and detecting threats, using AI to analyze over half a trillion daily transactions. Look, it's very simple. Hackers can't attack what they can't see. Protect your organization with Zscaler Zero Trust plus plus ai. Learn more at zscalercom security. That's zscalercom security. By the way, I didn't mention this, but alexis ohanian is also, uh, joining the bid to buy. Frank mccourt's bid to buy.

2:03:16 - Jeff Jarvis
Tick, tock, yep is frank mccourt a good guy or a bad guy on this?

2:03:21 - Leo Laporte
I don't know if you live in la. He's a bad guy because of how he ran the la lakers, but I, I have to say I feel like he's a good guy in his goals. He wants to make TikTok safer and he wants to keep TikTok in the United States. They say they have. I don't think they're even close. By the way, they say they have informal commitments of $20 billion for a TikTok deal and they don't care about the app's algorithm Then it's not TikTok.

2:04:01 - Paris Martineau
Yeah, I mean that's. The thing is, you're never going to get that algorithm no sale is because the chinese government has forbidden the sale of well. That's why they say it's okay we can.

2:04:09 - Leo Laporte
We believe that. And if you look up frank mccord, he's been funding this uh new way of um doing social, that's privacy forward, and um, here I'll, it's in his wikipedia, uh thing.

2:04:24 - Paris Martineau
Oh, not that frank mccourt I mean, I'm just curious to see what ends up happening with tiktok overall, because the clock is ticked off is not ticking, it's not.

2:04:34 - Leo Laporte
That's the funny thing. I don't think anything's gonna, it's just to fade off into the distance.

2:04:40 - Paris Martineau
But I mean this all opens up all of the companies that are hosting Tik TOK to huge financial penalties Whenever the next administration comes around, because the statue of limitation be a next administration. You said we're not allowed to talk about politics, Leo.

2:04:58 - Leo Laporte
So Frank McCourt, who is hated in LA because of how he ran the Dodgers the LA Dodgers. He's written a book called Our Biggest Fight reclaiming liberty, humanity and dignity in the digital age. I don't know if he has the expertise to do this, but he has put together let me see if I can find this here all this stuff about the Dodgers. That's all. Anybody cares. Project Liberty, which was designed to fund something they call the DSNP, the Decentralized Social Networking Protocol open source code, in other words, designed to build a social networking protocol. Open source code, in other words, designed to build a social network that is open. And that's why I guess he doesn't care about the algorithm. He wants to influence the direction taken by technology and its players.

2:05:57 - Paris Martineau
Yeah, it sounds complicated, to say the least. I don't think 20 billion less well thought yeah, it's the real thing.

2:06:05 - Leo Laporte
It doesn't. Yeah, yeah, but but you know, I and I like alexis, I think his heart is in the right place, so I think if he's in on it, that gives it more credence well, he's always been one of the good guys youtube says it has now more than 1 billion podcast viewers a month. We are not in that list. Actually, I'm sure we they include our viewers in there. We have a few million 70 that's why we stopped playing music.

Because you're on youtube, yeah in 2024 users watched more than 400 million hours of podcasts every month on living room devices, on your big screen, on your tv hi everybody crazy hello couches we do have people watching hello chair, place to scans hello moon, oh no hello coffee table. That's a different uh, that actually leads me into something that benito and I have cooked up. One month from now will be our 20th anniversary as a podcast network. The first twit was april, I think 14th 24 hour live stream.

2:07:14 - Paris Martineau
24 hour live stream. All right, we all get matching tattoos.

2:07:20 - Leo Laporte
Okay, that sounds like a good idea. Do you want to see mine? I do no, moon us, I would have to moon you to do it. We, benito and I, came up with the idea because we already had all the old hosts on on our thousandth episode. So I think April 13th will be the 20th anniversary twit, and I want everybody who watches our shows to send us a video talking about how you watch. Do you watch on a big screen? Showing us your setup, showing us your smiling face? Maybe talking a little bit about how you discovered twit? Share your experience. Do we have?

2:07:58 - Paris Martineau
it. Do you want them? Do you want them to film this horizontally or portrait mode? Do it any way you want, okay no, specify one, okay one.

2:08:10 - Leo Laporte
Do it like you would if you were doing a tick tock hi leo. Vertical video. I watch your show when I'm in the bathtub see watch your show when I'm in the bathtub, see something like that.

2:08:23 - Benito Gonzalez
You know that'll be great, okay. Um, okay, we'll need to set up like an email or something I'll talk to. I'll talk.

2:08:26 - Leo Laporte
I wasn't even thinking about it I started launched into the request and I didn't have a way to do it. Well, you or you, if you're in the club, you can upload it uh to the to uh discord. Just get put a link. Upload it to icloud or microsoft's one driver, google drive, and just put a link in the discord. That would be good too, you can. You can dm me in the discord, um, because it'd be fun to watch how people watch. But I think it's changed a lot. I think people do sometimes now watch on their tv set that'll be fun.

So what are you going to do with those?

2:08:58 - Paris Martineau
videos.

2:09:00 - Leo Laporte
Oh nothing. No, I'm going to play them on the show.

2:09:04 - Paris Martineau
We'll do a story and I'll say Nella, here's Jerry from Des Moines, Iowa.

2:09:09 - Leo Laporte
How do you watch Jerry? And we'll show the video. It'll be fun. You know we honored all the people who started the network with us. But you know who really counts is the people who started the network with us. But you know who really counts is the people who enjoy our shows, who watch and listen. That's who really. That's what we do it for, so I it's a chance to see them who is a club twitter member number one?

2:09:26 - Benito Gonzalez
who's club twitter member number one? Oh, I mean, you know who paid for it? Who poop? Who was the first?

2:09:32 - Leo Laporte
seven dollars. The first paying. Who's the first paying? Member of club twitter. Patrick, you would know that he's checking. He's gonna.

2:09:39 - Paris Martineau
I think that we could uh play people's um twit videos between 5 pm and 8 pm on the 24 hour live stream. Listen, I'm filling out the slots guys, it's, it's very hard.

2:09:53 - Leo Laporte
You have to entertain people while you're exhausted.

2:09:57 - Paris Martineau
You really want to do this. I think it would be kind of fun. We could get a bunch of people to take up shifts. You know we used to be. I romanticize these sort of things.

2:10:06 - Leo Laporte
It just seems fun to do a 24-hour live.

2:10:09 - Paris Martineau
I'm up all night anyway, so I remember what having that kind of energy was like wow listen, I'm one of the sleepiest people I know. It's not that I have a bunch of energy. I just think it would be kind of fun to participate in a 24-hour live stream.

2:10:24 - Leo Laporte
Well, okay, I'm going to take it on advisement. I would too.

2:10:28 - Paris Martineau
We could game for a couple of hours.

2:10:31 - Leo Laporte
I could just leave the lights on Jeff could talk about Linotype for an hour.

2:10:37 - Jeff Jarvis
Hey, yeah, three hours at least.

2:10:39 - Paris Martineau
Three hours. Listen, we got a good half of the day figured out already.

2:10:46 - Leo Laporte
Jammer B says, if we do another live stream, he's coming down. Whoa, that's a promise.

2:10:52 - Paris Martineau
Well, he was instrumental. He could show out his little saws for a couple of hours His saws.

2:10:58 - Leo Laporte
He has special saws.

2:10:59 - Paris Martineau
Burke always has a saw hours, his saws, he has special saws. Burke always has a saw right, a chainsaw.

2:11:04 - Leo Laporte
No, that's Elon Musk you're thinking of. We did start 20 years ago. We used Skype first for audio, then for video. Now here we are, 20 years later. It's all over.

2:11:28 - Paris Martineau
Bye-bye Skype, we hardly knew ye rip.

2:11:31 - Leo Laporte
Do you know what skype is? Paris, I used skype. I'm not that young. Uh, microsoft is going to take all the skype users and turn them into teams users. That's not going to happen, no I yeah, I don't like teams pre-microsoft skype was amazing jammer b confirms burke does have multiple saws. I'm not crazy oh, first paying member of club twit is seth graham. Seth, are you still with?

2:11:57 - Benito Gonzalez
us Still an active member, Seth Patrick.

2:11:59 - Leo Laporte
We salute you, seth.

2:12:02 - Paris Martineau
Wow, still an active member See, we could have a one-hour interview with Seth Graham.

2:12:06 - Leo Laporte
I'm feeling it out.

2:12:08 - Benito Gonzalez
Yeah, we can get him on. Let's try to get him on.

2:12:10 - Leo Laporte
Seth, we'll get you on. I want to thank you and honor you.

2:12:14 - Benito Gonzalez
If you're listening to this episode.

2:12:23 - Leo Laporte
Anything else in the talk about the la times stupidity. Oh yeah, that was crazy. What is he doing?

2:12:28 - Paris Martineau
so one of the hot new things in uh journalism right now is what if we put ai on the news articles? Uh, and that will somehow equal profit and for the only times that turned out poorly, jeff, if you want to explain well, it wasn't on the news articles.

2:12:46 - Jeff Jarvis
What they did, what he, what he's, what he's promised to do, is to a use ai to determine bias. This is patrick.

2:12:53 - Leo Laporte
Oh my god but so every, every opinion piece in the la times. From now on we'll have a little, not even opinion pieces. Right it's just now it's well he'll do his opinion, editorials, commentary and more but.

2:13:07 - Jeff Jarvis
But here's the other thing he did that I didn't know he was going to do. Is that he to both sides things he has the ai. Respond to the opinion piece. Yeah, it's a little and makes a point of saying that no human touches the AI, so it just goes up. So they had the AI defending the KKK.

2:13:30 - Leo Laporte
Which I'm sure it did with confidence and hallucinations.

2:13:36 - Jeff Jarvis
So I put up a PDF of an example of this, because you can't do it if you use archiveai, dot, dot, whatever ph.

2:13:43 - Leo Laporte
Yeah, so where, where did you, where did you put this?

2:13:45 - Jeff Jarvis
it's on the it's on if you look at the rundown I put. I put a lot of care into the rundown today if you click on jeff's little icon it'll take you to it jeff's little icon.

2:13:54 - Leo Laporte
He has a little icon do you know? How google? Docs works oh, I see. So I click on his icon and then what? Oh, it highlights everything he wrote. I didn't know 20 years and I still didn't know that.

2:14:10 - Paris Martineau
Well, you know here's a little Jeff headshot.

2:14:13 - Leo Laporte
I did, I clicked on his head and it highlighted the article.

2:14:17 - Paris Martineau
Oh, yeah, great documentaries reveal history's truth.

2:14:20 - Leo Laporte
Unregulated AI threatens to distort it. This is a voices column. Written by humans, right? Rachel, Stephanie and Jennifer. Okay, so you go down to the bottom.

2:14:32 - Jeff Jarvis
I didn't capture the KKK one, it was gone so keep going, keep going.

2:14:36 - Leo Laporte
Insights is the AI, so stop there. This is a center and left. Go up a little bit. Go up a little bit Go up a little bit yeah.

2:14:45 - Jeff Jarvis
So, yeah, it says here that it aligns with a center left point of view. Sorry, you were right. And then now go below perspectives plural Ideas expressed in this piece and then different views on the topic. So, for example, it says that AI could democratize historical storytelling, allowing creators to visualize poorly documented events, that is to say, make them up or marginalize perspectives. Tools like MidJerby enable amateur historians to generate images of ancient rome I don't want to see it.

2:15:20 - Leo Laporte
Nope, no, it's fake. On the other hand, regulation risks stifling innovation is seen in creative applications like there's the real val kilmer's voice for top gun maverick. Right, he, because he can't speak. So in top gun maverick, they used an ai to generate his voice, right, I mean?

2:15:38 - Paris Martineau
similarly they used AI to do Anthony Bourdain's voice, despite protests from his estate. No that was creepy. Yeah, yeah.

2:15:50 - Leo Laporte
Although it was saying things he said it was just simulating the voice because it was actual quotes from him. Now here's my thoughts on this, because it was actually quotes from him. Now here's my thoughts on this. This, just this, is just a way of ginning up more crap to fill up the pages of the la times, but he's also trying to do this thing where he wants more balanced opinions.

2:16:08 - Paris Martineau
Yeah I'm just curious as to what sort of money the la times is paying for perplexity and how that compares to the uh salaries of the journalists they've laid off for the last couple of years yeah so they use perplexity for this it's.

2:16:24 - Leo Laporte
It's credited there, yeah yeah, well, that's cool because that's what I use for my stories and those are never wrong, of course well, okay, so generally, perplexity's right is accurate and reliable, and it because it gives you footnotes, as it this LA Times Insights does. I was trying the other day and it was extremely frustrating. I have this idea. I wanted AI to take the spreadsheet we have you've seen it of all the people who've been on our shows, make a list of everybody's one name and then how many times they've been on, what dates they've been on.

2:17:00 - Jeff Jarvis
That's the kind of scraping it can do pretty well, isn't it?

2:17:03 - Leo Laporte
I thought, oh, this is going to be great. I thought everybody was on in the year 2023 for some reason. So I said well, you know, the dates range from like 2005 to 2025. And of course, AI does this. Oh you, you're right. I just noticed that too, so let me give you another list. And then they were all from 2024.

2:17:28 - Paris Martineau
have you tried messaging it please? I make a lot of points about how you can do everything on my podcast. You've got to come through for me it did bring home to me.

2:17:37 - Leo Laporte
Now, probably there is a way to do this.

2:17:39 - Jeff Jarvis
I don't know yeah, I don't brought it home to me to scrape data like crazy.

2:17:43 - Leo Laporte
Yeah, I would think that it seems to me that should be something it can do, especially since I'm giving it the source material right.

2:17:50 - Paris Martineau
It says I wonder if it's something to do with um, the way that the date data is stored in google sheets, because I know sometimes when I'm copying over columns of dates from like one sheet to another, something happens with how it was stored, to where it gets all weird. Or all the dates are the same one in the paste.

2:18:10 - Leo Laporte
Yeah, it's good. It's a little strange.

2:18:12 - Jeff Jarvis
Dates. Line one always line one. Column E always has the date.

2:18:17 - Leo Laporte
Yes, you did, wow, yes. Still, I was very specific in the prompt part of these because we are looking at little chunks that are smaller than the actual chunks that the ai's llm is looking at. So it you know. That's why I can't tell you how many r's in strawberry. It doesn't see the word strawberry or r's for that matter. Patrick says I have a page on our system that will show all the guests and sort by number of appearances leo, thanks, patrick it's just, you don't want to hear that I don't understand why you're making this so hard.

2:18:50 - Paris Martineau
All right how many trees were burnt for that, leo?

2:18:54 - Leo Laporte
I don't know, because I spent most of an afternoon beating on the thing. Yeah, I know we have an API, Patrick, that was what actually came to me is I thought I should probably just query the database. It would be. But yeah, or ask Patrick.

2:19:09 - Paris Martineau
A human being, a human being in your database.

2:19:15 - Leo Laporte
Oh yeah, we have. We did the right thing when we created it some years ago. But anyway, somebody's saying in the Discord and I think this is kind of interesting I just want the facts, man. I don't want balanced opinions, I just want facts. Report the news and give me the facts of the story without opinion. Is that Ant? That's what Ant says. No, I know Ant likes that, but no, it's Trust Noan in our Discord.

2:19:40 - Benito Gonzalez
What do you think?

2:19:40 - Paris Martineau
I agree, it's not really possible, though I mean every, every, yeah every article as a writer that has their own biases and ways that they're seeing the world like it's something people always say, that's like it's actually impossible, like that's not a possibility.

2:19:57 - Leo Laporte
Yes, because humans have the choice of AI could do it is reflects perspective, right and actually you think you want just the facts, but you really don't, you, you need. You need more than just effects. So, um, patrick soon Shong, who owns the LA Times, is a physician, a surgeon, a scientist, an inventor, a technologist and a philanthropist. That's why he's good at nothing. He is the CEO of a holding company devoted to the transformation of healthcare and using artificial intelligence to win the war against cancer and Alzheimer's. How's that going?

2:20:42 - Jeff Jarvis
He's liking RFK Jr, so is he. I'll leave that there.

2:20:47 - Leo Laporte
Yeah, he has 230 patents and is obviously a big fan of AI. We'll see, I mean, you know, if that, if the readers of the la times uh love it, then maybe it was a good thing I sure rename it the ai times the ai times.

um, okay, I, we're gonna wrap this up pretty quick here. So, uh, yes, jeff, you put a lot of wonderful, wonderful energy into this. What would you like me to? What should I bring up Timnit Gebru's post that's a good one On LinkedIn where she says there is in, says it in caps she is. Of course, we should mention the former google ethicist who was hired by google to help them with ethical ai but then fired when she margaret mitchell and was it emily bender?

2:21:48 - Jeff Jarvis
wrote yes, the stochastic parrots article beetle I always forget her name.

2:21:52 - Leo Laporte
Yeah, the fifth beetle, um, she says, by the way, temnet was fired then after that for having the temerity to imply that google's ai wasn't perfect in every way. She says there is no such thing as an agi, and attempting to build it is a eugenicist power centralization project that no one should support.

2:22:16 - Jeff Jarvis
I actually don't disagree with that yeah, and she said she said at the beginning, and caps in lowercase. She says the Transformers and what they were designed for. She remembers she didn't remember anybody talking about artificial general intelligence until open AI. This is why I asked Gary this came on the scene and started talking about large language models as stepping stones toward AGI. Most natural language processing people I come across were talking about creating pre-trained models, the way we had ImageNet pre-trained models in computer vision. I don't remember any of them talking about AGI, but people decided to follow this trope now and so she's constantly screaming it and I'm agreeing. And Gary says that he came up with his definition for the sake of the bet, but I don't think there's any clear definition out there, and eric schmidt in the wall street journal goes on about the changing, changing definitions of agi as well, um, and so we're constantly seeking a macguffin yeah, her piece that she points to is called the test creel bundle.

2:23:22 - Leo Laporte
Eugenics and the promise of utopia through artificial general intelligence is what you've been talking about, and that's, by the way, with emil torres was. Is he the fifth beetle?

2:23:30 - Jeff Jarvis
they are uh the uh no, no, no.

2:23:33 - Leo Laporte
They've worked with uh, with uh, timid on this okay, um yeah, I'm glad you uh brought that up. I don't disagree with it. It definitely seems like the people pushing agi are pushing another agenda as well that's what I'm trying to say, and the agenda is eugenicist.

2:23:51 - Paris Martineau
Yeah um paris, anything um I thought this was kind of cool. The new york city subway is piloting a program to use google pixels strapped to the subway cars to listen to track defects using ai okay that seems okay, I don't, I don't dislike that listen. Anything that helps uh these poor trains run marginally faster is good for me, and this seems to be one way to help detect uh potential breakdowns in cars or service without having a bunch of human inspectors having to walk the tracks every day seems great.

2:24:30 - Benito Gonzalez
This is the stuff that's ai is for, like these little, tiny little jobs that get no, no love that no one really cares about, but much data. This is the stuff. This is the stuff that it should be doing yeah how many did they say?

2:24:43 - Leo Laporte
how many they're gonna?

2:24:46 - Paris Martineau
let me try and get here uh, you know who does care about it a new york city transit assistant chief track officer, robert sarno, a 14-year veteran of the MTA, who, in order to help train this AI model, basically threw on some noise-canceling headphones and listened through five to 10-minute snippets of track noise to sort regular subway din from audio signatures of track trouble. So he'd listen to it and mark down loose joint, loose bolt, rail defect when he hears like a certain shake. And he by himself hit an 80% accuracy rate with just hearing this. But that was after 14 years of work. People like him can actually work on like fixing the problems rather than listening to videos or audio on the train in order to hear if a bolt is maybe going to come loose wow which we want them to hear.

2:25:47 - Leo Laporte
Six google pixel smartphones hitched free rides on four subway cars. I guess you don't need that many. I mean, you just got to go over the track, yeah kind of interesting.

2:25:57 - Jeff Jarvis
Of course there actually were seven, but one of them got stolen. I'm joking, did it, I wouldn't be surprised.

2:26:03 - Leo Laporte
No, nobody steals pixels. If it had an iphone that'd be another matter hey hey, youtube has introduced premium light.

You don't have to now buy youtube music to get youtube premium for eight dollars. You can just get youtube without some of the ads, not all the ads. They don't promise to remove all the ads for your eight bucks. I did say that I was going to talk about Manifest V3. I got pushed out an update to Chrome. I imagine most of you did as well that without a word or a hint or anything, just deleted the ad blocker I was using. You block origin because it's not compatible with an update to chrome called manifest v3.

2:26:44 - Paris Martineau
Serves you right, google I will not be updating that until I switch over to firefox, so what are you using now?

2:26:50 - Leo Laporte
um well, let me just say why google says this is a good thing. They say this uh, in order for an ad blocker to work, it has to look into the pages you're on, and that's a security problem. They don't want third-party extensions to have those permissions. Gorhill, who created ublock origin, has created a light version that still works under manifest v3, but doesn't work as well, so I switched to a new browser called Zen Browser. That is kind of actually well, it hits my sweet spot, let's put it that way Oops all.

2:27:25 - Paris Martineau
AI, because it is no AI involved.

2:27:27 - Leo Laporte
In fact, yeah, there's no AI. It's an open source project. Firefox uses the Firefox Blink Engine, but it has a UI like my favorite, now abandoned browser, the Arc browser. The browser company decided that there was no money in making their Arc browser, so they're moving on to something else that does have more AI in it, so I get really what I wanted. The Arc browser is based on Chromium, so I get a Firefox-based browser with the Arc UI and, yes, ublock Origin continues to work just fine.

2:28:03 - Jeff Jarvis
And the browser they're working on is really interesting, Jason. I talked about it last week. It expends, for example, the cursor. This is how AI is going to be used. You're talking about the browser, company's new tool.

2:28:15 - Leo Laporte
It's an agentic browser.

2:28:18 - Jeff Jarvis
Yeah Right, be your cursor becomes a window onto functionality about this that you're pointing to. It's really interesting yeah, uh, yeah.

2:28:27 - Leo Laporte
This is all the rage now in ai is agentic, where the ai does things for you and, uh, if you give it control in your browser, it can find the best porn. Anyway, I like the zen browser and that's what I'm using.

2:28:41 - Paris Martineau
It's free, it's all roads lead to porn as usual, yeah, uh what else is a series of tubes filled with porn that's disgusting I bet they are quite sticky this is important for uh, paris to know.

2:28:58 - Jeff Jarvis
Line 149. Michael masnick is coming up with a new podcast called objectionable, a limited term series about guess what section 230 it's otherwise objectionable is the name of it sorry.

2:29:12 - Leo Laporte
Yes, yeah, sorry, I screwed that up he has, uh, of course, been a great defender of section 230, probably more than anyone else, including this page that he created. Hello, you've been referred here because you're wrong about section 230 of the communications decency act. He wrote that five years ago, but it still holds. It's still good, it's still true. Uh, yes, I you know mike's been doing more podcasts, which explains why he's hard to get on our shows, I guess. Uh, good, I'm glad to hear it. Um, the competitive enterprise institute is is sponsoring this as a documentary podcast exploring the past, present and future of section 230 between mike and jeff kossif.

2:29:57 - Jeff Jarvis
Those are the two and our friend here on the show.

2:30:03 - Leo Laporte
You know they're trying to defend 230. Chris Cox, ron Wyden. Yeah, oh, kathy Ellis is on it.

2:30:08 - Jeff Jarvis
Oh good, these are more people who want to run up. No, no, I'm saying she's a defender as well. Yeah, oh yes absolutely the single biggest threat to free speech in this country is the power of big tech.

2:30:16 - Benito Gonzalez
It's time to fight back that was the wonderful Ted Cruz talking.

2:30:20 - Jeff Jarvis
there it's now time to repeal Section 230. Section 230. Section 230.

2:30:26 - Leo Laporte
Section 230. And, of course, elizabeth Warren is also in there. It is a bipartisan attack on what is the most important piece of legislation with regard to the Internet ever, and Mike will defend it, I'm sure. Good, I'll have to listen to that. Techdirtcom uh, I guess a little politics. One of the things that is being dismantled right now by doge is 18f, uh, which you may have ever heard of, but like the US Digital Service, which has also been kind of replaced by Doge, 18f were coders, were technologists who helped the government implement websites and other technologies, most famously the IRS's direct file system, and direct file is still running, but everybody at 18f has been dismissed. So the question is how long it will keep running. Um, this is just a part of a, just a continued assault by elon musk, doge and the president on science and on technology and, um, how it makes any sense at all to get rid of these groups that helped american government be efficient, work better and serve the people. I don't know. I don't understand. I really don't understand.

2:32:07 - Jeff Jarvis
I really don't uh but at least we're gonna have a, a cryptocurrency strategic uh reserve.

2:32:12 - Leo Laporte
Thank god, thank god, right, I have our dogecoin. All right, I think any, unless you tell me that there's something we got to say here. I'm just looking through it. There's, of course, so much good stuff.

2:32:27 - Jeff Jarvis
You want to spend $100 million buying AIcom $100 million.

2:32:31 - Paris Martineau
The information says oh, this is a great feature. My colleague Akash did this weekend.

2:32:35 - Leo Laporte
Larry Fisher, a veteran Brooklyn domain broker, aka a domain squatter, hopes to convince the likes of he's not the squatter.

2:32:43 - Paris Martineau
He's the intermediary between the squatter and a fun detail, and this is the person who owns AIcom, who declined to be named or interviewed for the article did, I believe, send a written response back to my colleague Akash's inquiries, described himself as being very lucky that they bought this domain years and years ago, not because of anything related to AI, but because it was their initials.

2:33:10 - Leo Laporte
Damn, I wish I were named Adrian Ignacio Fisher. A 62-year-old Brooklyn native, says if somebody's looking to sell a top tier domain, I'm the guy they call.

2:33:23 - Jeff Jarvis
I have a reputation for being able to get the best price and he is from queens, so it's the right, uh, right, I hope that's all right.

2:33:29 - Leo Laporte
There he sold. Let's just give you some of his greatest hits. According to akash, he sold skincarecom to l'oreal. He sold skincarecom to L'Oreal. He sold teamscom to Microsoft, chatcom to HubSpot, who then flipped it to OpenAI and made a little money. I'm sure that that really grinds his gears.

2:33:55 - Paris Martineau
To begin attracting attention to prospective buyers. The owner has adopted a clever sales tactic in the last few months, continually redirecting AIcom to different sites. For a while it went to ChatGPT's website. These days it goes to DeepSeek. And the ploy has had an amusing ripple effect. A bunch of different news outlets, like TechCrunch and things like that, have written a slew of headlines mistakenly claiming that OpenAI bought AIcom or XAI bought.

2:34:23 - Leo Laporte
AIcom. I got a deal for you, August D'Anascio. Why don't you point it to twittv and fool them all?

2:34:31 - Paris Martineau
No one will ever see that coming.

2:34:32 - Leo Laporte
No one will ever see that coming A hundred million. Take a Paris honey.

2:34:38 - Jeff Jarvis
Brooklyn accent to your queen's accent.

2:34:40 - Paris Martineau
I'm sorry, I'm sorry ron bergman, a wharton school associate professor who focuses on digital advertising, says I don't think it's going to get to nine figures of value, but if it does get there, that's a deal breaker, ladies, that's a deal breaker ladies, I think that's a good sign that we may be starting to get back into a dot-com bubble a bubble. It's a bubble um that's a bubble ladies, bubble ladies initially.

2:35:11 - Leo Laporte
So he had he he did buy a lot of uh. He was snapping up domains. In 1995 he bought brotherscom, sisters and StockQuotescom. My wife put her foot down. If I couldn't turn this into something that generated cash flow, it was over.

2:35:28 - Paris Martineau
Honey, you got to sell your domains. You got too many of them.

2:35:33 - Leo Laporte
I just sold Portageecom for $3,000. He sold Returncom for a low six figure sum after buying it for a hundred dollars. That put him back on top and officially on a roll. You know, akash did a great job of capturing this guy. I love it. I love it. Um, all right, 100 million dollars. Aicom I mean, I don't know how much. How much did open ai pay for chatcom? We don't know. Hubspot bought it for $15.5 million. I flipped it almost instantly. Oh, he got equity. Oh, oh, big mistake, buddy, he got stuck.

2:36:14 - Paris Martineau
Over the years, fisher himself has developed an attachment to certain names he's unwilling to part with. He grew up an avid follower of pop art but could never afford his favorite artist's work. Over the years, fisher himself has developed an attachment to certain names he's unwilling to part with. He grew up an avid follower of pop art but could never afford his favorite artist's work. So in 1997, he bought AndyWarholcom and has been sitting on it ever since, despite several attempts from the Andy Warhol Foundation to claw it back. It's my painting, it's my artwork, he said. The site is the same Warhol fan page Fisher set up 26 years ago.

2:36:41 - Leo Laporte
Yeah, it looks like it doesn't it? Yeah, it does. Wow, Andy Warhol. It's proudly powered by WordPress, though. It is indeed yeah, he's vowed to keep it as is After scores of deals. Fisher said he's addicted to the feeling of outdoing himself, driven by the high he gets from closing seven and eight figure deals. It's just like when my kids are born. It's the same feeling. As we finished dinner and got up from the table, I asked how he'll feel when AIcom sells Triplets. He said Nice job, Akash Good story. Who says?

2:37:24 - Jeff Jarvis
technology news is boring.

2:37:25 - Leo Laporte
Triplets it's human Triplets, deutsche Telekom and my favorite AI company, perplexity. They're doing interesting things. They're going to create a new AI phone priced at under $1,000. This was announced at Mobile World Congress Low-cost headset. I don't think that's low-cost. Created in close. It's called oh. It'll have a new assistant on it called Magenta AI. You know, mark Gurman, who writes about Apple for Bloomberg and is very connected, has said that Apple has fallen way behind on AI Apple intelligence for its smartphone. There may be an opportunity here.

2:38:08 - Jeff Jarvis
Nobody's really figured out the AI device yet. Have we seen, you know, human, humane and all that failures. Google has certainly got AI on its.

2:38:22 - Leo Laporte
Perplexity is a very clever company. I love perplexity because I can use different models for 20 bucks. I get to use all the other ones that I'm also paying 20 bucks for, foolishly. One thing apple did do, though they put an action button on their phone and now chat gpt will talk to me. Actually, I think this is perplexity I'm going to talk to here. Let me see, want to have an?

2:38:46 - AI
existential debate. What is a nihilist? A nihilist is someone who believes in nihilism, a philosophy that rejects or denies the existence of meaningful aspects of life. Nihilists often believe that life is meaningless. Aspects of life. Nihilists often believe that life is meaningless, that there are no objective moral truths and that traditional values and beliefs are unfounded this perspective can lead to a sense of skepticism and pessimism about the world and human existence.

2:39:13 - Leo Laporte
That's not wrong, benito, that's Paris.

2:39:17 - Benito Gonzalez
I mean, that's not really the definition isn't it?

2:39:19 - Paris Martineau
it's not really yeah the two members of the optimistic nihilist club would disagree with you oh, you both are optimistic nihilists he's got. He's a card carrying member. I'm just a renegade that's hysterical.

2:39:34 - Leo Laporte
it doesn't have any uh, you know what. It doesn't have any footnotes as to the source on this. That's interesting, but if I ask it, it will generate a picture of oh, this is maybe where it got the information from Some pictures of nihilists. Those people look hopeless. All right, let's take a final timeout and come back with your picks of the week. I feel bad that Jeff missed your picks of the week. I feel bad that Jeff missed my pick of the week last week. Have you been playing Bracket?

2:40:05 - Paris Martineau
City. I've been playing Bracket City. I've sent Bracket City to friends. Isn't it great? What am I now? I'm like mayor, I'm like editor-in-chief of Bracket City.

2:40:22 - Leo Laporte
Each time it's different because it's how efficient you were in solving it right. I am high ranking in bracket city, like I. I know that a word for a bunch of fish is a school. Jeff. Did you see this? You, you?

2:40:27 - Jeff Jarvis
would? I saw you, I saw I couldn't figure it out I didn't.

2:40:29 - Leo Laporte
School event with chaperones is a dance, a dance. Uh, let's see. What are that? Cows and what is that? What are those emojis? I don't know. Um, hold on, I've got to pull it up on my screen in a big kind of stand where you are upside down Cows, and what is it? What?

2:40:38 - Paris Martineau
are those emojis I don't know. Hold on, I've got to pull it up on my screen in a big way, kind of stand where you are upside down, cowboy Handstand.

2:40:45 - Leo Laporte
Oh cowboy, is this the new wordle?

2:40:47 - Jeff Jarvis
Is this going to piss me off?

2:40:48 - Leo Laporte
Well, this?

2:40:49 - Paris Martineau
no, you know the story is great, so you've got to figure out everything in the brackets to then make something that connects to once you uh, reduce all of them down. It gives you something relating to this day in the past. It's really what I have a life no, this is the word.

2:41:06 - Leo Laporte
This is do you do crosswords? No, I hate them. Oh well, you wouldn't like this then? No hand hand is the next one ah, okay, I put creek for like a river, but smaller. But that was wrong. Something to do with a hand or money.

2:41:26 - Paris Martineau
You do the hand or money. Vent is the next one.

2:41:28 - Leo Laporte
Vent, release frustration. You don't have to do it in order. Invent as a phrase. We'll get that, like a river, is a Jurassic Park.

2:41:38 - Jeff Jarvis
That's why I hate this.

2:41:39 - Leo Laporte
I hate this if you have it, you might rip one. Suck me in if you have it you might rip one material associated with biker gangs leather, a glossy leather for fattened for fancy shoes, patent. If you have it, you might rip one.

2:42:04 - Paris Martineau
Ripper what.

2:42:06 - Leo Laporte
Jammer B says gas, but I don't think that's it.

2:42:12 - Paris Martineau
I don't think you rip gas is the thing. No stream, stream, yeah, stream is where we go.

2:42:17 - Leo Laporte
Yeah, I tried crick Stream of social media content and the next one is meter feed or feed feed.

2:42:25 - Paris Martineau
No feed is first meters after you put feed in.

2:42:28 - Leo Laporte
Sorry, I'm reading ahead, you're reading ahead, you feed oh yeah, we're not. Well, we're not it's not yellow. Yes, we can if you have it, you might rip one, invent as a phrase. It's something to do with a hand or money. We'll leave this as an exercise for the reader, but I just wanted to make sure jeff saw that, so he could hate it oh, coin is the middle one invent, isn't it fun?

2:42:52 - Paris Martineau
so the story is.

2:42:52 - Leo Laporte
This guy owns a bar and he does trivia contests. He loves puzzles and he just made this up. It's free, it's online, and the real question and we should have a pool is how long before the new york times buys this right, because that's clearly what's gonna.

2:43:06 - Benito Gonzalez
Six months wait, so maybe, maybe we should buy it first and then sell it to the new york times we could flip it maybe we should get that domain.

2:43:15 - Leo Laporte
Guy in on it, he seems to be our agent a broker, don't you think he's probably got a few calls from people who want to do that exact thing. Uh, all right. Um, just real quickly, a couple of things I'd like to ask of our fine audience. First of all, if you're not yet a member of club twit, we would love you to join the club. It helps us defray expenses, it keeps the show on the air, helps us grow. Um, you know, we're always hanging by a thread. We have lots of advertisers now, I'm happy to say, but it still only covers about 90, 95 of our costs.

The club makes up the difference. If you're not yet a member seven bucks a month. Ad free versions of all the shows, because I wouldn't charge you and then show you ads. That's terrible. You also get access to the Club Twit Discord, which is a great hangout Place for lots of fun events, including, by the way, tomorrow our photo thing with Chris Marquardt Photo time. That'll be 1 pm, pacific. Seven bucks a month. Twittv slash, club twit. Please join the club. We'd love to have you and uh access to leo's only fans account.

2:44:29 - Paris Martineau
Is that I don't get it I mean, this is an only fans of some sort. It's just not pornographic right.

2:44:38 - Leo Laporte
Actually, there are only fans that are not pornographic.

2:44:40 - Benito Gonzalez
There are plenty of them there's a lot of you know, there are a lot of porn hub channels that are non-pornographic because the yeah, you get.

2:44:48 - Paris Martineau
Yeah, because it's got a good algorithm hey, we should stream on porn hub guys there's a woman, uh, yeah we've covered it before.

2:44:55 - Leo Laporte
Oh yeah, we talked about it. Yeah, yeah, never mind, uh. The other thing I want to ask you, whether you're a member or not, leave a review if you would, especially on apple's itunes. We've been review bombed in the past. It's brought our numbers down, but if you like the show, share that on, uh, whatever platform you like. Uh, I don't, you know, it's okay with me if we are 4.1 or whatever, uh, but advertisers do care and so that helps us a lot. So just another little thing you can do um, just go to itunes or or whatever you I'm on pornhub. Are we streaming on pornhub? Or just somebody just make that. I think it's a joke.

I hope it's a disappointed users out there that's the kind of thing you're missing if you're not in the club so related.

2:45:48 - Jeff Jarvis
In the early days of blogging I had blog ads yeah and henry copeland came up with a new format. It was a big ad and we called it the big ass ad, and so I posted that I was going to have the big ass ad and I have rejected in a couple years, but for many years if you search for big ass, you'd get that that's smart. Well like it was either either definition.

2:46:13 - Leo Laporte
I wasn't very happy that I was associated with big ass and you generate a lot of traffic there. Very, you should have registered bigasscom. I should have, I should have yeah, yeah, larry fisher would be on the line. Jeff, give us. Oh, I start with paris.

2:46:30 - Paris Martineau
I'm sorry, paris your pick of the week? What?

2:46:31 - Leo Laporte
are you doing yeah?

2:46:32 - Paris Martineau
um, I saw this come across my blue sky feed recently. Have you guys ever heard of the vintage computer festival? Yes, I'm thinking of going this year is it new york? It's in new jersey, like an hour and a half drive yes, oh yes, in a month it seems really fun yeah, um, it's in wall.

New jersey. Jeff, you have any recommendations for me to check out new jersey from in the hour and a half drive from new york? Let me know where is it? We have 500 towns in new jersey. I always forget where they are. I couldn't tell you. It's near the. It's at the info age science and history museum, if you know that. Um, but it seems really fun. There's going to be a lot of uh, there's going to be computer classes where you can learn to uh intro to coco and coco basic, and then you can play around on old computers. Um, there's also going to be like a huge marketplace for vintage computers. Um, and I don't know. It just seems like a fun little event there's food pdp8.

2:47:32 - Leo Laporte
There's vendors april 4th through 6th at the info, ageage, science and History Museums on. Marconi Road in New Jersey, in Wall, New Jersey. That's pretty cool. Yeah, no, I know about this. I'm not sure why I feel like people who are on our shows are involved. Yeah, yeah. With support from the Computer History Museum, which I do know a little bit about it says the theme is the sound of vintage.

2:48:01 - Benito Gonzalez
What is it? What does that mean?

2:48:03 - Leo Laporte
well, I couldn't tell you probably that yeah, modem sounds yeah, modem sounds, you know, disc sounds, boot up sounds, that kind of thing. The sounds of vintage jeff jarvis pick of the week well, I should uh a late breaking one here.

2:48:24 - Jeff Jarvis
Um, underline 163, you asked for people to show themselves how they watch us. Well, meme artist showed us how they're watching us there.

2:48:35 - Leo Laporte
Look at that, look at that in the, in the garage or whatever. Yeah, you could post it on. Actually, that's actually a better idea than sending it to us. Post it on, uh, on blue sky. Oh yeah, add us on blue sky. That's the that's. Add us on blue sky, that's what I should have said.

2:48:50 - Jeff Jarvis
Videos less than 60 seconds on blue sky.

2:48:53 - Leo Laporte
Well, that's good, we don't want it too long yeah, he mime artist is in the uk and thank you if she says she's going to do a video for us.

2:49:01 - Paris Martineau
Wow, it's late there that's yeah. Staying up late, that's very sweet you'll be an integral part of the 24-hour live stream yikes yikes schmikes.

2:49:15 - Jeff Jarvis
And then amazon's delivery drones are grounded in College Station, Texas, line 142. Because it scares the boydies.

2:49:23 - Leo Laporte
And the dogs, and the dogs. Well, it scares humans too. What about the humans?

2:49:31 - Jeff Jarvis
Several times an hour, an Amazon Prime air delivery drone would noisily emerge about 800 feet away, just past a line of trees behind smith's home. They said neighbors began calling it the fleet flying chainsaws.

2:49:46 - Leo Laporte
Yeah, that's kind of it's kind of what they sound like like.

Imagine if you come from a modern if you come from a modern, war-torn part of the world and you hear a drum sound ptsd ptsd yeah yeah, yeah it's a very good point so they, they got quieter ones and so on, but they're pulling back all right, all right, uh, I thought I had a pick, but I can't find it, so I guess I'll just use that, uh, that browser as a pick and a reminder. I am leo laporteme on blue sky, sorry, uh, you can, uh, you can uh mention me and post how you watch. That's another way you can do it. We'd love that. That's really nice then. Uh, thank you to all in advance. Um, all right, everybody, we're done, we're through. I hope you've enjoyed this episode of Intelligent Machines. We are setting the clocks forward on Sunday, so the next episode will be at a well, it'll be the same time from our point of view, which is 2 pm Pacific, 5 pm Eastern, but it's 2100 UTC, 2100.

2:50:55 - Paris Martineau
UTC never forgets. They don't let us forget about it.

2:51:01 - Leo Laporte
Our guest be there on time. Our guest will be Ray Kurzweil, the man who came up with the term intelligent machines and the singularity. We'll talk with Ray about that. He says we're about to merge with the machine. Um, we do this show.

Uh, I told you the live time because we do stream this show. Of course, if you're a club member, you can watch in the discord, but also on YouTube, twitch, tiktok, xcom, facebook, linkedin and Kik, so you can watch live. And if you're watching live, we can chat with you live, because I see the chats right here on my screen. After the fact, on demand versions of the show at twittv slash twig. If you go there, there'sa link to the youtube channel great way to share little clips and, of course, where you are and leave a review. By the way, if you go to your favorite podcast player you go to itunes or google play store or whatever make sure you leave a nice review and subscribe. That way, you'll get the show automatically as soon as we're done cleaning it up, because gary said a few bad words hey, hey hey hey, hey did.

2:52:07 - Jeff Jarvis
Jeremy was in the chat. Did you ever be noticed when this is happening?

2:52:09 - Speaker 7
oh yeah, he said hey, the chat, he was going hey hey, um, what else?

2:52:18 - Leo Laporte
I think that's it. Paris martineau writes for the information you should subscribe. It is really even today, two or three scoops from the information. It's a really good source of tech news. Martineau.01 on signal. If you've got a tip for do not use your work phone, maintain security smart about it times. Yes, uh, jeff jarvis, a former professor, not former always, now and forever.

2:52:44 - Jeff Jarvis
He's teaching us every week yes for us whether you want it or not, no, we love it.

2:52:50 - Leo Laporte
We do learn, uh, well, some of us do. I don't, but you know, uh, he is a professor at form, emeritus professor of journalistic innovation at the craig newmark graduate school, journalism at cuny. He also is at suny now was cuny now suny, stony brook and montclair state university. Thank you, jeff. The author of as you see behind him the web we weave. That's his latest, the gutenberg parenthesis magazine in paperback cuny to suny, but never loony jeff uh, but is he a human being? I don't know.

2:53:32 - Paris Martineau
We'll have to uh stay tuned for next week has dropped in the chat.

2:53:40 - Leo Laporte
remind us that the gizFizz is tonight 8.30 pm, eastern, 5.30 Pacific, about 20 minutes from now at gizwiztv. Hi, decky D, I will be back tomorrow for Chris Marquardt, our photo segment. Thank you everybody for being here. Remember Ray Kurzweil next week. We'll see you next time on Intelligent Machines. Bye-bye.