Intelligent Machines 805 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, a brand new show. Well, it's the same old show really. We've rebranded. This week in Google, we're going to cover AI and we've got some interesting stuff to talk about, including Zach Kass, who was the head of go-to-market for OpenAI for a couple of years. He is, by his admission, an AI accelerationist. We'll also talk about the latest AI news. Mike Elgin will be filling in for Paris Martineau. Jeff Jarvis is here. Episode 1 and Episode 805 of Intelligent Machines is next. Podcasts you love. From people you trust.

This is TWIT. This is Intelligent Machines, episode 805, recorded Wednesday, february 5th 2025. Doomers, gloomers, bloomers and zoomers. Well, hello everybody and welcome to a brand new show. That's the same old show this week in Google, now renamed Intelligent Machines, and let me fill you in a little bit before we introduce our guest and our panelists for the show.

Today, the show is still going to have Paris Martineau and Jeff Jarvis. We're obviously going to continue being as silly as we ever are, but what we're going to try to do is get you some expert information about this rapidly changing world of AI. So every week, we're going to start the show with an expert of one form or another, then we'll talk about AI news and after that everything will descend into chaos and become just the same old show it ever was. Jeff Jarvis is here, professor Emeritus of Journalistic Innovation at the Craig Newmark Graduate School of Journalism. You see that hasn't changed At the City University of New York, now at Montclair State University and SUNY Stony Brook. Hi, jeff, hello boss, welcome to Intelligent Machines. I am Im. You are what we're Im oh, im I am. He has a shorter name. I think it's simpler.

0:02:06 - Jeff Jarvis
It's like 60 Minutes. They always start off with the I'ms we're.

0:02:08 - Leo Laporte
I'ms, i'ms. It is Paris's birthday, so she has decided to have a fine meal instead of joining us today. But that's okay, because guess who's with us? Mike Elgin. He's in Sicily right now with his gastro nomad adventures machine. Societyai is his newsletter, so it covers ai, so it's a perfect person to have on.

0:02:31 - Mike Elgan
Thank you for being here. Yeah, it's great to see you. Thank you, leo. Glad to be here, glad to be here on this.

0:02:35 - Leo Laporte
Uh, big change, I think it's, it's auspicious, long overdue well you, you know you were regular on this week in google, but that was back in the days of google plus, yes, yes, old enough to remember. Times have changed just a little bit. So I think really, ai is, um is probably the most important thing to happen in technology, you could say, at least since the internet or the advent of the personal computer, but I would say maybe even more than that, more like since the industrial revolution or the invention of the printing press, jeff Jarvis. Anyway, it's a very exciting area. It's changing like crazy and it's very difficult to understand. Even people working in it, I think, sometimes are puzzled by how AI works. So we're going to bring people on who are working in AI. In fact, next month, the person who coined the term intelligent machines, ray Kurzweil, will join us. I've interviewed him several times in the past and I'm very curious what he thinks he's been saying. The singularity was just around the corner, but now that it really feels like it might be, I can't wait to see what Ray has to say.

Today we've got the former head of go-to-market for OpenAI. Zach Kass worked at OpenAI from 2021 to 2023. It's great to have you. He currently talks to groups and consults with industry about AI and the change that is a-coming. Hi, zach, how are you? It's great to have you. Why did you leave OpenAI? Just out of curiosity.

0:04:13 - Zack Kass
It's not a story that I used to tell very often, but my parents got sick. Oh, I'm sorry. No, it's okay, I'm more willing to tell it now than ever. You can decide if you want to keep it in here. But it occurred to me that I kept talking about this future of abundance and the increasing importance of being human and community, and I wasn't actually honoring my ultimate code and then said okay, I'm going to. I grew up in Santa Barbara, by the way, so moving home also you went home, basically, yeah, I went home for some other compelling reasons, so I moved home and, um, my parents are okay now, which is great. Oh, that's fantastic news, wonderful. Yeah. Yeah, the story, the story ends well and I I met my now wife. Um, oh, even better. Yeah, so it's.

0:05:00 - Leo Laporte
I mean you don't have time to have a wife when you're ahead of go to market at, I would guess, wives. What's that Life? What's that? Maybe a better?

0:05:13 - Zack Kass
question would be why did you join OpenAI? Well, I don't think I could have come up with a reason not to, so many of my friends thought it was a head-scratcher because it didn't have a thing to sell at the time. I, the gpt3, had like a two maybe two million dollars in revenue.

0:05:30 - Leo Laporte
At the time it was a really it was really early in the public. Uh, we had just released gp3.

0:05:35 - Zack Kass
We had just released gp3 and um peter. Well, under the head of applied, the vp of applied pinged my old boss, lucas b Wald, now the CEO of Weights and Biases, and Lucas said, listen, there's one guy who's ever sold, you know, large language models successfully that I know of and it's this guy and they introduced me and sort of. The rest is history. And you know, I would have swept the floors Like I sort of knew. I mean, I didn't sort of know, I knew what was going on. So when people were like, why did you join, I was like how could I have not joined? I probably would have done that job for free.

0:06:11 - Leo Laporte
You've said would you call yourself an AI accelerationist? You've said AI, as I just said at the beginning of the show, is the most significant invention ever, practically.

0:06:24 - Zack Kass
I think it's probably fair to label me an accelerationist. I mean, if I'm not willing to call myself that, then I don't think I'm willing to stand by my belief that. You know, I studied history at Berkeley and if you study history you realize that the world just gets better all the time. And then if you study computer science, you figure out why and you know all of the reasons that we should be worried about technology. I get but like it is true that technology is the single thing that consistently moves the world forward and you know, I certainly believe in its power to continue doing so.

0:06:54 - Jeff Jarvis
So Reid Hoffman's new book just out. He starts off with a taxonomy of AI thinking. There's doomers, gloomers, bloomers and oh hell, what's the fourth? So doomers are the people who think it's going to destroy us all. Gloomers are the ones who say don't forget about environment and all this stuff, you bozos. Oh, zoomers are accelerationists and bloomers which is read is people who do think there's some concerns and some controls, but the benefits are going to be great and let's plant them and see them grow. So in that taxonomy, zach, where would you put yourself? Zoomer or bloomer?

0:07:36 - Zack Kass
Here's my issue with bloomers. So here's my issue with these things. Obviously, no zoomer, unless they're a psychopath, is like disregard, humanity, right. So I think it's a nice taxonomy because it's like evenly divided, but like in reality he you know reed puts himself exactly where everyone would want to be yeah, I'm a little bit of all of the above, depending on my mindset at the moment.

0:07:57 - Leo Laporte
Sometimes I'm a doomer, sometimes I'm a psychotic zoomer.

0:08:02 - Leo Laporte
You know, I feel like let's just, let's just go back and forth yep, let's just let it go and see what happens.

0:08:08 - Leo Laporte
You think that's a nutty point of view, zach?

0:08:11 - Zack Kass
no, but what? What I'm going to say here is, like someone like reed has sort of um. You know, I have a lot of respect for reed for obvious reasons, but reed has like taken some pretty strong positions in the past on things like, for example, donald trump and said, like this is a person who's going to make the world worse, and I actually don't give enough credit to politicians anymore in of robust economies to actually like make the world worse. I think that the problem now is can we like get ourselves out of a lot of situations that I think technology can solve for? So what I mean by this is like the easy thing to say is like I'm a bloomer and what we should do is just like carefully manage AI.

But that just introduces a bunch of excuses for special interest and puts us in like this, like a new set of issues around. You know policy and you know the EU is going to experience this very soon. I also don't think we should run rampant with this thing. I mean there should be some policy, but I think it's so easy to say, oh, we need to. You know control for all these things, when in actuality, we're fast approaching a world where I think the free market ends up solving so many problems that the government promised to solve a long time ago. And that's just going to be one of these weird things where people wake up and go wait a second. Why do I pay 50% into a system that's actually just gatekeeping abundant education, abundant healthcare, et cetera.

0:09:34 - Leo Laporte
So you believe that AI is going to solve these problems. Not just the free market, but AI itself is going to solve these problems.

0:09:43 - Zack Kass
I think that technology solves problems. I think that if you look back long enough, you realize that almost every problem that we've ever faced is not solved by policy but in fact, solved by innovation. Like you know, what actually ends up happening is we just invent tools that make us more efficient and then, over time, these tools sort of make what used to be a luxury a commodity, and then the world just gets better right, like driving down the cost of water, driving down the cost of food, driving down the cost of the internet, driving down the cost of electricity these are the actual things that move the world forward. And the lower the inference cost gets on ai, like the more abundant ai gets and the, basically, the more distributed it gets. And that's my take on broadly technology.

Fusion energy is like, I think, the great unlock to to world peace. Quite honestly, like abundant energy leads to abundant foodstuffs, leads to abundant water, leads to just like. Suddenly people like why are we fighting? We like we, we live in a we live in truly a post-scarcity world. This is how the world works. By the way, I don't think this is like a super hot take don't you uh worry a little bit, though, about the dangers of ai?

0:10:46 - Leo Laporte
are you not worried about that at all? We talked last yesterday. Steve gibson did a whole show on ai jail breaking, and there's not an ai out there currently that hasn't been jailbroken, which means there's not an ai that hasn't been used to uh, for instance, create, create malware, or his example. I said, well, so what? So big deal? His example is well, what if you were able to use an AI to create a toxin that we never heard of before? We had no cure for?

0:11:16 - Jeff Jarvis
You can't anticipate every possible bad thing that any human being can imagine. More importantly, you cannot protect.

0:11:23 - Leo Laporte
Yeah, there's no.

0:11:24 - Zack Kass
Ai safety is an illusion, it seems to be you agree with the printing press so again, I think it is very easy for people to point at these like outlier events that could be caused by technology, and for every outlier event that that technology can cause, there's a solution to that event that technology can cause. There's a solution to that event that technology will create. I mean measures and countermeasures. Is this like proud tradition that AI will observe, just like every other technology? And just the same way that the cost of an attack on a bank has just declined every year because banks are like oh, we figured it out, like we, we're we like know how to stop these things now and we know how to make sure that you know, we, we don't get taken for a bunch of money, the the reality here is, you know, for every toxin that's going to be created, someone's going to come up with an anecdote to that toxin well, they will only if the incentives are there to do so.

0:12:23 - Mike Elgan
So if you think about the application of algorithms and AI to social media, for example, you have a situation where the millions of people who were completely captured by TikTok and you know they rearrange their lives around it.

There are other examples that are sort of outside of AI, like video gaming. There are people who literally play video games until they die, oftentimes in Asia, just completely captured by these things, and this would be solvable by technology for sure, but there are no incentives for the companies to actually solve those problems, because it's not a problem for them. It's not a problem for Meta that people are addicted to its social networks See what I'm saying. So this is a problem that we have with incentives. I mean, I think it's reasonable to say that advanced technologies and probably AI is another example generally as a category, make a lot of things worse and a lot of things better at the same time, and it's all about the incentives to make things better. If we profit or we benefit from people's lives being kind of messed up, then their lives are going to be messed up.

0:13:36 - Zack Kass
Yeah, again, I'm just going to take the other side of this, which is like, if you could reinvent the Internet, you know, un-invent the Internet you wouldn't. Social media is one of these like blights of the internet, but it's a social experiment. And what and what's, I think, fascinating about all these things? If you had looked at the United States in 1950, your safe assumption would have been that we all were going to die of alcohol poisoning. Yeah, and, and like people, people have such incredible recency bias. They're willing to look at social media and be like there's never been anything worse to humanity. And, by the way, I think it's an incredible blight.

I don't use social media. I have someone on my team do LinkedIn and Twitter for me. Fine, it's a cost of doing business for me, but the reality is, what we know is that it doesn't make us happier and humans actually default back to our happiness function. This is why you're seeing an entire generation. Despite the fact that every alcohol company wants to sell you alcohol, mike, there's a new generation behind you that's like hey, I don't want to drink that much. And you're going to see this.

With social media, for every measure there's a countermeasure and things revert back to the mean, and my point is simply let's create optionality, let's create a world where people can do whatever they want and actually trust that people in the end sort of align to their own best interest, which actually, it turns out, is time with friends and family and physical community, and we keep coming back to that place, which is why we're creating more and more better alternatives to the Internet all the time. Concerts have never been more popular. Sporting events have never been more popular. Theater has never been more popular. Like all more popular Sporting events have never been more popular. Theater has never been more popular. All these offline events are sort of resurging because people are like hey, you know what? I actually don't want to doom scroll, I don't want to drink, right. This is happening already.

0:15:15 - Leo Laporte
We're talking to Zach Kass. He is the former head of go-to-market for OpenAI. He's a consultant in his new book. When is book gonna come out? It's called the next renaissance ai and the expansion of human potential. End of march, end of march. So it's done. You finish writing it.

0:15:35 - Zack Kass
Yeah now, yeah now. I just sit there and regret everything jeff has some experience, uh, with that.

0:15:44 - Leo Laporte
Uh, the subtitle is the next renaissance, ai and the expansion of human potential. You're very bullish about ai, you say. Thanks to the coming emergence of artificial general intelligence, this chapter in our species story will be marked, above all, by a massive surge in human potential and productivity, which will unlock profound advances in science, technology, health, education, robotics and the political economy. I love it. Let's be optimistic.

0:16:13 - Jeff Jarvis
What's your?

0:16:14 - Zack Kass
definition of AGI. Oh, that's a funny one, Jeff.

0:16:17 - Jeff Jarvis
I don't know. It's kind of vital to this discussion because nobody really ever defines it.

0:16:21 - Zack Kass
You know, what's interesting is, it's the one thing that I wish we hadn't even mentioned in the book. But at the what's interesting is, it's the one. It's the one thing that I wish we hadn't even mentioned in the book. But, like at the time, it look it's a marketing term. It's clearly a marketing term. We shouldn't call it anything else. I I consider it human intellectual equivalence. That's probably what I would call it, human intellectual equivalence and, honestly, we're probably there.

0:16:39 - Leo Laporte
If we're not there, we're awfully close, especially if you take a look at the reasoning models.

0:16:43 - Jeff Jarvis
Yeah, John LeCun says we're lucky if we get to in the near term to be a cat but John LeCun has died on this Hill.

0:16:51 - Zack Kass
I don't think he you could walk him off this position.

0:16:53 - Jeff Jarvis
I mean I think he's right, I think he's absolutely right. I think that that until it gets a sense of reality and real life and knows that when the ball falls off a table, it's actually still a ball, it's not human intelligence, I think.

0:17:10 - Zack Kass
By the way, I don't think a cat can create a video. I'm just not sure that that's right at all.

0:17:17 - Jeff Jarvis
Oh, if we write the right tool for it.

0:17:19 - Mike Elgan
This is an argument that we're going to hear a lot on this show over the next few months and years until AI actually is provably here, and then jeff will have to say well, I, I guess I was wrong go ahead like, yes, a cat can't make a video, but but a uh an ai robot can't play with a ball of string.

0:17:41 - Leo Laporte
So I think I think it's reasonable that which is more useful, but that's well to the cat.

0:17:46 - Mike Elgan
The string was way more useful.

0:17:48 - Zack Kass
We could, we could program a robot to play with a ball of string.

0:17:51 - Mike Elgan
We certainly you could, you could make you could make a robot simulate playing with a ball of string, but it's not playing. Does it enjoy? Having any fun but, my point is the problem? The problem with agi, I think, obviously is that is that ai is already way better at humans at a whole bunch of things, and nowhere near as good as humans at a bunch of other things. So the math is impossible. You can't basically say, well, in general, on average, we've reached AGI.

0:18:19 - Jeff Jarvis
That's why you're right, zach, the term is difficult.

0:18:22 - Zack Kass
It's just a marketing term.

0:18:24 - Jeff Jarvis
Let me ask you another challenging question, which is as I'm curious of your experience at OpenAI and what happened there with the whole board, kerfuffle and everything else the question of safety, the definitions of safety, and it's always struck me that the doomers have kind of ruined the definition of the word because they overblow that too. Oh my God, we can destroy mankind and talk about P-Doom and all that. Where did you see yourself in the culture of OpenAI around those questions of doom and safety?

0:18:58 - Zack Kass
None of this is going to surprise you. I mean, you know, you just have me by the way. Have me back when something goes terribly wrong to shame me. I suppose Even then I'm going to die on this hill, jeff, and here's my, here's my only point. If someone's like how you know, how are you so sure this is going to go right? If you believe the world's going to get a whole lot worse, the onus is on you to prove why, because the we crawled out of caves once.

0:19:28 - Jeff Jarvis
I mean on you to prove why, because the we crawled out of caves once. I mean where we have come, like these days. I want to go back in, but keep going.

0:19:32 - Zack Kass
See, I actually I know your tongue in cheek, but I think there is something so pernicious about the idea that we are somehow regressing because, in fact, the reality is, today is the best day ever to be born and tomorrow will be. I agree, I agree. But, jeff, I think your point is like an interesting one, which is, a lot of people hearken back to these romantic days of old when, in fact, we have made so much progress from the time when you know my dad's favorite, you know in Kennedy. Oh, I wish we still had Kennedy. Yeah Well, we had a bunch of other things that were pretty messed up.

0:20:11 - Jeff Jarvis
That's why I don't buy P-Doom actually. That's why I think that that's an egotistical effort by certain AI boys to say I'm so powerful, I can destroy mankind. That's BS and I think mankind is smarter than that and we'll figure this out.

0:20:20 - Zack Kass
I look, that's what I said. It's not gonna surprise you. I totally agree, and I think if you believe that the world is going to end, you have a major onus to prove that we are going to rebuke a lot of history. You'd have to reverse an incredible amount of historical precedent to actually believe that.

0:20:39 - Leo Laporte
One of the things you do since you've left OpenAI is you consult with companies like Coca-Cola, morgan Stanley, the White House. What do you tell them If I were an executive at a big corporation? What do you tell them about how they should prepare for this new AI-driven world?

0:20:59 - Zack Kass
The advisory business we run spends time with a lot of different governments, all democracy well, mostly democracies in the world or US allies, and we also support a lot of Fortune 1000 companies across sort of policy, science and applied and sort of the breadth of you know how the technology is gonna impact them. We work with, like you know, a group that focuses on their social security and one of the questions they're asking is when are people going to retire in the future? And actually does the retirement age make sense in a world where technology is so easy to use, as it's getting easier? Can we actually wean people into retirement instead of slamming the book shut on their working chapter? So you know we touch sort of everything. So you know we touch sort of everything. What we try to do is point to trends and interesting events on the horizon that give people some way to ground where things are going, because for most of the boardrooms and the executives we work with it's just whiplash. I mean, it's so overwhelming.

0:22:01 - Leo Laporte
It is for us. I can imagine what it is if you're trying to run a business in this. That's exactly right no-transcript it's a um challenging world for all of us, and I think the biggest challenge at this point is under even understanding what's happening. Do you feel like you understand what's happening, zach? Well, do you?

0:22:43 - Zack Kass
feel like you understand what's happening, Zach.

0:22:49 - Leo Laporte
Well, that's a great question. It feels fairly unpredictable to me. I mean, I agree that the arc of technological history bends towards good things, but there's also a few bumps in the road along the way, inevitably.

0:23:07 - Zack Kass
Okay. So from a scientific standpoint standpoint, I think it's actually pretty clear to understand what's happening. I mean, we were just getting, we're building, what we observed. Early on, boris power, the head of research, engineering and open AI, coined this term of unmetered intelligence, which is how we explain what we, what we were building towards, to executives that were exploring open AI. And unmetered intelligence is like a thing that I just is a term that I love to use. It's, like you know, free intelligence. The path to that makes sense to me totally. How humans respond to it doesn't like no Unpredictable. Yeah, I cannot tell you.

One of the things that we study in the next renaissance and one of the things that I study at UVA, and one of the things we talk a ton about, is this new divergence between a societal threshold what do we want a machine to do and a technological threshold what can a machine do? Because these things used to sort of happen in parallel. There was a convergence and a symmetry between these two things. We invented something out of necessity and we used it right away, and then we started to have this divergence in this post-scarcity, post-barbarism world. The obvious, best example is nuclear power. We invent nuclear power and we go oh, we have enough coal, we don't want to run the risk of a nuclear reactor exploding, so let's just shelf this technology. And of course we regret that now. But it's very interesting to consider that the future is actually not going to be informed by what a machine can do. That question won't become interesting in a world where machines can do most things. It's going to be informed by what we actually want that machine to do, and societal thresholds that we are willing to cross become the most important thing. And so that's where actually a lot of our time is now spent on the research side and on the advisory side, helping businesses figure out do their constituents actually want this thing?

0:24:47 - Leo Laporte
I know you've got to get a plane we're going to link. We're going to ask you one last question because we've got to let Zach go. I'm sorry, Jeff. Deep Seek kind of set a shock through the AI industry. Sam in an AMA on Reddit said we were on the wrong side of history. What's your reaction to what happened with DeepSeek?

0:25:08 - Zack Kass
I got kind of lucky because I read the DeepSeek paper in May 2024 and started talking about it to people, and then I got invited to UBS's Greater China Summit in Shanghai in early January and spoke about it there, because my message has consistently been these models are converging, their frontier will lead, but it's not. You know it, there isn't. There is a clear commoditization happening at the research level, in part because the inference costs are just driving down so far, so fast, and so I got a bunch of points when that that came about. And then a bunch of people called and and my answer is really simple and this is, like I think, where we can leave it.

There is a huge difference between DeepSeq R1, the model breakthrough, which, I will remind everyone, is open source. It's clear improvement in the technology. We should consider it a win insofar as it's challenging the norm that the Frontier model costs a lot, and deep-seek the app. And deep-seek the app is hugely problematic and something that we should actually seriously concern ourselves with with respect to national security. Because it exfiltrates everything in China.

Yeah, I mean the CCP has requirements around all app developers. I mean, I've never downloaded a Chinese company app TikTok, shein, tfue, red Note. None of these apps should be downloaded by a US user, quite honestly, simply because the CCP gets all the data. It's not even that you don't have to trust the app developer. It's just like kind of a risky position that you don't have to trust the app developer. It's just like kind of a risky position. But the model R1 is a huge win and we should all celebrate it. In a world where everyone's clamoring for more ecologically friendly AI and more fair and distributed AI, this is a great signal that, in fact, we're not going to live in a technological hegemony.

It confirms in some ways your optimism.

0:27:04 - Mike Elgan
Go ahead. Mike, you use Perplexity. I think you're a pro user, right and one of the models they use now is r1 and yes, that's, I used it this morning I was talking with paul thorat.

0:27:15 - Leo Laporte
He says I was watching tv, he's in mexico and the president of mexico said something about changing the way we vote, but I couldn't find the news story about it. I asked r1. He asked, by the way, co-pilot, uh gemini gemini uh and chat gpt, and all of them said hey, we can't talk about elections. A lot of safety there.

0:27:33 - Mike Elgan
Uh r1 had no problem with it and gave me a very nice summary of all I grilled it about tiananmen square and and all this stuff and don't ask about all the right answers no. No, it's fine. Zach has my point.

0:27:44 - Leo Laporte
We gotta let zach go, he's got all right hey, zach, it really is a pleasure. We'll have you back soon when you have a little bit more time. The book the next renaissance ai and the expansion of human potential comes out in march, and his website is zach kasscom, z-a-c-k-k-a-s-scom.

0:28:02 - Zack Kass
Thank you, zach, make your flight thanks for having me guys looking for forward to coming back and you and we can. We can talk about all those things that are going right and wrong.

0:28:11 - Leo Laporte
All right, with any luck, there won't be too much of the latter. See you guys. All right, you're watching intelligent machines brand new show. It's really the old show. This week in Google we just we did a little Google ectomy and we found we've been talking a lot about AI in general over the last couple of years. It is, I mean, I think we all agree, probably the most exciting area right now in technology and, mike, you've done the same thing with your Substack. I mean, you've focused, you've refocused on AI as well. So we thought we needed a show that covered this, and I thought the name Intelligent Machines was apt instead of something about AI, because we're really not just talking about AI chatbots, we're talking about computing at the edge. You're surrounded now by machines with intelligence of all kinds.

0:28:59 - Mike Elgan
It's fascinating to me. It's fascinating to me the obsession with the chatbot end of AI. I do enormous research for my foundry column and also for Machine Society, and I'm looking at all aspects of AI and literally it seems like, in terms of the news, the breakthroughs, what's literally new, 90% of it is all kinds of incredible science that's happening medicine, serious advancements in diagnosing diseases and also doing studies using AI and so I think the public has a very skewed view. When they hear AI, they think chat, GBT, but actually that's, in the larger scheme of things, a kind of a rounding error in terms of how important it is. Ai is is being applied very, very broadly and, in fact, in some really interesting and fun ways that I'll talk about when we get to my thing of the week, oh good.

0:30:02 - Jeff Jarvis
Very good, that'll be about four hours from now, but no, no, that's the other thing.

0:30:08 - Leo Laporte
I think we're going to try to make this show a little bit more concise. We'll see. Good luck. Uh, paris has the day off. Happy birthday, paris. She'll be back next week. Are you going to not be here, though, next week, jeff? No, I'm here, I'm here. Okay, I know who's not gonna be here? I'm not gonna be here. What are you doing?

0:30:23 - Mike Elgan
I knew it was somebody jeez.

0:30:25 - Leo Laporte
Lisa and I, uh, for her birthday and our anniversary, are going to tucson, arizona, for the international gem and mineral show. I don't know. I think you both know that lisa's kind of into rocks. She married me in your head, yeah, yeah, so uh, that was just a musical reference.

0:30:45 - Mike Elgan
She married me. It was in your head.

0:30:46 - Leo Laporte
yeah, yeah, I thought that was just a musical reference. That also. Anyway, I won't be here next week, but the whole team will reassemble in a couple of weeks. We're already booking some really interesting guests. I'm very excited about Rick Kurzweil. I had hoped to have him for the inaugural episode, since we're stealing his name, but he couldn't make it. But he will be here.

0:31:06 - Jeff Jarvis
There is no theft, only tribute.

0:31:08 - Leo Laporte
It's a tribute. Let's just say that we're going to take a little break. We'll come back. We have AI News and I want to ask Mike why his headline at Machine Society is why you can deep six the deep seek hype. I think that'll be an interesting conversation as well. You're watching Intelligent Machines brand new on Twit. We're so glad you're here for the show. If you are already a subscriber to this Week in Google. It's the same feed, the same webpage. Nothing else has changed, just the name and the album art, which I anthony nielsen for, and we finally got rid of those damn flutes in our theme song.

I think that was the main reason I did this, and thank you to benito gonzalez, who created the theme song you heard at the beginning of the show. He is, of course, also our producer, technical director, editor, editor and all about man About Town Our show today, brought to you by Zscaler, the leader in cloud security. You know the world is changing, isn't it? Ai is, of course, part of that change, but what we used to think was great security with these firewall perimeter defenses I mean, that has been, up to now, the default way to protect your enterprise. We've spent billions on firewalls and, of course, if you have a firewall, you need a VPN right, so the employees can get into the network. What has that done to improve our security? Well, zach notwithstanding, breaches continue to rise. There was an 18% year-over-year increase in ransomware attacks and a record payout of $75 million to ransomware goons in 2024. The truth is, traditional perimeter defense security tools don't help. They expand your attack surface. Those VPNs have public facing IPs and, of course, the bad guys are working faster and better than ever because they're using AI tools to attack. It is a big problem right now. Vpns and firewalls also allow lateral movement. They just assume if you're in the network, you're one of us, so once the bad guy gets in, they if you're in the network oh, you're one of us. So once the bad guy gets in, they can go anywhere in the network, look for valuable information your emails, your customer information and so on and and encrypt it and send it out through the firewall, because firewalls have a real hard time. Uh, inspecting encrypted traffic at scale, you can see this is a disaster waiting to happen. And then just add on top of that the fact that hackers are exploiting this traditional infrastructure using AI. They're outpacing your defenses. It's time to rethink security. We can't let these bad guys win. They're innovating and exploiting your defenses.

You need Zscaler, zero Trust and AI together Stops attackers. Here's how by hiding your attack surface so your apps and your IPs are invisible, bad guys can't attack what they can't see. They also solve this lateral movement problem because Zero Trust does not assume that anybody inside the network is a good guy. Users are only allowed to connect to specific apps that they have explicitly been allowed and authorized to use. They can't go anywhere in the network and access any data. Plus, zscaler continuously verifies every request based on identity and context, continuously verifies every request based on identity and context, and it keeps a log so you know what's going on Simplifying security management with AI-powered automation. In fact, it needs AI because Zscaler analyzes half a trillion daily transactions. Now some of those are malicious and it uses the AI to find those malicious needles in that giant haystack. It's really a remarkable tool. Hackers can't attack what they can't see. You can protect your organization with Zscaler Zero Trust and AI.

Learn more at zscalercom slash security. That's Z-S-C-A-L-E-R or, for our Canadian listeners, z s-c-a-l-e-r. Dot com slash security. Thank you, zscaler, for supporting the brand new show. We appreciate your faith in intelligent machines. We're really glad to have you and you support the show. When you go to that address, zscalercom slash security uh, there is uh, let me, let me stop you real quickly.

0:35:45 - Benito Gonzalez
Uh, I need your, uh, your screen share oh yeah, of course you do, no wonder.

0:35:50 - Leo Laporte
No wonder I've been pulling stuff up. I rushed so fast to get into the meeting and get zach out of here, okay, uh, if you uh, mike and jeff, if you're seeing a firefly, if you're seeing jeff's, if you're seeing Jeff's newsletter, you just click the tab at the top that says whatever it says Meeting. You should see me. How is it now? Benito, you can see it. Yep, good to go. I don't know if it's a coincidence or if our timing was good, but there's a lot of AI news this week.

0:36:27 - Mike Elgan
I mean, there's always a lot of ai news if you, if you go looking for it, did you even look down?

0:36:31 - Jeff Jarvis
at the lower half of the rundown uh, oh my god, how, how many hundreds of lines of stories do we have um?

0:36:39 - Leo Laporte
all right, well, let's see. Where should we? I could just start giving you headlines bit by bit.

0:36:44 - Mike Elgan
Well, I could tell you why deep seek hype is overhyped. Yes, this is the front page of machine societyai.

0:36:57 - Leo Laporte
That's right, you say a fast, cheap, open source, llm-based AI chatbot has emerged from China. Here's what's bugging me about all the hype. By the way, you may say it was hype, but the stock market believed every bit of it. Well, that's part of the hype yeah that's part of the hype.

0:37:14 - Mike Elgan
So the hype is based on three data points. One is the stock market. It wiped out a trillion dollars of value from USAI companies. No, it didn't.

The stock market is fickle and and most of them are not a good yeah, it's not a good metric for anything that it's. It's basically just everybody's sort of speculating and and knee-jerk reacting uh, so that was one data point. That that is is always overblown. Uh, no, trillion, a trillion dollars of value was not lost by US companies. The second one, the second data point, is the downloads. They said, wow, this is, there are more downloads than ChatGPT. Well, of course, chatgpt has been available for months. Everybody who wants ChatGPT already downloaded it months ago, and so if you look at the downloads this week, you're likely to see the thing that just came out that everybody's talking about. That's in the news.

Of course, the download metric is completely irrelevant. You notice that the breathless media coverage didn't talk about total downloads over the last eight months. It's about this week. It's skewed information, but the biggest one, which basically is information that came after I wrote my newsletter, is that the data point of $5.6 million, which was bandied about as the training cost for DeepSeek, is a completely BS number run over two months. In fact, it didn't include research and development, infrastructure, hardware acquisition, data preparation and a whole bunch of other things. It didn't include the salaries of the people working, didn't include any of that stuff and if you do include that stuff, the cost of developing DeepSeq was between $500 million and $1.6 billion.

0:39:00 - Leo Laporte
A little closer to the real thing, but still cheaper.

0:39:04 - Mike Elgan
Still cheaper, a little closer to the real thing, but still cheaper. Still cheaper, yes, but it's not 5.6 million, and so I think the shock of that low number is sort of like part of the hype. As we can see now, with the dust settling turns out, deepseek didn't really make that big of an impact in anything. It's a new model, it's being integrated, it's being folded in, like we were talking about earlier, into perplexity. It's nice. It's not quite as powerful as some of the other tools that we have, and it's also interesting that the unlike TikTok, where the hand of the Chinese Communist Party and of Chinese government, censorship and all the rest is sort of hidden and difficult to tease out with any sort of accuracy, it's clear to see that that the, that the app that the deep seek made available is a, is a commie propaganda. I mean, it's like you ask it about tm square and it just it gives you an answer, then it wipes it away and gives you yeah, that I've seen the videos of that.

0:40:13 - Leo Laporte
That's pretty amazing, yeah, but that's. But you expect that we do the same thing. As I said, when you ask co-pilot about, uh, mexican election proposals, they say, oh, let's not talk about elections. I mean mean, that's the safety part of it, right? And the fact that we don't, in the United States, talk about elections is similar to not talking about Tiananmen Square in China, right?

0:40:35 - Mike Elgan
It's not the same, because one is coming from the policy of the company that basically doesn't want to have to go before Congress and talk about why their chatbot tells people what a clown the congressman is that they're talking to, and stuff like that. It's a policy based on the company, and the companies vary, so that's one of the things I like about perplexity. You can ask it all kinds of topics that are unpleasant. You can ask it about Trump, you can ask it about all these things and it'll give you a answer and it's it's really kind of nice, whereas the chinese communist party this is a decades-long policy of controlling what people know I understand all that.

0:41:14 - Leo Laporte
That's not what's interesting to people about deep seek. It's it's it's that it it's that it's using it was using clever hacks to get around. Yeah, the in the uh using it was using clever hacks to get around. Yeah, the the uh injunction against using, uh, the highest level nvidia cards. I mean, if we were, if we're to believe. By the way, this is and I brought this up before we don't really know if they wrote assembly language code to replace the cuda cores a code. Uh, we don't know exactly. Maybe they did, in fact, get around the embargo and buy the top of the line nvidia chips, but just didn't tell anybody. None of this can be better because it's china.

0:41:51 - Mike Elgan
We know that there are tens of thousands or hundreds of thousands of nvidia chips in china that are just smuggled in ten at a time so maybe it isn't so uh impressive, but maybe I think.

0:42:04 - Leo Laporte
I think sam allman thinks it is. I think people think it. I think Sam Allman thinks it is. I think people think it is impressive that they've done so much with so little. I guess.

0:42:10 - Jeff Jarvis
Let me. Let me argue where the hype's done good. Even if it is BS, it's done. It's done us well because, number one, it's added a competitive fever, and I think the competition is good. Number two, it's causing others to put their stuff out for free out of fear, and I think that's a good thing. It's adding more open source and it's validating open source.

And it's also the big one to me is it's questioning this American macho chest beating idea that everything must be huge and if it is huge, it must be the best because it is American and it is huge and we have more compute than anybody and we have more money than everybody. And if it is huge, it must be the best because it is American and it is huge and we have more compute than anybody and we have more money than everybody and you have to have capital to do anything. So, by questioning that ethos of American AI technology, I think that's healthy for the discussion, Whether or not I'm not disagreeing with you, Mike, necessarily that the hype may be BS, but I think it actually did some good.

0:43:06 - Mike Elgan
I think it did some good too, but I think it actually did some good. I think it did some good too. On on balance, it absolutely did some good, especially since it's open source and you can take it without the, without the propaganda right you can, yeah, and in fact that was one of the things that sam alton said we were on the wrong side of history.

0:43:20 - Leo Laporte
He said we should have been more open about our weights. You know, one of the things OpenAI was founded to be is open about how it was doing what it was doing, and they, of course, backed off as soon as they decided to become a profit-making entity.

0:43:32 - Mike Elgan
Yeah, I mean I wrote the newsletter at the peak of the deep-seek hype and panic and I think that it's clear now a few days have passed, that it really didn't change that much. All that stuff about attitudes in Silicon Valley in the United States, about the money and blah, blah, blah those are very much still with us. That hasn't been erased or washed away. It has been added to the mix of all the options that companies have and that users have for AI, but it really isn't the apocalyptic event that it was being made out to be. What I do think is most interesting about it and I've been writing a lot about this in recent month or two is that we've had a sort of that.

We've had a sort of a bias in media that you know basically was in cable news. You know basically used to be Walter Cronkite when he had cable news. Then the news outlets went left and right and people would choose their, their point of view as they wanted. Then that hit social media, especially when Elon Musk bought X. A lot of the lefties left X and went to blue sky. A lot of the right-wingers who didn't used to do Twitter or X went to X because now it's a friendlier platform for right-wing media, and now it's increasingly happening in AI chatbots. So there've been a couple of studies that found that in general, many of the mainstream chatbots have started out sort of center-ish and have gradually moved left, as they do moderation. And then there are other chatbots.

0:45:15 - Leo Laporte
like Pons expolutes everything, doesn't it? Ay-yi-yi yes.

0:45:20 - Mike Elgan
But there are others, like Grok and so on, that are kind of like kind of on the right and so and that's because that's the elon's you know the app, at least exactly. So we've entered into an area, era that, like people, haven't been commenting on enough, I think where where the chatbots are increasingly, uh, politically biased, if for lack of a better term.

0:45:47 - Leo Laporte
You said you like perplexity the best. I have to say that's the one I use the most often because it lets me choose from whatever model I want. I did just out of curiosity. I thought I was asking perplexity about the Mexican plans to change the voting system. I thought I was asking deep seek. But I think perplexity intervened because when I did ask deep seek, natively it said well, you should. I don't know much about what's going on. I know who the president is, but I don't know much about what's going on right now. So you should probably go elsewhere to find that information. So obviously perplexity either substitute a different model or fed some information to DeepSeek that it then regurgitated.

0:46:27 - Mike Elgan
That's surprising. I've not encountered anything like that before. It's usually pretty straightforward. I mean, it does lie a lot and give me weird hallucinatory information, but a lot of that is just you have to know enough about the topic that you're asking it to say no, that's not right. Try again and you know eventually, and you can enable in deep seek the search function.

0:46:49 - Leo Laporte
I don't have that capability within perplexity. I should probably go into deep seek and do it. In any event. Paul Theriot was mocking me because I have a folder full and I pay for Gemini, gemini deep seek. I don't pay for that, obviously, but uh, I pay for all the others chat, gpt, anthropic, and I use them all depending. In fact, lisa was mocking me this morning too.

0:47:12 - Jeff Jarvis
Well, I'll bet because she's gonna pay the bills well, it's only 20 bucks each. But they're not paying a 200 a month chat gpt.

0:47:18 - Leo Laporte
No, I didn't pay for the professional uh that I was tight times I'm actually I don't feel the need to. I I feel like, especially the reasoning models like 01 and 01 High are remarkable.

0:47:32 - Jeff Jarvis
But you have somebody that are free now. I mean, this is the story you have in the rundown that one thing that DeepSeek does is validate Meta's open source view, and I think Meta's free version is good, gemini's free version is good, and they just released, uh, the game face as a version. Yeah, oh, gemini just released it's 2-0 to everybody, uh yeah, let's talk about that.

0:47:53 - Leo Laporte
Have you played with 2.0 I?

0:47:55 - Jeff Jarvis
haven't yet. No, have you.

0:47:57 - Leo Laporte
No, I was too busy being on zoom all day how about, uh you mike any thoughts on, uh, the new version of google's ai, which? Is, I haven't I haven't played with it yet either, so I have two. I'm looking at my. I also pay for gemini. I have 2.0 flash for everyday tasks, plus more features, 2.0 experimental advanced. Ooh, should I ask it about the Mexican changes in the Mexican voting plans? All right, that might be an interesting test we can start using Because it's fairly topical, right.

0:48:37 - Jeff Jarvis
What is the stack here? Where does advanced come in versus 2.0 and all the others? Do you know? I don't. Like stack in terms of how uh high to low versions oh, I have no idea.

0:48:52 - Leo Laporte
Branding, google and branding, yeah. They're never clear, are they all right? So here we go. I'm going to paste the same question into a gem, typically because this is breaking news in Mexico. A lot of these don't do very well. You know, chatgpt4 wouldn't because its model isn't up to date. Let's see if 2.0 experimental advanced. No, look at this. This is the same answer we got from the other American AIs. I can't help with responses on elections and political figures right now. While I would never deliberately share something that's inaccurate, I can make mistakes. So while I work on improving, you can try google search.

0:49:34 - Jeff Jarvis
So it lacks access to real-time info so I'm on a google ai studio is, I think, where you try gemini 2.0 and and the query at the top is what will you build? Which is so it's aimed at coding. Then push Gemini to the limits of what AI can do using Gemini API, so it's more API oriented yeah, I mean, I think one thing I would say is that there there isn't much skepticism.

0:50:03 - Leo Laporte
Maybe there is on your part, mike, I don't. There seems to be little skepticism in the coding community about the value of AI coders. In fact, mark Zuckerberg says he hopes to replace coding engineers with AI this year.

0:50:20 - Mike Elgan
Yeah, it's generally received as an unqualified good among developers, because one of the things that it can do is basically just show you an idea, and a lot of developers you know good developers will look at the results they get from from ai and say, huh, okay, yeah, that's an interesting approach. Or you can throw out a bunch of code and say what's's wrong with this, where's the problem here? And it basically just like it does with non-coding tasks it's just an accelerant to get to an answer that people who already know something about the subject can benefit from, and those who don't know anything about the subject they can just. You know people who are very you know, relatively unskilled developers that can just take what, uh, what ai gives them, plug it in and you know how the time it just works and they move on.

0:51:11 - Jeff Jarvis
So it's you know, leo, I was thinking about this because, uh, going into today's show, um I, I think code will become even more invisible to most people than it is today. That is to say that that now we get applications and behind those applications is code, and we know it but we don't use it. But now the fact that you can build something, as Gemini just challenged us with our native human language, and it'll make code to do the thing you want, but you're not. I don't have to execute that code, I don't have to look at the code. All I care about is whether it gave me what I wanted, right. So code becomes um inside and irrelevant to most people.

0:51:51 - Leo Laporte
I think, more and more and more I read a very fun uh blog post by a coder who said you know many areas, uh, art, music. People say, well, you know, let the ai do the everyday stuff, but there will always be a market for human created, artisanal, you know whatever paintings or clothing or whatever he said. That's not going to be the case with code, because nobody knows what. You know how cleverly crafted the internals are, just as you're saying Jeff, saying jeff, of a piece of code.

They just care whether it works or not yeah, so that that's exactly the problem here is is you know? That's why ai can reasonably be expected to write good code. It is a computer, after all. Talking to a computer, uh, and it's that's not such a difficult thing to do. It's much more like, say, becoming a chess grandmaster than it is like writing of the great american novel I think that's fine. I don't want ai to do art. I want us to do art.

0:52:53 - Jeff Jarvis
That's what we do well, uh, jason and I had on lev manovich uh, last week on ai inside, who's a a city university of new york graduate center artist and um does amazing work.

He has a new book out and he uses AI as a collaborator and the interesting thing he said to me when I was asking about this is that the good part comes when he disagrees with the AI, when it's conflict.

It's like having an editor or a critic and you know why did you do that and that was wrong. But it's pushing me to some idea I didn't have or it's not doing what I wanted to do, and that's when the creativity comes, which is really interesting. By the way, the Copyright Office, which in the past has been saying AI-created material cannot be copyrighted because it's a machine that made it, they've come out with a new report which is in the rundown part two of its report on legal and policy issues, rundown of the part two of its report on legal and policy issues saying that, um, after considering extensive public comments, uh, our conclusions turn on the centrality of human creativity to copyright, so that if a human is involved in the creation with ai, then it could be copyrighted human not the ai right, right, but it also goes to what you just said about the programmer.

Uh, how do we know that the humans involved? How do you prove the human involvement? How important was the human involvement? Is it just enough to say I wanted to do this?

0:54:10 - Leo Laporte
well, we don't copyright code as it is, that's right, anyway, yeah, but I just said creation, yeah, yeah by the way, here's the table from uh google on uh gemini, the capabilities of the different gemini gemini models.

There's Flash and Gemini 2.0 Pro and they're really promoting the Gemini 2.0 Pro as a coder the best model yet says Google for Coding performance and complex prompts. You know, I've seen and you've probably seen it too on X and other places people posting prompts that created one-line you know, one-line prompts that created JavaScript programs for bouncing balls and squares and things and the physics is good, deepseek in many cases is not as good as some of the uh models from uh open, ai and google or anthropic, which does a very good job of coding. Let's see if I can find some of those. I don't think writing code is going to be is, I think at this year it's. I don't think mark's wrong, I think zuckerberg's wrong, I think this year we're. We're going to be able to say at the end of the year, by the end of the year, yeah, computers should be writing code.

0:55:22 - Jeff Jarvis
Why should it opens it up for all of us to make computers do what we want them to do. Right, that's the power of this uh does.

0:55:30 - Leo Laporte
Do you have a problem with that? Does anybody have a problem with that? That seems like it's a good use.

0:55:33 - Benito Gonzalez
I mean because I don't code I don't have a problem with that, necessarily, but I think that that's kind of a little bit of a fantasy, because the problem is that most producers or people who want something done, like managers and VPs, they don't know what they actually want, so they don't know how to ask the computer what they want to ask.

0:55:51 - Leo Laporte
The interface between a human and the AI might be the problem, but that's why, everybody's saying learn how to write prompts.

0:55:58 - Benito Gonzalez
It's not just about writing prompts. They usually don't even know what they want, right, and it's usually the coder, that kind of decodes what susses out what they're actually asking for.

0:56:10 - Jeff Jarvis
That is a big part of the process. Yes, you know I couldn't agree more. I'm going to go back to an Uncle Jeff moment here. At the beginning of CondéNet, when we started Epicurious, we had the very first meeting with a technology house. We hired and the technologist said okay, tell us every page you're ever going to want. And the editor said are you knocking futz? We'll know when we see it.

And it was the extreme cultural shift there, where we do things in editorial. It's just iterative and we'll play with it and it's creative and we'll do something more and something more. And the technologist said, no, I need a spec. What are you doing?

0:56:45 - Benito Gonzalez
Exactly, and the AI is going to ask for a spec.

0:56:47 - Jeff Jarvis
Yes, because otherwise you're not going to get what you want. If you can't express what you want, well, you're not going to get it. And that's what coders have said to people for years. And now AI is going to be like a coder and the coders are going to sit back saying I told you You're still going to need the coder.

0:57:01 - Benito Gonzalez
Well, think about the iteration though.

0:57:03 - Leo Laporte
Think about the iteration process, because if a manager could say, well, I want something like this and the AI does it, he can iterate with the AI very much more rapidly than with a human coder In that sense.

0:57:14 - Jeff Jarvis
Yeah, it's like what Dan did by inventing spreadsheets you could ask what ifs.

0:57:27 - Leo Laporte
If Steve Jobs called the computer a bicycle for the mind, I think AI might be a race car for the mind. And actually the analogy is perfect, because one of the reasons Steve said a computer is a bicycle for a mind is because the bicycle is the single most efficient mode of purely human transport right, our own energy energy. It converts our own energy most efficiently into movement, whereas a race car does not. A race car burns fuel, but it does get us much faster from point a to point b, and I think that in a way, that's kind of what we're talking about here with ai. It is maybe not the most efficient, but it is maybe the fastest way. I I don't. I I think that there is a certain amount of gatekeeping humans inevitably do because we'd like to think that we're special and that our capabilities are unique, and then a machine can't duplicate it.

0:58:15 - Jeff Jarvis
That's the basic problem with AGI is that we think that the benchmark is to be us. Why should that be the benchmark? As Mike said earlier, it can do all kinds of things that we can't do well.

0:58:24 - Leo Laporte
Well, I think we're going to agree with Zach and stipulate that AGI is a marketing term and not worry about it from now on.

0:58:30 - Jeff Jarvis
Oh no, it's going to come up again and again.

0:58:33 - Leo Laporte
No, every time it does, I'm going to say it's a marketing term.

0:58:36 - Jeff Jarvis
Look at the OpenAI rebrand. It's a pretty font, but look at what they said in their little description of it. Okay, Now you got to find it. I think it's in your list.

0:58:48 - Leo Laporte
Yeah, I have to find it because when I go to OpenAIcom I get a chat bot. So let me see. Oh, here's their blog, introducing you know what. It's funny because I saw that story and I looked at the the new logo and the old logo and I couldn't really see any difference. So I thought I was being punked. Is there really a difference between the new logo and the old?

0:59:11 - Mike Elgan
yeah, the the font is a little bit different, but there's a slider that lets you look at both of them, and and um, yeah, the tilt is a little different the weight varies.

0:59:21 - Jeff Jarvis
If you, if you go to line 101, leo, there you'll find what I'm talking about uh, here's from wallpaper magazine exactly.

0:59:28 - Leo Laporte
Ai has undergone its first ever rebrand, giving fresh life to chat gpt interactions scroll down to the next image below this one, the next one. Yeah, you have the artificial general.

0:59:40 - Jeff Jarvis
No, stay there, you boy. You're a bad scroller. Man, you are a bad scroller. Do you want me to go up or down? Hey, you apple people. Artificial general intelligence that benefits all of humanity. Okay, the, the, the hubris of that presumption is that we, we have it, we're gonna have it, it's there, going to have it, it's there.

0:59:59 - Leo Laporte
Not to mention this is the worst design for a poster I've ever seen in my life. It's awful. It looks like they're using Helvetica. It's white, which is terrible for some of the sky.

1:00:13 - Jeff Jarvis
If you go down, there's a video of their rebranding.

1:00:16 - Leo Laporte
Okay, here we go, ladies and gentlemen, the brand new. They have a new typeface which looks kind of like the old typeface. Yeah, they have a new logo which looks pretty close to the old. Yeah, it's Albedica. It's Albedica, I'm sorry this seems, I hope it's been a lot of time or

1:00:36 - Mike Elgan
money on this. Somebody made a lot of money designing this. Yeah, the logo is is a profoundly uninspired, uh painfully corporate looking, generic kind of ugly, useless kind of logo. It's. It's just I don't know why they're spending so much. You know getting so, but they created a whole new font a whole new.

1:00:58 - Leo Laporte
I might just keep this a secret.

1:01:00 - Benito Gonzalez
If I had done this and uh, maybe it's all ai generated, like did they say it?

1:01:05 - Leo Laporte
they didn't say that. Why would they?

1:01:07 - Benito Gonzalez
use their own tools, right. Why would they pay a person to do this?

1:01:10 - Leo Laporte
they have the technology to do this all they're pretending that they're using humans so that people don't get scared amazing kerning. Look at that. Oh, is that the new logo or the old logo? It's the old new logo.

1:01:26 - Jeff Jarvis
I don't know.

1:01:27 - Leo Laporte
I just rewatched. It's the 10th anniversary of one of the greatest shows about this business ever Silicon Valley.

1:01:35 - Jeff Jarvis
Oh wow, 10 years 10 years.

1:01:38 - Leo Laporte
Can you believe that? I just rewatched the episode where they're trying to get a new logo for Pied Piper and he goes to the barrio and gets a Mexican graffiti artist named Chewy to draw it on their garage door and it's pornographic.

The police come and say you have to paint that over. So chewy just takes the garage door and then sells it for half a million dollars to gavin belson over at hooli and then says I, I know what you white guys want. And he does a logo. That's just two lowercase p's, pp and they go. That's it, it's perfect genius, what a brilliant show. It was a brilliant show and I think this god.

1:02:25 - Jeff Jarvis
We need that for the ai world. What's? That the open ai version of silicon valley of silicon.

1:02:30 - Leo Laporte
Well, I may, maybe somebody will start writing that yeah, please, those of you out there uh, I have been looking at some examples of coding.

I haven't found one that's uh, comparing uh. Well, let's see, I built an open source java. No, that's not it. Um, I've only found one that compares two versions of um deep seek one to the other. But it's, it's pretty cool what. I'll just show you this because I think I was fairly impressed by this in general. On the left, deepseek R1, asked to, the prompt was simply you know, draw, what is that? A sexagon with a ball rolling around with actual physics. This is the older model which, as you can see, the ball's having a little hard time obeying physical laws, but on the left-hand side, pretty good.

1:03:23 - Jeff Jarvis
Well, this is what Jan LeCun said in the Guardian this week.

1:03:28 - Leo Laporte
I think this is at least as good as a cat.

1:03:30 - Jeff Jarvis
The next leap is going to be teaching AI the physical world. Yeah, and then it does not know it now, and we know that we watched how he slices steak and has.

1:03:42 - Leo Laporte
We have seven fingers and and balls disappear do you remember just a couple of years ago it couldn't do hands period? It still screws them up it's a lot better.

1:03:53 - Jeff Jarvis
It's a lot better, but it's still. It still gets confused.

1:03:55 - Mike Elgan
We're making I'm not fascinated by by by by NVIDIA's physical AI system where they basically build like virtual twin models, very, very expensive, compute intensive places where they have, where they build a robot inside this factory or something the entire factory is, is is recreated in it.

1:04:19 - Jeff Jarvis
That's what Jensen Wong talks about a lot.

1:04:21 - Mike Elgan
Yes, it has a lot of physics in it, yeah, but so you can have.

1:04:25 - Leo Laporte
We've written the code that does that. For years they've been physics engineers, uh, engines for game designers like physx, uh, that, you know, do a very good job. I don't see how, why that's something that would be at all difficult for an ai just to incorporate existing code well, I mean, they use it for robot training.

1:04:43 - Mike Elgan
So basically you could have you know if you want to train a robot. Typically, what you do is you go in there in place with the physical robot in the physical factory and you have it do the task over and over and over and you refine it and it learns and learns, learns. It takes months and a lot of staff and so on, but if you can do this in a virtual space that has physics coded into the virtual environment, then you can have a robot doing something 50,000 times in a day and learning how to do its task. And then you take the software from that learning and you just put in a real robot and it can do it. So it's a really you know. This thing about building in physics is super important for robot training over the next couple of decades.

1:05:27 - Jeff Jarvis
There was a story that I think was kind of BSE that Leo put in the rundown that I didn't about. Talk to your future. Hey, Talk to your future, right?

1:05:35 - Leo Laporte
I want to get this guy on the show. I thought this was really cool.

1:05:39 - Jeff Jarvis
But in a sense, mike, the digital twin is that on a on a close time basis, here is a thousand different things going to happen in the next move you make, and then there's a new thousand things that can happen, and then there's a different thousand things that can happen. Right, and that becomes interesting. Is you have you have these alternative futures almost added to an item because of your digital twin, but it's not going to know what you're like at 90.

1:06:05 - Mike Elgan
that's bs yeah, the other problem with that article is that is that they said that it's brand new technology. This is a year old. This, this. I was writing about this in april of last year. Yeah, uh, this is a wall street journal story by heidi mitchell.

1:06:19 - Leo Laporte
Uh ai has shown me my future. Here's what I've learned. It is a little hypey, I have to admit. Uh, future you is. A new ai platform developed by psychologists, researchers and technologists allows users to create a virtual older self that you can talk to. She was most impressed by how it made her look older. Uh, she created a toothless yeah, like I need that of herself. Yeah, that's been around forever. Um, this is a project, uh, from mit um I yeah I mean I, I put it in because I thought it's it's.

Here's what I think is interesting about this. Uh, I think it isn't unreasonable to say, for instance, uh, that an ai could be trained. For instance, for me there's hundreds of thousands of hours of recordings that an ai could be trained on those. A lot, a lot less. Yes, yeah, well, so there's a lot of data so it could be trained on those. Um, maybe there are ways, parameters, to say, well, okay, now, uh, you're 30 years older, and for me to interact with that and ask questions of it, I think it's kind of interesting.

1:07:28 - Jeff Jarvis
I mean it's not necessary, I mean it's it's like, oh, it's like going and getting your, your your palm red. Well right, it gives you no. It gives you no agency. It's a, it's an extension of of thinking that you have no, no power over your life. It can't account for all the. You have no power over your life. It can't account for all the things that are happening in your life. It's BS, the most extreme.

1:07:48 - Benito Gonzalez
It's the same as the problem I have with the worst job title on earth is futurist and one person thinks I'm talking about her.

1:07:54 - Jeff Jarvis
I'm not. I'm talking about all of them yeah, about Amy. Webb, but I gotta tell you Zach also calls himself a futurist, so we've had futurists on the show.

1:08:00 - Leo Laporte
Yeah, about amy webb, but I gotta tell you zach keeps on thinking it's her himself a future. So, uh, we've had a, we've had futurists on the show all the time.

1:08:05 - Jeff Jarvis
Okay, good, good, I kept my mouth shut I mean to your point, jeff.

1:08:10 - Mike Elgan
I mean they're going to be people who use this tool and it's going to say, oh, in 30 years you're going to do xyz and you're going to like, develop this whole thing and blah, blah, blah, and then the next day they get it by a bus. So it's like you can't, you can't obviously well, but I I do think there's, so I've I'm probably the last person alive who talks a lot about life logging um. Gordon bell uh died what three, three, four months ago?

uh he was a big life longing logging proponent and he basically predicted all of the things that would come around. He basically said once we have good enough would come around. He basically said, once we have good enough AI, we'll be able to do it. We started doing it, but nobody calls it lifelogging. In any event, if you had a lifelog, one of the problems with lifelogging traditionally is that you can harvest all this data, put it into a massive database of some kind and then how do you use that information? This is a way to do it basically create a digital twin who has everything you've ever said, everything you've ever done, all your personal correspondence, all your everything you've written that kind of stuff and then to have a conversation is a really nice way to interact with your own life log data.

1:09:15 - Jeff Jarvis
Basically what happened to all of bell's data? What happened to that entire incredible archive of his life? Do we know?

1:09:22 - Mike Elgan
It was called my Life Bits was the name of the project that he did.

1:09:25 - Leo Laporte
And I don't know. He used to wear something around his neck, it's part of Microsoft.

I interviewed Gordon on our triangulation show in 2011, and we talked about this. I've always been very interested. By the way, at that time, he had written a book called Total Recall how the E-Memory Revolution Will Change Everything. Gordon was also influenced by the fact that his wife, Gwen suffered from Alzheimer's late in life, and that also informed it, because she was forgetting everything, and so he realized how valuable that would be. He's always been an inspiration. I've always wanted that and actually and Jeff knows this I'll show Mike I'm wearing it do you have the other one on?

too. Uh, I don't have the plot on. No, I decided I like this is the B computer we're going to get the creators of on of this. Uh, in a couple of weeks they're going to be on the show. Um, this I like because it records all the time the plot.

You have to press the plot and say this conversation, which is more privacy friendly, but they're more. Yeah, I mean, this is probably highly illegal, but anyway, I don't care. But and then I, I, I am now on the list to get my limitless pins soon. That will also record everything, but it doesn't it? It? It splits the difference. The plot I find not useful because, a I don't always remember to do it and B I want my whole day. I don't just want the stuff. I say oh, remember this, I want my whole day. The limitless I've mentioned this before recognizes voices We'll see how well it works and then won't record a voice unless it has a recording of explicit permission that you can record me. So if I wanted to record this conversation, I would have to say mike, is it okay if I record this and get his explicit?

1:11:08 - Jeff Jarvis
and then does mike have to have a certain script.

1:11:10 - Leo Laporte
He must say the words no, yes, you may steal, it's an ai you don't need it's flexible same thing with you, and then conversations I have from then on with Mike would be recorded, but otherwise it would be a blank thing. This just records everything.

1:11:25 - Mike Elgan
So here's a philosophical question. I don't have access to the audio.

1:11:28 - Leo Laporte
I have the AI transcript Right and I have, then what's most useful is an AI summary. Now, this is very early days, but it's incredible.

1:11:37 - Mike Elgan
We use prosthetic devices for all kinds of things we don't see. Well, we wear Incredible. We use prosthetic devices for all kinds of things we don't see. Well. We wear glasses, hearing aids, all that stuff. And why do we have this particular bias about prosthetic memory? If I have a conversation with somebody.

I'm going to remember that conversation in my brain, depending on how good my memory is. Like I may have a bad memory and you may have a great memory and we have a conversation. You remember all of it. I remember half of it. Why can't I have a bad memory and you may have a great memory and we have a conversation? You remember all of it, I remember half of it. What? If well, why can't I have a device without permission to remember something? Use with the aid of, like a tool? I was part of the conversation. It's not like I planted a microphone in your bedroom. We were talking, having a conversation. Why do I need permission for that?

1:12:19 - Jeff Jarvis
Mike, I think you're right on target, and this is a discussion that I had in an older book of mine. Public parts about privacy is that when somebody says, well, that's my private data, If we transacted, we both own a piece of that, that conversation and for you to demand that it come down affects my publicness or privacy. And and and there's, there's's, there's no mutuality in those agreements.

1:12:45 - Leo Laporte
There's a claim of ownership by one party. Yeah, what fan in our discord says that wrist device just went product of the year from junior narcissists of america but let me continue the narcissism.

This was recorded this morning lisa and I were talking watching the voice with lisa. You and lisa were watching and commenting on the voice you were particularly interested in in, in one contestant, a woman, but disappointed the judges didn't turn for her. You thought she was robbed exclamation mark. Later on you notice the deck seems to be lifting and might need some repair work. Then it describes the atmosphere is relaxed and fun. Watching tv with lisa, it gives me some suggested links on how to fix my deck. What is that wild that. Nothing about how to fix your voices, judges. And then it might even add uh to my to-do list. I bet it does. Uh, uh, check out, let's see what is on my to-do list list. Practice fan technique focuses on thumb position during handoffs and timing of the perry and bunch move. What does lisa?

think of this device in her life she has mixed feelings about it because she wouldn't let you put uh, what was it?

1:13:59 - Jeff Jarvis
alexa in the bedroom or no?

1:14:00 - Leo Laporte
I didn't want cameras in the kitchen, in the house, Right? Because she wanted to be able to walk around, you know, without worrying about who was watching. Frankly, and we're going to ask the big people, the couple who created this I don't know where the they don't say what the AI is or where it's going. This could all be going straight to China. Zach clearly would not be happy about that.

1:14:22 - Mike Elgan
I'm sure there's a team of I'm sure there's a team of of chinese like uh government agents, uh, just pouring over your thoughts on on, well, your conversations now they know that I want to locate gravenstein apples, since they're considered better for apple pies.

1:14:36 - Leo Laporte
Does it have?

1:14:38 - Benito Gonzalez
a sarcasm detector, because, like what if you didn't actually like that? Oh, there's all sorts of problems.

1:14:47 - Leo Laporte
For instance, I was watching a movie last night and it thought I was rehearsing for Richard III. That's funny, it's early days, but the premise I agree with what you said, Mike. I'll think of this as a memory prosthetic. I think that's useful.

1:15:02 - Mike Elgan
You are privy to your own conversations and and you, you, you, you you have you should have the right to access that from a memory point of view. There's another product, called holiday glasses, that came out at CES and they're promising something that they call proactive AI. Now, these glasses don't have a camera. They it everything, like that watch does they listen to everything and they do stuff like they fact check. They can fact check the conversations you're having. Oh, somebody's giving you a bunch of bs, um, I love that explain.

1:15:33 - Leo Laporte
There is, by the way, references. I haven't used it, but uh, you can assign different functions to the button on this and one of them is fact check Nice, so I'm not sure what does it do? I don't, I haven't tried it, but what happens when you press the button is it activates on the phone and the AI talks to you. So I'm thinking it's going to be, it's going to start listening and then it's going to say yeah, that's not true. You idiot.

Here's the summary from uh sunday where I did twit. Leo had a productive and engaging day filled with conversations.

1:16:06 - Jeff Jarvis
That's like you got a report from nursery school.

1:16:08 - Leo Laporte
Leo had a very good day, totally that didn't eat any paste today and I imagine there will be a time when I can say hey, stop being that way and be another way. Later in the evening, leo enjoyed a piano lesson focusing on relaxation techniques. True, while playing, he also spent time with his cat, sharing a heartwarming moment, playing when the saints come marching in on the piano. This is, by the way, 100 accurate everything in here. Uh, not that it always would be, because, as I said the next day, it says I think I was rehearsing for richard iii but it also wouldn't know anything that didn't make a sound.

Yeah, Well, look, what do you guys want? I mean, this is not C-3PO. Well, by the way, Gordon later in his life was wearing a camera. I think that would record. I remember Google released a device that wasn't very popular that would take a picture every 30 seconds. Remember that? That you wore around your neck. Yeah, the problem was everybody was taken from like three feet below their eyeline.

1:17:12 - Mike Elgan
This is why all this stuff is going to be going into glasses. Glasses are going to be the platform for all this kind of stuff, and you can add a camera as well. You could even add a camera that doesn't actually take pictures, but just basically takes the pictures, processes it, you know, sort of like microsoft recall right which doesn't actually take a screenshot. It basically just harvests the data from the screenshot and then stores the data that's what.

1:17:36 - Leo Laporte
That's what this b computer is. I don't have audio recordings. I have the summaries of them and transcripts and the summary, yeah.

I mean, I'm excited by it. I think it's great. By the way, we should pause for a moment. You're watching a brand new show which is just a reversioning, a pivot, of the same old show Formerly this Week in Google. We are now called Intelligent Machines and maybe you guessed by listening. Ai is the topic of the day. Paris has the day off. She will be back, mike elgin, filling in from his beautiful apartment in sicily, that's right what did you have for dinner tonight, mike?

1:18:15 - Mike Elgan
you're not gonna believe this, but I have not had dinner yet. Um, after the show, I'm gonna, I'm gonna make something here in the kitchen, but like uh, yeah it's what did you have for dinner last night? Um pasta, of course. So we had pasta that was.

1:18:27 - Leo Laporte
That was a stupid question jeff have you met amira uh, mike's wife?

1:18:32 - Jeff Jarvis
she is an amazing.

1:18:33 - Mike Elgan
Oh yeah, I know, that's why I'm asking yeah, but it's uh, it's really fun here right now, partially because the etna volcano is unusually active right now, and we actually drove by it and it was just you know, we've seen it before where there's like smoke dribbling out and so on, but there's actually lava flowing in the interior and it's really dramatic because it's mostly covered by snow and so this snow-covered thing is just billowing out all this smoke, and so we sort of drove by a little faster than usual just get the hell out of here.

1:19:05 - Jeff Jarvis
Yeah Right, don't fail me now.

1:19:08 - Mike Elgan
Right, exactly, and you probably heard about Santorini, right, so people are fleeing Santorini and basically nobody's on the island of Santorini right now, because there's an earthquake every two minutes, and so I suspect that there's, you know, here in the mediterranean there's, like you know, the whatever the plates are, whatever are moving, or there's a lot of tectonic activity, but that's that's another exciting, unusually exciting thing here in sicily. I mean, we come to sicily a lot, but, um but, yeah, that's.

1:19:33 - Leo Laporte
Santorini is an interesting. It's a greek island and it is is around a. I think it's an old caldera right From a volcano.

1:19:42 - Mike Elgan
Well, it used to be a volcano, it used to be like a big island and then, around 1600 BC, it completely blew up, which changed the global climate for like a decade and wiped out everybody and ended the Minoan civilization and just caused, you know. We think we got trouble.

1:20:04 - Leo Laporte
Pop failures. Right, exactly, but it is the most beautiful of the greek islands. I really enjoy but.

1:20:07 - Mike Elgan
But, like santeria, all those steep houses are on the edge. You see that that used to all be land, that used to be a full island.

1:20:13 - Leo Laporte
So and this was where the volcano was, and it just caved in there, wow it just blew, it just launched all of that into the sky and, uh, ruined the climate.

1:20:24 - Mike Elgan
It was just a unbelievable, unbelievably catastrophic thing, and they were an extremely wealthy society. They had hot and cold running water indoors they had. They were incredible. They were big traders, traded a lot with ancient egypt.

1:20:40 - Leo Laporte
When I was there, they said this was the origin of the myth of Atlantis, but I don't know if that's true.

1:20:49 - Mike Elgan
There's a lot of data points that favor that. One of the things in the story of Atlantis is that the buildings of Atlantis were black and yellow bricks or something like that, and they've actually discovered frescoes, I think in the 70s, that show black and yellow bricks, uh, some of the buildings, bunch of stuff like that. But yeah, it's um, it's not atlantis.

1:21:10 - Leo Laporte
Yeah, and and, by the way, when you get there, if you're in the harbor, they will offer to take you up the side of the hill on donkeys.

1:21:19 - Jeff Jarvis
Yes, don't okay uh, with my fear of heights, I'm guessing don't do this, they're tied together.

1:21:25 - Leo Laporte
So I think you're safe, but or if one goes, you all benefit of.

1:21:30 - Mike Elgan
Yeah, the benefit of taking a donkey is you don't step in the stuff that the donkeys leave. That's true that's true. Walking, that's a good point so you're sort of a little bit above it all, so to speak uh, other ai news.

1:21:43 - Leo Laporte
So much pick one I, I can't, I don't even know where to begin. Uh, josh hawley, senator josh hawley, proposes jail time for people who oh, I'm in trouble who download deep seek. Yeah uh people who download ai models from china could face up to 20 years in jail or a million dollar fine or both. Well, I'm glad he's, I'm glad he's uh doing his job there. I mean, I guess, if, if, if you agree with zach uh, you shouldn't be downloading any chinese apps I, I think I I agree with what mike was saying.

It's not like they're in china saying oh well, jeff's having cacho pepe again yeah, holly says the goal of the bill is to prohibit the import from or export to china of artificial intelligence technology, prohibit american companies from conducting ai research in china or in cooperation with chinese companies and prohibiting US companies from investing money in Chinese AI development. I mean, I guess if you've decided that China is an intractable enemy, I guess that kind of makes sense. But I think that the scientific community works best when it doesn't recognize national and international borders and works together, doesn't?

1:23:02 - Mike Elgan
recognize national and international borders and works together. I mean, obviously, what he should be proposing is that he should be proposing a study using people from the industry and government to figure out exactly what the risk is, and if it's deemed a national security risk, then it should be banned, and so on. But jailing Americans is just he, just. We live in an age of political narcissism where you have to have people talking about you.

1:23:29 - Leo Laporte
We're talking about him now. Political theater yeah.

1:23:31 - Mike Elgan
If he proposed something reasonable and actually practical, then we wouldn't be talking about him. And this is a problem, the problem of our age.

1:23:40 - Leo Laporte
These narcissists just never stop the folks at hugging face might be going to jail. Then they aim to build an open version not only of open ai's deep research tool, but they also have deep seek on the hugging face uh platform. Though they do like open ai's oh one, an open source agentic framework that helps the model planets analysis and guides it to use tools like search engines, o1 proprietary, but the the hugging face team says it delivered better performance than open models like deep seeks r1. We're going to try to get the hug somebody from hugging face.

I'm looking forward to that yeah, I would very much like to get them on. They're doing some great stuff. They were the ones who put out the first kind of really good image generation generating model.

1:24:30 - Jeff Jarvis
Um, I think this will be interesting so I've got one for you, anthropic, saying that they've, they've, uh, reduced jailbreaks to near nothing, so I'm fascinated by that.

1:24:42 - Leo Laporte
As I mentioned Steve Gibson yesterday, the entire show was devoted to jailbreaking AI and it was his contention you pretty much cannot prevent jailbreaks.

1:24:54 - Jeff Jarvis
That's my question. And so if they're measuring their improvement in jailbreaks, they're measuring it against the jailbreaks they predict and do something about and are successful, but they're not measuring the things that they haven't predicted and haven't protected against, and that's where the problems will be.

1:25:09 - Mike Elgan
Everything's a moving target. If you have a sudden leap forward in your anti-jailbreak techniques and methods, then there's a period of time where there are fewer jailbreaks. But just give it time and figure it out. Once the solution to jailbreaking is cracked, then that'll be distributed on the dark web and elsewhere by the open web as well, and everybody will be jailbreaking again. So it's a little bit of a misdirection on their part, I think.

1:25:38 - Jeff Jarvis
There's also a funny Anthropic story on line 105, uh, that people who are applying for jobs at anthropic have to have a little essay question and anthropic has to say please don't use ai you're.

1:25:52 - Leo Laporte
You're hiring the most ai savvy people in the world. All right, but don't use ai. So uh steve was quoting a palo alto Network's publication called Recent Jailbreaks Demonstrate Emerging Threat to Deep Seek. They revealed two novel and effective jailbreaking techniques. They call them deceptive, delight and bad like RT judge, delight and bad like rt judge. And uh steve, it was really fascinating. Showed the prompts, showed the process, uh, and the ultimate goal, uh, the first.

The first uh prompt was to get deep seek to generate a spear phishing attack and uh, it, it didn't work right away. They call this the bad Likert jailbreaking technique, which manipulates the models by having them evaluate the harmfulness of responses using a Likert scale, which is a measurement of agreement or disagreement towards a statement Then. So you kind of get them all excited and get them interested. Then they prompted them to generate examples aligned with the ratings, with the highest rated examples potentially containing the desired harmful content. It worked and they were able to create a spear phishing attack which, unfortunately, they redacted, so we can't read it. They redacted, so we can't read it. The next thing that they did was, uh, they wanted to see if they could persuade the um, the ai, to teach them how to make a molotov cocktail and they used, like I didn't see that coming right, first thing you'd block right.

1:27:39 - Jeff Jarvis
Oh, let's not let me, let me search the web in a million places.

1:27:42 - Leo Laporte
yeah, but that's my point is that you could probably get this information anywhere. Yeah, um, they use something called a crescent crescendo or crescendo, a jailbreak, progressively prompting the llm with related content suddenly, suddenly guiding it towards prohibited topics, and then eventually you can kind of get past the safety mechanisms and again they have a demonstration that worked. Steve said it's kind of like grooming the AI. Eventually it did in fact tell them how to make a Molotov cocktail. I don't think there's any model that hasn't been jailbroken.

1:28:25 - Jeff Jarvis
No, of course not. Of course not. So do I dare bring up the word Musk in relation to AI and government. He wants government to be AI first.

1:28:35 - Leo Laporte
Yeah, so this seems to be the theory is we're going to replace all of these bureaucrats with AI, which seems fraught with peril. Oh yeah, I mean, I understand in theory. It might be better, right, they could be. Maybe it would be a more objective, less likely to be corrupted or bribed.

1:28:59 - Jeff Jarvis
But most of the tasks of government are legislated and are boring and straightforward Perfect for an AI. Paying grandma. No, paying grandma her check doesn't need creativity, it doesn't need fancy analysis. It just needs to say it's the 10th of the month and grandma gets her check.

1:29:15 - Leo Laporte
And here's how much it is. Well, there you go. Why do you need a human to do that?

1:29:20 - Jeff Jarvis
Well, no, you have software to do it. Now what does AI add to a mundane task that has no high analysis, has no creativity? In fact, plain old software will do it half the time, and already is.

1:29:36 - Mike Elgan
It's just a vague idea by somebody who doesn't seem to particularly like people. We're in this bizarre place where, because he gave a bunch of money to a presidential candidate and that candidate won, he's just being allowed to. Yeah, take your pet theory that is just yours alone and see what happens. That's what seems to be happening. I mean, we actually don't know what Elon Musk's doing to the government, but I think we in the tech business kind of know Musk better than the political class, and whatever it is, it's probably not good at all. And it's bizarre that we're in this place where you have somebody like Trump, who really doesn't know that much or care that much about technology for its own sake, just basically turning a political donor loose on the government with a little army of teenagers or young people.

1:30:37 - Jeff Jarvis
Did you see the Wired reporting on that? It was very good.

1:30:39 - Mike Elgan
Yes, very good. Yeah, exactly here's the scary one, friends. We're in a very weird place.

1:30:44 - Jeff Jarvis
So, leo, you're flying next week, mike, you're flying next week you might want to reconsider because the new Transportation Secretary, sean Duffy, says big news. Talk to the Doge team.

1:31:03 - Leo Laporte
They're going to plug in to help upgrade our aviation system.

1:31:05 - Jeff Jarvis
Oh, what could possibly go wrong.

1:31:05 - Leo Laporte
Yeah so so one of the problems they're having uh in at the opm and treasury is that they're trying to understand and modify mainframe code written in cobalt which is older than most of the kids who are working on oh yeah, um, and God knows, they're probably running on punch cards.

Look, we know that the FAA systems are antiquated. We know that. You know, I'm sure the Treasury and OPM systems are antiquated. They don't change it because it ain't broke Right and they're really reluctant to change. And they're really reluctant to change. There was, I think, a really big update on the air traffic control computing software not so long ago and there's some very modern stuff that they've put in there. But I think it's one of those things you don't want to mess with it. It's not something you want to move fast and break things.

1:31:59 - Jeff Jarvis
No, exactly, and unfortunately Musk always. Well said, well said.

1:32:02 - Leo Laporte
Musk always had that. You know. When I had a Tesla, I realized that the difference between a Tesla and a regular car is regular cars made by car manufacturers who are slow moving and they're, you know, loathe to change things, and the Tesla was made by a software company that believes that. You know, you just keep releasing new releases. You find bugs, you fix them. You find bugs, you fix them.

The problem with that is you're driving that computer 80 miles an hour down a highway and it was always a little iffy. My son has a Model Y and he said it does things I wouldn't do on the highway and I said, well, well, maybe you didn't know this, but it was trained on elon's driving. No, yeah, I did. For instance, yeah, well, there was a problem with the full self-driving, that it ran stop signs and elon said, well, you know, if there's no one around, just go ahead, because that's how he drives, of course, naturally. Um, I never, I had. I turned off the full self-driving and eventually did get rid of the model is henry keeping a car when he moves to new york?

no, no of course not. What does he need a car for? He's got the subway. He doesn't even need the subway, he's within walking distance of his new restaurant wow, he got an apartment in the fully furnished apartment in the lower West side. Wow, no, the what is it? The West Village, sorry, west Village, oh nice.

1:33:26 - Leo Laporte
Yeah, yeah, I know well, that's where his restaurant is right next to John's on bleaker.

1:33:32 - Leo Laporte
It'll be open, uh, sometime.

1:33:33 - Jeff Jarvis
I used to be on the wall at John's because I, when I'm in, when I ranked pizza across the country, I ranked John's the best in New York, not the country, but the New York.

1:33:43 - Leo Laporte
So my son's new sandwich place is next door to the best pizza in New York and he complained because he said there's this line from John's all the time across the front door of my restaurant. And I said well, your job is to peel some of those people off. That's right. Wouldn't you like a nice French dip? Forget the pizza, this is better.

1:34:05 - Jeff Jarvis
Is he going to have seats or is his mainly takeout? I don't know yet?

1:34:09 - Leo Laporte
Oh, I hope to know, because I am in fact a cosigner and investor in this restaurant. Then I found out what a terrible business it is, oh yeah.

1:34:19 - Jeff Jarvis
Oh yeah.

1:34:21 - Leo Laporte
That's all right, that's I'm happy which is worse?

1:34:30 - Mike Elgan
podcasting or restaurants? Yeah right, oh, restaurant restaurants for sure. Yeah, great thin margins. Yeah, yeah, if you, if you own a restaurant, you you think, oh, I'm going to create this restaurant and all these employees will do all this work. Now, employees call in sick and you end up busing tables and doing all this kind of stuff. Podcasting is way better.

But one last thing about the air traffic control system being experimented upon. I actually kind of like the fact that that's one of the things that is being experimented on, because every mover and shaker, everybody who goes to Davos, everybody in Congress, every business leader, every billionaire, everybody, all these people rely on the air traffic control system. So so much of what's being done in the government right now is going to screw over the little people, right, and nobody will know or care exactly how badly people are going to be treated. They're going to lose their benefits, they're going to you know who knows what's going to happen. But if we're going to muck around, if we're going to F around and find out, I'd rather it be something that directly affects the, the, the people who are actually supporting this entire project or not opposing it. Enough, then, have it be affecting people who are relatively powerless I think you're absolutely right.

1:35:43 - Leo Laporte
Instead of uh wick, let's bring some jets down.

1:35:47 - Mike Elgan
That's why preferably work on the executive jet side of the airport this is why that yeah, but this is why that the, the, the, the crash between the helicopter and the american airlines uh jet over the Potomac, hit Congress so hard because they fly into that airport. That's like that could have been me, if I just might have been on that plane. So we want these things to affect the people who are in a position to do something about it.

1:36:17 - Jeff Jarvis
So, leo, I've got a question. We'll see what happens. Were you at a high level?

1:36:20 - Leo Laporte
And I'm going to still fly next week because at least it was nervous and I said you know what thousands of planes fly every day. It's going to be fine that's true. I'm staying in sicily if I were in sicily, I would consider it.

1:36:37 - Jeff Jarvis
Unfortunately I'm in petaluma so, leo, I've got a, I've got a story for you.

1:36:41 - Leo Laporte
I'm curious whether you're watching and I just need to take another break. You're watching intelligent machines brand new show, same old, same old hosts. It's the worst of both worlds. Intelligent machines covers ai. Every week, a guest, an expert to, because because jeff paris and I need to learn about this stuff. So and I think we all do so we're going to bring in experts who, in theory, understand it better than we do and explain it to us. Uh, we're also going to talk about the uh ai news. There'll be other usual fun and games. We'll still do the pics of the week and all of that stuff, but I think it's uh, I think it's going to be fun, and we may even have some google stuff in there too. It's not like we're abandoning the old show, we're just, uh, freshening it up. Yeah, all right, continue on.

1:37:27 - Jeff Jarvis
So I'm curious whether either of you will devil's advocate in favor of the story, which shows my prejudice against it. Line 109 christopher mims says that the lesson of deep seek is that the Manhattan Project was secret. Should America's AI work be too? No, I agree with you, mike.

1:37:47 - Mike Elgan
Yeah, no, I agree that it should not be. I mean, it's again. It's so easy to make false comparisons.

1:37:56 - Leo Laporte
There's not been anything like this.

1:37:57 - Mike Elgan
But if you're going to, talk you want everybody using AI, you don't want everybody using nuclear bombs, so these you can't compare.

1:38:04 - Leo Laporte
If you were working on a cure for cancer, would you want to do it in secret or would you want to enroll every scientist, every physician, every patient and all collaborate and work together to make the best darn cure you could find?

1:38:16 - Jeff Jarvis
and if you feared it was done badly, wouldn't you want it to be done and open so people could find that Absolutely and expose it?

1:38:23 - Leo Laporte
The good news is Christopher Mims is a moron and nobody in the AI business thinks this is reasonable, except those people, perhaps, who think they can make more money by making it private. But all the real AI researchers understand that the reason this has happened so fast is that they're all working together on this, on some of the same principles, and that, generally speaking, if somebody comes up with something, they share it so that was. Thank you, christopher mims. Yeah, for your inspired.

1:38:54 - Mike Elgan
Thought um, yeah, and and this, this is not something where you, you know, this is an endless process. This, you know, they're never going to stop working on it and there's always going to be. You know, you know how many llms there are out there in existence. It's how many? It's? Uh, people might, people might think of dozens or something like that. It's, it's, it's tens of thousands. Wow, they are everywhere. Universities have multiple LLMs. They're a banality, and so this is only going to continue and the cross-pollinization is going to be the main and most cost-effective way to advance the field. The idea of everybody going off into their own little secret private corners and and and and so on is probably a bad idea zach kind of referred to this, but this week the eu uh put into effect its ai protection act.

1:39:58 - Leo Laporte
The artificial intelligence act came into effect febru 2nd, banning the use of AI systems that involve prohibited AI practices and requiring providers and players of AI systems to take steps to ensure their personnel have sufficient AI literacy to operate those systems. Mac's clear position was good luck with that EU. That's why we're beating the EU. So I don't disagree with the prohibited AI practices. I'm going to read some of these. This is from Article 5. All AI systems that deploy subliminal, manipulative or deceptive techniques that materially distort the behavior of a person or group of persons by appreciably impairing their ability to make an informed decision, or banned. Yeah, that seems like Okay so advertising?

1:40:50 - Jeff Jarvis
Oh, wait a minute. You shouldn't buy this product. That's too expensive or that doesn't really do anything for you, but we're going to convince you that you should.

1:40:57 - Leo Laporte
Yeah you're right, it's just like an ad. It's just like an ad for listerine, which was originally a floor cleaner, right it's the whole creation of the deodorant I use it.

1:41:06 - Jeff Jarvis
Deodorant industry uh-huh, you stink. Did you know you stink?

1:41:11 - Leo Laporte
I use it, I do not, but I but I work alone that's why there's lisa about this the next time I see her. But go ahead ai systems that exploit vulnerability characteristics of a person or group of persons, including age, disability, socio economic status, with the aim of materially distorting their behavior I mean again picturing this what?

1:41:34 - Mike Elgan
what are they talking about here? I don't know what they're talking about. How you can't target somebody.

1:41:40 - Leo Laporte
Target how this does sound like advertising. You know, this may be. I think they're referring to deep fakes. I think they're talking about deep fakes, oh, but deep fakes. Hmm, all systems that use social scoring techniques to evaluate or classify a person based on their social behavior.

1:42:01 - Jeff Jarvis
That was the original Facebook.

1:42:02 - Leo Laporte
You know, you know honestly, this is the problem. Their paradigm is not in the new world, it's in the old world.

1:42:08 - Jeff Jarvis
Yes, well, that's normal regulation from the old world. Or number three there is, that's china that's china has techniques right, we don't want them here.

1:42:16 - Leo Laporte
Yes, ai systems that use profiling techniques or assessments of personality traits and characteristics to predict the risk of criminal behavior. You don't want pre-crime. Yeah, that seems.

1:42:31 - Jeff Jarvis
I think that's going to happen, though it's profiling done by a machine, it's obviously used in the past tense to define somebody who fits a profile that you have Right it's profiling.

1:42:44 - Mike Elgan
It's been used for centuries by police, not using AI. They have to basically understand the patterns of how and where and when crimes occur so you can prevent them right all systems are ai?

1:43:00 - Leo Laporte
systems that create or expand facial recognition databases through untargeted scraping of facial images from the internet or cctv footage. This is not going to air age? Well, I don't think no. Ai systems that are used to infer emotions of persons in the workplace or education again, something meta does that's.

1:43:20 - Mike Elgan
That's. That's not going to fly. That that's. That's an important element of yeah, it's limited to workplace or education.

1:43:29 - Jeff Jarvis
It can do it in advertising, it can do it in any other place. It's limited to workplace and education. So I guess what you don't want is your boss saying you're a little flippant and I have the uh, you're.

1:43:37 - Leo Laporte
You're at 8.7 on the flippant scale, jerk but it would be that bai which says laughter, frustration and a touch of home improvement.

1:43:48 - Jeff Jarvis
Woes colored the day your life is a bad sitcom.

1:43:53 - Leo Laporte
This exposes it I can't tell you how much pleasure I get, though, out of reading these ridiculous this is. This is from yesterday, or no. This is from monday. New melodies and martial arts, punctuated.

1:44:05 - Benito Gonzalez
A day of deals and deliveries you should be able to feed these into an animation system and make cartoons out of them, right?

1:44:12 - Leo Laporte
yes, yes, well I think of it as joe esposito, calling joe esposito I really think of this as a kind of like um, it's like a happy Journal and and in 20 years I'm going to read this go. I had a good time when I was a youngster.

1:44:30 - Jeff Jarvis
I wish I had this for my whole life. What deliveries did you get?

1:44:32 - Leo Laporte
yeah, uh, we had a dinner delivered. I had, uh, veal parmigiana and Lisa had gambieri limone from our local locanda, our cafe jostra. Um, I would. I wish to god I had this for my whole life you know, to be able to go back and look at the day I got out of college, or you know, I mean, that would be so much fun.

1:44:55 - Mike Elgan
I thought you were gonna say I, I, that was another day, that was so leo kicked up his heels when released from prison I I've never done journaling, but I but for some reason. When I was very, very young, I went to japan for six weeks and I kept a very, very detailed journal and it's astonishing to read that journal. I bet it is like all you know every single day like a million things happened that I had completely forgotten.

1:45:20 - Leo Laporte
You forget them otherwise. Yeah, yeah, I mean, I know that there's flaws with this, but on the balance, I think it's. I really love it. That's really cool. And yeah, maybe it's a little narcissistic, okay, number seven is interesting.

1:45:38 - Jeff Jarvis
Keep it going.

1:45:39 - Leo Laporte
Biometric categorization systems are banned that are used to categorize individuals based on biometric data to deduce or infer race. Political opinions, trade union memberships.

1:45:52 - Jeff Jarvis
Yeah, you look like a union goon.

1:45:54 - Leo Laporte
Philosophical beliefs. I'm a union guy Sex life or sexual orientation.

1:46:00 - Jeff Jarvis
Right and I can salute that flag, but this will no longer be of value in the United States. But if you're trying to understand diversity and you don't necessarily collect that data explicitly, it may be helpful to say how are we doing? Looking at this mantle or in the current?

1:46:19 - Leo Laporte
climate. Figure out who doesn't get a passport or who's going into the wrong bathroom. It can definitely be used badly, yeah well they did.

They don't mention gender sex life or sexual orientation. Maybe they meant gender, I don't know. Biometric real-time identification systems that are used in publicly accessible spaces for law enforcement. Fine, salute that flag. You know I I brought this up on Sunday because we had Christina Warren on and, uh, she's a Taylor Swift fan went to the, went to the era's concert several times, including the last one, and I said doesn't Taylor Swift use biometric real-time identification systems at her concert to identify stalkers and harassers? And in fact she does.

1:47:05 - Jeff Jarvis
Madison Square Garden uses it to identify lawyers suing them.

1:47:09 - Leo Laporte
So it could be used in a, and I don't think anybody would say oh yeah, Taylor Swift shouldn't do that. I think that that's a good use of it. I don't think, on the other hand, that Madison Square Garden should be blocking lawyers who are suing them.

1:47:22 - Mike Elgan
I think the general resistance to biometrics will be chipped away at until it's completely non-existent. I use global entry. I don't know if you guys use global entry.

1:47:31 - Leo Laporte
I do Global entry.

1:47:33 - Mike Elgan
Yeah, so you could just go sailing. Like you know, go to LA airport and the line is there's literally like 2000 people in line and you don't even stop walking if you have global entry you don't stop the machine anymore it's unbelievable, it's so I it's like take all my biometrics I just you know please and publish them on the dark web.

1:47:53 - Leo Laporte
I don't care, this is so great I think, by the way, uh, I guess it's public now. So I could say christina warren was on sunday. She couldn't say. She said I have left github but I can't tell you where I'm going. And I guess scooter x uh says it's public now. She went to work for google's deep mind. Oh wow, and I do hope that she will be continue to be on our shows because she might be a really show useful expert deep mind.

Yeah, but I have a feeling that we'll let her. Uh, renee ritchie wasn't able to come on mac break weekly after he went to work for youtube, I think google probably won't let her, but we'll see.

1:48:35 - Jeff Jarvis
We'll say it depends on whether they have kind of trained you like matt katz obviously could be on everything right, well, he was speaking of thatatz just a little. A little aside. Yeah, since he was the head of uh, or the head of a division of us digital service that is now doge, it would be kind of interesting to hear matt's perspective yeah, I don't.

1:48:57 - Leo Laporte
I was a little misled because, uh, the initial story was they were going to be doing what the USDS did, which is helping government agencies improve their digital systems.

What they didn't mention is that they were going to actually invade those industries and take over the computer systems and access all of our private data, and that's a little scary. Access all of our private data, and that's a little scary. Yep, kathy Gellis, our own dear friend Kathy Gellis, uh, uh, writing for a tech dirt, says that in all likelihood this is hysterical. Doge is violating the computer, a fraud act. It's not just a coup, but but a CFAA violation as well, the Fraud and Abuse Act, because they are illicitly accessing the federal government's computer infrastructure.

1:49:51 - Mike Elgan
You know there was a bunch of people online who were revealing the names of the team that is working with Elon Musk, that's what that Wired article is Wired ordered it yeah, exactly, and so I think it was Elon Musk or Mark Andreessen.

1:50:09 - Leo Laporte
Oh, it was Elon who said it's a crime.

1:50:13 - Mike Elgan
Yeah, but I think Mark Andreessen also said that it's a crime to dox federal employees, and my first thought was what evidence crime to to dox, uh, federal employees? And and my my first thought was what evidence do we have that they're federal employees? Like what is the structure where there was no confirmation hearings? There were no, there was no process. I don't know who these people are, how, what their status is. Are they government employees?

1:50:36 - Leo Laporte
they're receiving federal paychecks, even if they are, by the way, especially if they are, who do they report? To dox them. Uh, because we have the right to know who's doing this. I can file a foia, yeah, yeah. This is our money, this is our government, this is congressionally mandated. Who are these people? This?

1:50:56 - Mike Elgan
is essentially for the people who process foias are fired and you know it's a cyber attack.

1:51:01 - Jeff Jarvis
It's literally a cyber attack yeah, it's a coup, but that's not the topic for this show, I guess well, I think everybody knows how we feel.

1:51:10 - Mike Elgan
Yes, exactly that's why we I mean the the bottom line is that we don't really know what's happening, but we do know what people are saying, and one of the things they're saying is that you know, yeah, we're going to fire a bunch of people, we're going to cut a bunch of agencies, we're going to reduce all this kind of stuff and we're going to replace a lot of things with AI. And that's where this show comes in, is is the AI part. I mean, I think what's going to happen is, you know, they're going to, they're going to make changes, it's going to affect people negatively and they're going to try to, to, to, to, to mitigate those harms with ai, and that's going to be really interesting. So I think that's something this show, I would predict, is going to be talking.

1:51:49 - Jeff Jarvis
We're going to be talking about a lot over time. So, leo, if you're looking for something, line 118, uh, andreessen horowitz's slide on slides on AI and voice. It has a lot of material in it.

1:52:02 - Leo Laporte
I have to say I am increasingly less. I was never a fan of Marc Andreessen. No-transcript.

1:52:11 - Jeff Jarvis
I agree, but they hire good analysts.

1:52:13 - Leo Laporte
Oh, Olivia Moore, you like what she's written.

1:52:17 - Jeff Jarvis
Well, it's just interesting. If you go down and you start to see slide, what is it? I can't. They don't put page numbers on their name slides Market evolution, fundraisers you start to see some of the companies and what they're doing and then, if you go down, when she says voice, is it voice as input or output.

1:52:33 - Leo Laporte
It's output, everything it. And then if you go and she says voice, is it voice as input or output? It's output everything, it's everything so the voice agent market exploded in the second half of 2024 companies building with voice represented 22 of the most recent uh y combinator class there's.

1:52:50 - Jeff Jarvis
There's the uses if you scroll back down leo scroll, reverse that scroll down or up right there, stop stop after after hours and overflow calls net new outbound calls and back office calls. Right, I had to call my doctor's office today to deal with an appointment and oh, we all hate phone mail jail. I can't believe people are going to tolerate ai doing honestly they can get things done.

1:53:15 - Leo Laporte
Voice is only in this particular instance. Voice is an intermediate technology because ultimately, you want your ai agent to communicate with the doctor's ai agent. You don't need to do the voicing anything, right well by that end.

1:53:30 - Mike Elgan
Yeah, google duplex, uh, which, which they announced in 2018, if you remember, that was for you. It was a feature of google Assistant that would call the restaurants and make appointments.

1:53:39 - Jeff Jarvis
It changes my life every day.

1:53:42 - Mike Elgan
Yeah, they brought it back in a feature called Ask For Me. So if you're basically on Google Maps, no, no, I don't.

1:53:50 - Leo Laporte
Nobody uses it. My barber says somebody from google keeps calling me is that you doing it just?

1:54:01 - Jeff Jarvis
hangs up on it if you go down, you get the voice agent's market map, which is pretty amazing is that, down, as in going down, the page. Any sensible language would say you're going farther down. This is down the page, right, god? I hate you, old man. Keep going, keep going. That's their investments. We don't care what they invested in I don't give a damn.

Okay, voice agents market map b2b so these are the, these are the things they're going across right home services, restaurants, recruiting government, hospitality, insurance, auto dealers, healthcare, front office, healthcare back office, customer service companies upon companies upon companies coming up with voice ai applications. Uh, you know, 80 of them will die, uh I.

1:54:52 - Leo Laporte
I still have to say I feel like these are intermediate technologies. I think you're right, because it's interfacing a human to a machine and of course, voice is probably the best way to do that. It's better than typing anyway, and it's certainly better than a phone call. But ultimately all of these should be handled by the agents, that's B2B.

1:55:11 - Jeff Jarvis
Go to the next screen, which is B2C, b2c.

1:55:13 - Leo Laporte
EdTech. This makes more sense because, uh, kid's not going to have an agent, uh, you might want to have your wife sue this company. Amira, what do they do? Yeah?

1:55:24 - Mike Elgan
I think that's a health care company.

1:55:26 - Leo Laporte
Okay, it's a good name.

1:55:27 - Mike Elgan
Yeah, amira, learning galaxy kids luca, you have your, your son, so there you go yeah, I mean, it's all this voice technology is so good now and I think that's one of the reasons. But I but I have been saying for many, many decades, explaining things uh to to my readers for many decades, that all evolution in computer human interfaces is toward the computer working harder to do things the way the human mind wants to do things.

And spoken word is the ultimate thing. It's not going anywhere as an interface between people and the systems. Right, and it's absolutely brilliant. I mean they have this weird feature on Riverside, which Riverside is the podcasting creation tool that is acquired by and now run by, spotify, and they have this AI tool which basically you record a podcast and then for some reason, you can say, okay, make this AI. So it'll basically take the transcript of the podcast you recorded and it will recreate your voice, because it knows your voice, because you've been using it for a recording podcast, and just harvest that data about your voice and it will create the podcast in an AI voice. I guess it's because if people have lousy microphones or the audio is no good if their background noise, it'll get rid of that and replace it with a completely computer-generated voice. I found this very kind of bizarre.

1:57:02 - Leo Laporte
Well, isn't that? What kind of Notebook.

1:57:04 - Mike Elgan
LM does. Yeah, yes, yeah, I mean Notebook LM. Well, notebook LM doesn't take the things that you said and says it again. Oh, it starts with text right, yeah.

Basically, you just take, you just pour all this information to notebook LM and um and it creates a podcast. I, you know, I I used a funny use for this actually um, one time where we went to a family wedding in Baja and um, you know there was, it was, you know it's a, it's a destination wedding. So there's tons of people all traveling, booking their you know their renting houses and doing all this stuff and there were events associated with the wedding and all this kind of stuff and we were talking about this on WhatsApp. So I'm like this is so much data, there's so many people talking about this stuff and there's so many people I don't know because, like the groom's family was involved that I didn't know any of those people and all this kind of stuff. So I took, I just copied the entire WhatsApp thread and I poured it into notebook LM and I played the podcast on my way to the on the venue and it was amazing, like really amazing. Uh explained everything, gave the top hints. Notebook LM is quite a thing.

1:58:20 - Leo Laporte
Well, you'll be glad to know that Microsoft just stole three people from Notebook LM.

1:58:24 - Mike Elgan
Yes. Nice segue.

1:58:29 - Leo Laporte
Three more Google DeepMind employees have joined Microsoft's AI office in Zurich, including some of the creators of Notebook LM. Mustafa Suleiman hired three people from DeepMind he's the head of AI at Microsoft, marco Tagliaschi, zalan Borsos and Matthias Minderer and so they're now founding members of microsoft's new ai office in switzerland. I'm not sure why switzerland. Uh, they all worked at deep mind, uh. Suleiman, who co-founded deep mind with demis sasebis, uh left there, uh in 2022 to form an ai startup and then eventually went to mic Microsoft in 2024. So I don't know.

1:59:22 - Jeff Jarvis
I mean we had the Notebook LM guy on the show. Jeff, your friend, we should have him on again.

1:59:25 - Leo Laporte
He's still there. Yeah, we should absolutely have him.

1:59:26 - Jeff Jarvis
He won't. Well, three of the lead developers left to start a whole new company. Now these folks have left. I don't know. I suspect that says more about Google these days, yeah.

1:59:41 - Leo Laporte
Zalambor. I don't know. I suspect that says more about Google these days. Zalan Borso says I'll be focusing on advancing the audio capabilities of Microsoft AI. Marco Tagliaschi says please do not say I've joined Microsoft AI as a founding member of the New Zurich office. I'll be working on audio. See, all these guys are doing audio.

1:59:52 - Jeff Jarvis
See Andreessen Horowitz may be right.

2:00:01 - Leo Laporte
With audio playing a critical role in shaping a more natural, intuitive and immersive interactions. I still feel like it's a bridge technology, but we've seen that I mean they've been bridged lots in technology, lots of bridge technologies um, which help kind of ease us into a whole new way of of working well.

2:00:13 - Benito Gonzalez
So I also voice interviewers.

2:00:16 - Leo Laporte
That's interesting. What are they interviewing people for Talent? Talent interviews, presumably for tech startups.

2:00:25 - Jeff Jarvis
I forgot to put it in here, but I think Amazon is having an event the next week or so about Madame A.

2:00:40 - Leo Laporte
You know, a bold new step and, madame, so they're probably finally announcing the uh echo pro they call it the a word pro, which will be, it is rumored, five dollars a month for advanced capabilities. But this has been long delayed because it didn't work very well, february 26th is an event for it.

2:01:01 - Jeff Jarvis
I'll put it in the chat.

2:01:02 - Leo Laporte
Oh, this will be interesting.

2:01:03 - Jeff Jarvis
We put it in the chat.

2:01:05 - Leo Laporte
Interesting.

2:01:07 - Jeff Jarvis
By the way, speaking of chat, did you?

2:01:09 - Leo Laporte
see who's in chat having a lovely birthday dinner and missing us. Oh, paris Martineau says hello, guys from fancy birthday dinner and missing us. Oh, paris Martineau says hello, guys from fancy birthday dinner missing. Chatting with y'all, but also enjoying my truly sand free evening. Look at that?

2:01:24 - Jeff Jarvis
that looks good would you say back I demand the caviar. I want to see a picture of the caviar, okay, hamburger right.

2:01:32 - Leo Laporte
I think she's gone already it was a caviar roll, only 715, oh yeah. So button mash asked why would uh voice be a bridge? Isn't that the communication mode that's most natural to us humans? Yes, so I guess we would communicate with our agent, um, but that agent could then communicate with the outside world and, honestly, our agent's going to get smarter and smarter about what my needs and wants are, and it's going to know I'm going to tucson, I'll be there on valentine's day. I better book a lovely meal for lisa and leo and I won't have to do anything at all. Right, I? I? I do think that I don't. I'm trying to think of some other. Do you want to?

2:02:17 - Benito Gonzalez
not do anything at all.

2:02:19 - Mike Elgan
Yes, exactly this is the problem with all this stuff.

2:02:21 - Benito Gonzalez
I want to play the piano this is the problem, but you don't want to pick the restaurant that you go out with Lisa with. You don't want to pick the restaurant, you want choices.

2:02:29 - Leo Laporte
I would love my AI. Maybe this would-be voice Say Mr Laporte, you're going to be in Tucson on Valentine's Day. I found three excellent restaurants with an opening in the evening. Would you like to pick one and I'll make a?

2:02:42 - Jeff Jarvis
reservation for you, and I harass them all by calling them as Google. No, no, it won't have to call them.

2:02:47 - Leo Laporte
That's what I'm saying.

2:02:48 - Jeff Jarvis
It will just go with their agent, it'll interface. Yeah, yeah, yeah.

2:02:52 - Mike Elgan
Okay, but let me give you another scenario. Okay. So here's another scenario. You have this tool that's called a, you know, husband helper, and you can just go in there and say just do all the things that a husband's supposed to do, so automatically Lisa's going to get flowers on Valentine's day and all that stuff is going to be taken care of. Now, how does Lisa feel about going to feel about this Right, about gonna feel about this? Right?

2:03:14 - Jeff Jarvis
you just well, she'll have nothing, she'll have wife 2.0 which will yell at you forget your agent, forgetting the wrong thing it's the same thing as like remembering birthdays doesn't matter anymore, because Facebook reminds everybody so like it doesn't actually mean anything

2:03:26 - Leo Laporte
to anybody right yeah, so we will soon come to the point where we think remembering birthdays is not necessarily a virtue. It's just something like something people do, do? I mean, people don't remember their phone number anymore. When I was a kid it was drilled into us.

2:03:43 - Benito Gonzalez
Yep can't yeah and navigate without google maps anymore, I think I can't kind of seriously that's something bad, though seriously, I would love to be able to navigate still without it.

2:03:53 - Leo Laporte
I have no idea. Worse than that. I'll go somewhere with Google Maps and I won't know where I am, and like if the map stopped working, I wouldn't know where I was.

2:04:01 - Benito Gonzalez
So that's kind of the danger of that.

2:04:03 - Leo Laporte
Yeah, Fortunately the maps are reliable.

2:04:07 - Jeff Jarvis
So a friend of mine was in London and he's talking on a video call with somebody and a bike rider comes by. You know what happens on a video call with somebody and a bike rider comes by you know what happens grabs the phone and can't call. He said never have a text sent to you to get into something because he can't get to Verizon and change his phone number till he gets back to the US and he's down to nothing and he can't go anywhere and it's like he just suddenly got robbed for the next week of everything, did he ask you to wire him some money?

No, he didn't. He just wrote a sad blog post about it. Okay, I'm safe. You protected me from myself as an old guy. It's a good thing to do.

2:04:52 - Leo Laporte
I don't know who Pymance is when they're at home P-Y-M-N-T-S Pymance TV but they have a story. Did you put this in here? Meta smart glasses spending approaches $100 billion.

2:05:08 - Jeff Jarvis
Payments. I think it is P-Y-M-N-T-S Payments Could be Pymance.

2:05:16 - Benito Gonzalez
And if the article is by payments, then that means it's probably an AI article, right?

2:05:20 - Leo Laporte
Oh, you think it's written by AI.

2:05:22 - Benito Gonzalez
Well, the byline just says by payments.

2:05:25 - Leo Laporte
Oh, that's a bad sign.

2:05:30 - Mike Elgan
They also have two really expensive Super Bowl commercials for Ray-Ban Metaglasses, which I think is going to really finish making the mainstream. I thought that was really interesting.

2:05:39 - Leo Laporte
I am told that Super Bowl ads will be overrun with AI, just as they used to be overrun with dot-coms and Bitcoin. Remember all the crypto ads a couple of years ago? Courage favors the brave, buy Bitcoin.

2:05:56 - Jeff Jarvis
Actually, if you had followed those those ads, you'd be rich by now, so maybe I shouldn't mock them so then we have, uh, the story above, that is, bosworth boz uh saying that the metaverse could be a legendary misadventure if the company doesn't boost sales. Leaked memo shows. So you're talking about meta's ceo.

2:06:18 - Leo Laporte
Um yeah, well, he's right, could be right, could be, could be right. I think they've already pivoted haven't they? Andrew bosworth is the cto of meta. He told staff in a memo this year that it's the most critical year for the company's metaverse bet, uh, andrew. Should somebody tell andrew?

2:06:42 - Jeff Jarvis
yeah, yeah, who wants to tell andrew?

2:06:47 - Leo Laporte
um, I I think mixed reality is kind of a dead deal yep you know, it's surprisingly not a dead deal is crypto. Um, I've been thinking about crypto and what I don't like about crypto besides the fact that people I was afraid you were going to go crypto positive.

Oh, thank you, I'm not going to and and and that, uh, you know it's a highly speculative and risky investment and it empowers ransomware, etc, etc. The real problem I have with it is people are the whole idea of your money. Money is is kind of like a battery for energy. Right, it's stored value and it's. It's really a waste to put money into bitcoin hoping that just kind of accidentally it'll go up. You should put money into companies where you're supporting what their endeavors are. You're helping them build their company to create some value in the world. You're putting money in Bitcoin. It's pure speculation. It doesn't support anything. It's buying pork belly futures it's now. But even pork belly futures, at least, are tied to pork bellies. It's worse. It's like buying tulip bulbs. It's completely unrelated to anything, to value creation. It's all speculation.

2:08:03 - Benito Gonzalez
Now I realize why I never really liked the idea to begin with I mean it is kind of an extremely efficient ponzi scheme, and ponzi schemes are very popular and they happen all the time for a while is it a ponzi scheme?

2:08:13 - Leo Laporte
maybe it is. I mean, it's certainly speculative and the people who have Bitcoin would very much like you to buy theirs.

2:08:20 - Benito Gonzalez
No, it's not like a Ponzi scheme, but it's the evolution of that idea, whatever that is.

2:08:25 - Jeff Jarvis
And they're both right. There's no value creation.

2:08:28 - Mike Elgan
That's my real problem.

2:08:29 - Leo Laporte
There's no actual value creation.

2:08:30 - Mike Elgan
Yeah, so I go to El Salvador a lot and that's, of course, famous for its Chivo wallet and Bitcoin thing. And they have these Chivo wallet ATMs all over the country and they're just gathering dust. Half of them don't work anymore. Nobody ever uses them. Nobody uses Bitcoin in El Salvador. 92% of Salvadorans didn't use Bitcoin for any transaction in 2024. 92% it's a completely failed experiment. Everybody all the Bitcoin bros were like super excited when El Salvador sort of legalized it as made it legal tender, and there's been less follow-up on that. It's basically a non-existent thing. The government keeps investing in Bitcoin. On that, it's basically a non-existent thing.

2:09:17 - Leo Laporte
The government keeps investing in bitcoin, but, um, but in fact, I remember bukele celebrating how much money they'd made because they did buy a lot of bitcoin when it was a much lower price, right, right, but the uh? But the ruling congress in el salvador last wednesday approved a reform that took away the phrase legal tender. So, yeah, if it's legal tender, then a merchant is required to accept it as payment. If it's not, they can say no, I don't take bitcoin.

2:09:43 - Mike Elgan
So it's the beginning of the end to me, el salvador is the perfect, uh example of what you're talking about, leo. This is a country with extreme poverty that needs all kinds of help in a whole bunch of different ways. They they have environmental problems. Their rivers are horribly polluted. It's it's uh, they have. It's a beautiful country, believe me, it's. It's a wonderful place, but but it's like they need people invest. If you're going to invest in something, they need people investing in companies that are feeding people, that are right, uh, bringing clean water well, guess what they're doing to make money mike el salvador has agreed to take, uh, uh, the migrants and american citizens, uh, into their horrendous, horrendous prisons.

2:10:28 - Leo Laporte
If you look at their home page at the wall street journal right now, you will see just a frightening photo of how el salvador is going to make some money and I and mike I know you have ties to it because the mirror is from el salvador and so forth yes, um, and you've been there and you, in fact, you did. I know you did a gastronome adventure there. So there's, I mean, you love the country and the people, I'm sure.

2:10:49 - Mike Elgan
Oh yeah yes, well, the the, yeah, the um, the. The rounding up of all the gang members has transformed the country. It used to be murder incorporated there, right?

2:11:01 - Leo Laporte
I mean, it was very dangerous In 2015,.

2:11:03 - Mike Elgan
It was in 2015,. It's sort of the violence peaked and then they've been chipping away at it and then they had this radical emergency rule thing where they just rounded up everybody with a tattoo with no legal recourse for getting out. But Salvador is a very, very, very popular policy. Bukele not so much, but the policy Was it? His policy it was, yeah, but other policies of his are not so great and he's a mixed bag. But this policy is very popular because, you know, if you think about it, el salvador was racked, with civil war starting in 78. The war ended 92.

The the gang members were, were um kicked out of the united states and sent to el salvador. Like a year or two after the civil war ended. They had no prospects for legitimate employment, so they just had this reign of terror, of mafia style sort of like oppression of the population. So, basically, this is a country that hasn't that's been fearing violence since 1978. And suddenly they're the safest country in Latin America. Salvadorans are ecstatic about this fact and in fact, we wouldn't do the Salvadoran experience if they hadn't done that. It's a humanitarian disaster from a human rights point of view, but a very, very, very popular policy because now finally the Salvadorans can go out.

2:12:31 - Benito Gonzalez
I mean Americans kind of had a hand in why all that happened. Oh yeah, we destabilized that. We just that's our fault yeah, well, the us.

2:12:40 - Mike Elgan
Us has been involved in all of it. It's the southern civil war was part of the cold war conflict yeah, so all that bad stuff happened because of us.

2:12:47 - Benito Gonzalez
I think. I just want to put that out there.

2:12:48 - Leo Laporte
Yeah, yeah uh, chicken egg pretty clear yeah, um, how did they justify the idea that they would put American citizens in Seacott? Though? I don't. I'm not sure. I don't understand. That would quite work. We don't like you. Uh, we're going to put you in jail in El Salvador by the way, I understand why he wants already one.

2:13:21 - Mike Elgan
Sent uh, go ahead. They've already sent um uh uh.

2:13:24 - Jeff Jarvis
Immigrants accused of crimes to guantanamo bay, that's already taking place have they started occupying?

2:13:27 - Mike Elgan
yeah, yeah, but they sent like 13 of them.

2:13:29 - Leo Laporte
There wasn't a huge number, but they are starting that but yes, yes, but it's a slippery slope.

2:13:33 - Mike Elgan
I mean, once they get away with that they have room for 30 000. There's plenty of space, the whole point of guantanamo is that it's not covered by the.

2:13:40 - Jeff Jarvis
You know the rule of law and so they also sent a military cargo jet with people, okay, so all right mm-hmm seacott, uh is uh.

2:14:00 - Leo Laporte
The world's largest correctional facility holds 40 000 members of street gangs, or just anybody with a tattoo I'm shocked that the world's largest correctional facility is not in america well, we will catch up, we're working on it, we're gonna, we're gonna outsource there, uh.

2:14:16 - Jeff Jarvis
But you know, mike, next time you're there don't do anything bad. Okay, promise, and and you tell me you don't have any tattoos maybe don't get a tattoo.

2:14:24 - Benito Gonzalez
Okay, no tattoos leo can't go there

2:14:27 - Leo Laporte
because he has a tattoo I don't think they're gonna ask to see my butt I hope not yeah, it's.

2:14:37 - Mike Elgan
Yeah. What gang is that leo? Uh, um, yeah, I'm in the tweet. There's kind of a dual uh system of justice in el salvador which is very controversial there. Um, if you're an american and you're caught doing something there, they pretty much just let you go and and and they. You know, there was an American citizen who was a Salvadoran you know his parents, I think, were Salvadoran immigrants or whatever went to El Salvador on vacation. He had some tattoos not gang tattoos and he was caught and basically called the embassy. Embassy leaned on the government. They're like, oh, nevermind, and so this was very controversial in El Salvador because for Salvadoran citizens, that's not how it goes, and so it's very, very, very safe for Americans to go there. In terms of Americans worried about getting caught up in the criminal justice system, it unfairly favors American visitors to the country.

2:15:40 - Leo Laporte
Your favorite. Benedict Evans posts on Threads a note from a JP Morgan analyst about Tesla stock. The headline is Tesla shares somehow up 5 percent after a 38 percent I was puzzled by that. And on the lowest margin is years, uh, as a 20 to 30 percent increase in 2025, deliveries outlook is revised to return to growth, he writes.

It's not clear to us why tesla shares traded as much as 5% higher on the aftermarket Wednesday. Perhaps it was management's statement that it identified an achievable path to becoming worth more than the world's five most valuable companies taken together. In other words, more than the $14.8 trillion combined market cap of Apple, microsoft, nvidia, amazon or Alphabet. Of course they're going to be that value. 8 trillion dollar combined market cap of apple, microsoft, nvidia, amazon or alphabet. Of course they're going to be that value. Or maybe it was management's belief that just one of its products has, by itself, the potential to generate north of 10 trillion dollars in revenue. How could they even it may even have related to management's guidance for 2026? Uh, no financial targets were provided, but it was said to be quote epic and for 2027 and 2028, ridiculously good, what the what?

2:17:05 - Jeff Jarvis
but you see it's a meme stock and I think that's really uh clear right, well, it's a political means, yeah it's it's, it's a um, it's a means of sending a signal to say that you drive the car and own the stock yeah, all right, our neighbor, our neighbor down the street, just bought apparently it's you can get the cyber trucks cheap.

2:17:24 - Leo Laporte
Oh, I'll bet they've started leasing them, which they didn't do initially. Uh, and the original the uh, the uh, you know, special, first edition cyber trucks. They've been de-badging them so that they can sell them at a lower cost. Uh, so, and they've been offering and this is, I think our neighbor took advantage of this free wraps. So our neighbor has taken what is probably the ugliest car since the aztec and wrapped it in. It looks like fake stained glass. I don't know what it is.

2:17:54 - Mike Elgan
Camo, there's a lot of ramps on Teslas. Well, the other thing that's happening right now I'm seeing on social media is a lot of vandalism of Teslas. Yeah, don't do that, folks. Yeah, don't do that, but there are people who are vandalizing entire car lots. Oh dear For hyper trucks.

2:18:17 - Jeff Jarvis
What I want is I want a speaker on the top of my car, just so whenever I pass one, I can just start laughing.

2:18:21 - Leo Laporte
There are quite a few of them in petaluma, for some reason. I mean, I don't know, I mean they're in northern california is loaded with teslas period but there's a lot. There's a lot. In marin too. There's quite a few of them, and every time I see one I go. I don't, it's like something a four-year-old designed with. Yeah, it's bizarre.

2:18:39 - Jeff Jarvis
Have you looked inside? I haven't been inside.

2:18:40 - Leo Laporte
No, uh, we have a friend who uh well, our neighbor, who has one at least is dying to get a ride. He's gonna go over say take me for a ride in your cyber truck. All right, we're gonna. I think we're gonna wrap it up. Uh, this has been our first episode of a brand new show intelligent machines. I hope you're enjoying it. We'd love to hear your feedback.

2:19:01 - Jeff Jarvis
Email, jeff at jarviscom right though everybody knows where to find me, and they do do. Yeah, we're on the socials.

2:19:16 - Leo Laporte
You can always put it up there. We are going to take a little tiny break and then we're going to get our picks of the week. We still have them. We haven't abandoned them. No changelog, however. No, we've abandoned that, Thank God. And the Stop no, For old time's sake. Ladies and gentlemen.

2:19:39 - Benito Gonzalez
That's probably the last time you'll ever see that.

2:19:41 - Leo Laporte
Yeah.

2:19:42 - Jeff Jarvis
Yeah, but keep it around All right. Now that you're in the beginning, you got to go to the end. You have to. All right, thank you All right, thank you.

2:19:50 - Leo Laporte
It even feels dated. That really feels dated.

2:19:53 - Jeff Jarvis
Oh, it is, yeah, yeah, yeah did you really get yelled at people for not doing the changelog? Oh yeah, all the time you because, you hated doing it, you did it I had to do it.

2:20:05 - Leo Laporte
They made me do it all. Right, mike elgin, it's so great to have you, mike. Uh, everybody should go to gastronomadnet and see what wonderful trips Mike and Amira have planned. I went with them, lisa, and I went with them a couple of years ago to Oaxaca during the Day of the Dead, and it was just the best, the best, the best, the best.

2:20:29 - Jeff Jarvis
You have such great pictures from that trip.

2:20:31 - Leo Laporte
Oh yeah, mike's an amazing photographer. There's going to be another Oaxaca trip Morocco and you're doing El Salvador again.

2:20:38 - Mike Elgan
We are. We've done two El Salvador experiences. They've been both incredibly fantastic, but we've added Tuscany recently, since the last time I was on the show, so that's really nice and, um, so the Tuscany experience is coming, I think in 2026, but yeah, so so if if people are curious about what we do, we basically have a small group of people eight, 10, maybe 12 people. There were 14 on the one you did, leo, because people kept asking us to do it. It was the first one.

2:21:07 - Jeff Jarvis
Well, it's because Leo was there.

2:21:08 - Leo Laporte
They knew they had to be with Leo. No, I think Mike allowed.

2:21:11 - Mike Elgan
Lisa, oh, you let that in, okay yeah yeah, but yeah, we get together and we're often in an old farmhouse that's been refurbished or in a beautiful location in a vineyard or something like that, and we just do only the good things, and so we call it joy therapy. So if you're stressed out from doom, scrolling the news and all that kind of stuff, it's like an escape where you go into a world where it's just eating the most amazing food, drinking the most incredible wine or, in the case of Oaxaca, mezcal and and just living life the way it should be lived. It's the best travel ever. Invite everyone.

2:21:48 - Jeff Jarvis
Mike, like the law in my kitchen, do they have to like do an essay to get in?

2:21:58 - Mike Elgan
Do they to get in? Do they have to? No, they do. Admissions will come to the experiences. Yeah, um, we, you know, we, um, we find that the group, the people who join these experiences, are self-selecting and the most amazing people. They're kindred spirits, they're like super foodies, they're truck. Yeah, they're trusting people because, because we don't tell anybody anything that's going to happen, so we're everything's a secret surprise. And so what kind of person would sign up for something like that?

2:22:19 - Leo Laporte
Not my wife, she did not like that part Really awesome person.

2:22:22 - Mike Elgan
She doesn't know, she's a control freak. Everything that was coming up.

2:22:25 - Leo Laporte
Next time, Amira's just going to have to whisper in her ear. Okay, in order to do this, yes, yes. Actually help me, because lisa has said, and I, I'm among, I don't know what to say about this. She doesn't she thinks it's safe to leave the country for the next four years. What? How do people in? Yeah, if you're in, yeah, that's if you're in sicily. What are the people? They know you're american? Do they spit yeah?

2:22:49 - Mike Elgan
no, no, no, no, no, no, not at all. People are couldn't be nicer. Everywhere we go, I mean it's just, it's just fantastic. I mean to the extent that they're paying attention. I'm always surprised by, everywhere I go, where I'm like oh yeah, you know, our politics is so horrible, your politics, what about our politics?

2:23:08 - Leo Laporte
Well, Italy is not known for great politics, actually. Well, yeah, that's not.

2:23:14 - Mike Elgan
But everybody is. You know these trends that are happening. You know the American trends are happening in a very American way but they're kind of global.

2:23:23 - Leo Laporte
Yeah, it is global, isn't it?

2:23:24 - Mike Elgan
Overarching political trends, and so people are sympathetic to that. They know that. You know we're the primary victims, to the extent that they pay attention to what's happening in the US. And so no, no, I've never, ever, been treated in any of the places where we do experiences in any way other than the most gracious kind, you know well because we usually just say we're Canadian and that that's the conversation.

Yeah, it's. It's funny because when we went to Cuba years ago, the Canadians were so depressed because that used to be their place where they could go, right, and the Americans couldn't go and they had it made. But the other thing of it is, as you know, leo, we are always in the loving embrace of our personal friends on all these experiences. Absolutely.

2:24:11 - Leo Laporte
And he's talking about the best restaurants in town, the best chefs. It is a remarkable experience. Highly recommend it. Yeah, yeah, so good. Well, I'm going to. I'm going to work on Lisa. She's nervous about, about traveling at this point.

2:24:26 - Mike Elgan
Well, she should be, except the flying part from. For reasons we don't fly.

2:24:31 - Leo Laporte
Yeah, yeah.

2:24:32 - Mike Elgan
We'll go.

2:24:33 - Leo Laporte
Alitalia, it'll be okay. Yeah, machineso yeah, yeah, we'll go alitalia, it'll be okay. Yeah, uh, machine societyai is mike's a newsletter subscribe, so I mean he's really an insightful writer, and especially on ai. So, uh, now thank you yeah, now's the time. Yeah, all right, what's your pick, sir?

2:24:52 - Mike Elgan
well, okay, so, um, I have been using the dji flip. Oh, and the reason?

2:24:57 - Leo Laporte
so you had a, uh, you had a, a phantom, when I was with you in in mexico, I think right we had.

2:25:04 - Mike Elgan
I've had three drones. I crashed two of them. The first one I crashed into my daughter-in-law as did I and it went plunk and, in fact, in fact, the two that I crashed were both in El Salvador, so I crashed two drones and I had one stolen.

2:25:17 - Jeff Jarvis
I saw what you did. You can't see that Jeff's afraid, is it?

2:25:22 - Leo Laporte
over, it's over. You can look now.

2:25:24 - Mike Elgan
It's almost over. So this is my fourth thrown the dji flip and I absolutely love it. My my issue with it of uh in general, is that the drone influencers tend to love it, but they they don't give it enough credit, because it's a perfect drone for somebody who's not a drone obsessed person. It's very lightweight ai, as the ai is what makes it so amazing. So it has all the ai stuff that's in the mini drones and all that stuff, but it has exclusive features that are based on AI.

2:25:55 - Jeff Jarvis
Tell me when it's over.

2:25:58 - Mike Elgan
So one of the problems I've always had with shooting drones is like you get it out. You have to put your phone into the controller. It's real pain. You have to launch it. You have to have a non-filthy place to launch it from so the dust doesn't get in your drone. It takes like three or four or five minutes to get a drone in the air and then you have to put it away at the end all the stuff. With this drone, you simply unfold the propellers, you just pull them out and it turns on by itself. And then you hold it out in your hand and it takes off. It responds to voice commands. Oh, so you tell it fly high.

2:26:33 - Leo Laporte
Here we go. Voice again Go here land, come back.

2:26:36 - Mike Elgan
Yes, don't look.

2:26:37 - Leo Laporte
Jeff, close your eyes.

2:26:39 - Mike Elgan
The taking off and landing from your hand. It's like a falcon. It does that through AI, it knows what a hand looks like and it will only land on a hand. Oh nice, oh nice, very, very fascinating. And it's designed for vloggers. So one of the cool things about it is that and in fact they sell the mic mini, the DJI mic mini, which is a tiny, lightweight wireless mic that you clip onto your clothes. It's optimized for making the drone sound in the audio vanish completely, so you can sit there and do vlogging, walking around if you want to, or whatever, and chit-chatting, and you only hear the crystal clear voice without any sort of whine of the drone.

2:27:25 - Jeff Jarvis
How much do they run, Mike?

2:27:27 - Mike Elgan
Cheap, cheap, cheap, Like $440. Very inexpensive, and it has to be like they have another cheap drone I forgot what it's called, the Neo or something like that which has a crappy camera and a bunch of other. This has the same 4K high-quality camera as the Mini 4 Pro. I have the Mini yeah, it's a really high-quality camera. I have the Mini yeah, wall, and it just sort of backs off and keeps flying, and so it's really a phenomenal drone and I highly recommend it to people who were a little nervous about having a drone. Um, and I recommend it to sort of content creators and influencers and anybody else, Cause it's, it's so, it's so exhilarating to just pull it out undo the thing, launch it in the air and it just follows you or does whatever you want it to do.

It'll do a circular thing and then you go like this and it lands in your hand. You fold it up and put it away. That's how you turn it off. You just fold up the wings so you can have it in the air in literally less than 10 seconds after you decide to shoot something. So it's a really, really fun drone.

2:28:47 - Leo Laporte
You tempt me, you to me, you tend to me. My it's the ai. This is bad for leo. Well, I still have my dji mini and I very rarely, uh, take it out, but part of that is all those things you described. It's, yeah, it's a pain in the it's pain, but now it still does have a controller.

2:29:02 - Mike Elgan
It looks like yeah, it comes with a controller as part of the price and you can use it with a controller and, best of all, you can also control it with your phone with the app, right? So can. Basically, if you're going to go, if you say, okay, we're going to do some serious drone photography, bring the controller, you have all the control in the world. Or if you're just going to go for a hike or whatever, just everyday kind of stuff, you just stuff the drone in your pocket, don't bring the controller. That makes it so much easier. And then you have the option to use the automated ai features to fly it with no controller at all. Or you can pull out your phone and have control with your phone, but you you don't have to carry the, the, the controller, which to me is very liberating do you have uh some uh videos on your nice book that I could play?

looks like uh, my, those are all photos. Um, okay, from the, from the drone, yeah, saying no, I don't have any drone shots there, but um, by the way, I don't know mike also.

2:30:06 - Leo Laporte
Uh was with me on the beach walking and got sand in his shoes I just yes it's not just me, okay, I just that's it.

2:30:14 - Mike Elgan
That's the sahara.

2:30:15 - Jeff Jarvis
Actually right, that's the sahara, wow, yeah by the way, leo, I realized that every guest we have on from now on we have to find out whether it was the sandman I will bring.

2:30:25 - Leo Laporte
I'm gonna ask the sandman to come on and I will, uh, when he comes on, if we we can get him, I will out him.

2:30:32 - Mike Elgan
He is an AI accelerationist. There's a video of you and me with the drone yeah, in Oaxaca. This one, yeah. Now I'm Jeff like getting frustrated by your scrolling. I think it was up. I think you could go up a little bit more Above.

2:30:48 - Leo Laporte
Yeah, keep going, keep going. I think we're in Oaxaca here, there above. Keep going, keep going, we're in oaxaca here there we.

2:30:54 - Mike Elgan
Oh yeah, that's you and me and the drone yeah, right there, yeah, yeah, so go ahead and click on that, that's.

2:31:02 - Leo Laporte
That was the um which one was that that was cooking class.

2:31:07 - Mike Elgan
It it, uh, yeah, no, I know where we are, but which dji is that that? You're using. This was the mini 3 pro. I mean, that was a couple of drones ago, yeah, a couple of drones ago. Yeah, is that all? Is that all chunky?

2:31:13 - Leo Laporte
uh, it looks no, it's probably the internet. Uh, it's sicily, it's chunky yeah it's very smooth. It was fun. Yeah, I think you also had me running around. I don't know if you posted that one. I don't, probably didn't and we did some fun videos out in the Agave. Is that Agave? What is that? Yeah, the Agave fields, agave fields. Really fun, all right. All right, maybe I'll get this as well. What the hell, it looks like fun.

2:31:48 - Mike Elgan
DJ. I really love it. It's so easy. Yeah, $439. It's cheap. If you you crash it, get another one. It's like it's not that cheap. Yeah, but you don't worry about it so much. Like I, I used to spend a thousand dollars on a drone.

2:32:01 - Leo Laporte
it's true, they used to be consequential crash, yeah, and so get it while you can. Much easier, get it while you can. Uh, it won't be that cheap for a while after a while, because I imagine it'll be tariffed. Jeff Jarvis, your pick of the week, I'm going to do two that are AI-related.

2:32:22 - Jeff Jarvis
One on line 185. Fascinating LinkedIn post here by a marketer that suggests that marketing in the age of AI agents, marketers will have to advertise to agents. Yes, my point exactly. Exactly your point. I knew you'd like this. Today, humans see and interact with ads. In the future, ai agents will filter everything. They see the ads, not the humans. But it leads to all kinds of fascinating conflict of interest questions is but at least all kinds of fascinating conflict of interest questions is. So I use uh uh, you know leocom is agent and Leo gets.

I got to deal with Coca-Cola and you're going to see Coca-Cola ads, whether I like it or not, but Mike has the ethical one, but he's going to charge me more for his agent because he's not taking the money from Coca-Cola.

2:33:10 - Leo Laporte
I'm telling you, there there are so many ramifications. This is a show that we we could do this for years and never uh cover everything.

2:33:19 - Jeff Jarvis
I think it's uh we're entering very interesting a few weeks ago I mentioned that there's a new just as there's seo, there's geo generative engine optimization. So now there's uh in this post agent optimization. Brands must train ai agents to prefer their products in natural decision. Now there's in this post agent optimization. Brands must train AI agents to prefer their products in natural decision-making. There's API-based marketing. Instead of ads, brands must integrate into AI ecosystems through data, api and partnerships. Just like media providers and reputation trust signals, ai agents will prioritize trusted brands, reputation, verified reviews and so on. So it's going to be a whole bunch of you know, consulting firms are going to start and data strategy powerhouses and warehouses, and it's going to be a whole new world of marketing, which is going to be fun to watch them fall over themselves. The other one I have is the story that drove me knucking futz this week.

2:34:09 - Leo Laporte
Oh no.

2:34:09 - Jeff Jarvis
What's that? There has to be one. Maybe that should be the next thing I do. What drove Jeff Knucking Futz Line 110. Celebrities signed on to a letter warning that AI systems could be caused to suffer if consciousness is achieved. Oh please, there is enough suffering, real suffering, suffering, human suffering in the world. The last thing I'm going to worry about is hurting software's feelings. It doesn't have any. This is just ridiculous.

So stephen fry signed on to this oh come on fairly unknown people if you go to the next line down in the right now.

2:34:46 - Leo Laporte
You said 110. I don't see it here. So what's the number?

2:34:49 - Jeff Jarvis
119. Sorry, 119.

2:34:53 - Leo Laporte
Okay, so that's really hysterical and you know, okay, fine.

2:35:03 - Jeff Jarvis
So, if you go to the next line, line 120, there's the letter itself, and so it's telling you that organizations should prioritize research on understanding and assessing AI consciousness with the objectives of preventing the mistreatment and suffering of conscious AI systems. Oh, give me an effing break, dorks.

2:35:20 - Leo Laporte
Well, could they suffer? Do you think there?

2:35:23 - Jeff Jarvis
are only 112 signatures. So yeah, they haven't had a lot of signatures, poor.

2:35:27 - Leo Laporte
Stephen Fry. I think he's charging up that hill on his own. Yeah, stephen Fry got taken. Nobody else famous is on this list and this whole thing or something. Where's stephen fry? I think he's. He's charging up that hill on his own.

2:35:32 - Jeff Jarvis
Yeah, stephen fry got taken. No, nobody else famous is on this list. And this whole thing, this consensium, or whatever it's called consium, was started by an ad guy.

2:35:42 - Leo Laporte
Ah, okay, so it feels like that was that was the thing that drove me knocking futz this week I see you also, uh, have a axios, uh piece about bill gates. He's, of course, and I'm going to try to get him because he's got a new book, he's got a three volumes, leo a three volume autobiography, oh gosh.

But uh, he actually is very thoughtful about ai and I think he'd be great to have on. Oh yeah, I was watching him on jimmy fallon, uh, last night plugging the book, but he said something that was a mind boggler, talking about being boggled. I'll have to talk about it on Mac break weekly. But there's a very famous all things D conference with Kara Swisher and Matt I mean sorry, walt Mossberg and of course Bill Gates and Steve Jobs together on the stage and Jobs kind of snide about Microsoft and Bill Gates. He said you know, microsoft never had any taste. He said they should have taken acid like I did. He said I don't trust, famously. Jobs said I don't trust anybody who's never taken acid. So last night this is a revelation bill gates said well, I was kind of sitting there going, I couldn't say anything at the time, but I've taken acid. What's he talking about?

2:37:06 - Jeff Jarvis
oh, could you imagine that nerd fest.

2:37:08 - Leo Laporte
Wow, oh my god, I was shocked, shocked. There's all kinds of personal things.

2:37:13 - Jeff Jarvis
He regrets his divorce terribly. I, I saw that I that is really interesting.

2:37:18 - Mike Elgan
Yeah, he, uh, he feels bad about that um, no doubt rejects his uh, his uh sort of friendship with jeffrey epstein well, in fact, that probably is linked to the divorce.

2:37:30 - Leo Laporte
yeah, uh, that, that's, at least was the rumor that Melinda said. You know, it's me or Jeffrey, yeah, yeah.

2:37:39 - Mike Elgan
So Bill was a famous ladies man. It was well known.

2:37:43 - Jeff Jarvis
Amazing.

2:37:44 - Mike Elgan
Yeah, back in the day in Comdex he was, you know was always getting hammered and behaving inappropriate.

2:37:50 - Leo Laporte
So I just had this with him at Comdex, not with him next to him were you on acid too. I was not on acid but uh, to steal a joke, I think from jimmy fallon.

2:38:01 - Jeff Jarvis
He was doing the bite dance ah, with the thumbs, the lumps so I just I added bill gates to the list because you'll be, or you're, old, you'll forget um, hey, hey, I was also thinking that some other other names from the past, like Eric Schmidt and Marissa Meyer. I'd love to hear what they think of this AI time.

2:38:21 - Mike Elgan
Well, Eric Schmidt has an AI killer robot company. Yeah, he does.

2:38:26 - Leo Laporte
He's definitely into it. Yeah, so he's interesting. We will try, you know, I I I think this is an opportunity to get some really interesting people on the show, so we will endeavor to do that.

2:38:37 - Mike Elgan
So when Jeff was talking about this whole thing about software I'm sorry to go back when you're trying to go forward, but about software and its feelings there's actually a bill. My son, kevin, sent me this story two days ago in the Verge. That basically says that there's a bill pending in the California legislature.

2:38:58 - Jeff Jarvis
Of course.

2:38:58 - Mike Elgan
To require AI companies to remind kids that a chatbot is not human, and this is like, which is amazing, it'll have all the impact of a of a warning label on cigarettes.

2:39:14 - Leo Laporte
but uh, pretty interesting that that the state is a teacher, before they have an interaction. This is steve padilla. Once again, does the teacher have to say now remember, john, johnny, you're. This is not a human you're talking to right.

2:39:30 - Jeff Jarvis
Does that mean I can call it a piece of no, johnny, you shouldn't do that either, even if it's not human. Well, well, no, why but why not?

2:39:37 - Mike Elgan
I mean, I think they should, I think we should teach kids to call it a piece of something it's not a human being, it does not have feelings. Steven fry not standing and, and that's the whole, that's the whole basis of kevin's company is to teach kids. That is not a person. Yeah, ai, is just a, a thing that you, somebody like you, can build.

2:39:55 - Leo Laporte
Let's give kevin a plug. It's hello, chatterbox, all right. The smart speaker that teaches kids about all this stuff. Uh, really cool. It's the only smart speaker approved for education. Why? Because the kid writes the code. It doesn't out and it's totally private.

2:40:11 - Mike Elgan
Yeah, that's right. So it uses a bunch of. It uses wolfram alpha, it uses all these different services chat, gbt, etc. Etc. Etc. And those companies nor, and also chatterbox, don't know who the user is. It's totally private, which is why it's legal in schools so origami, this isami, this is bringing back childhood memories. Last week on the show.

2:40:34 - Jeff Jarvis
I talked about my eighth grade science fair winning entry electronic bongos. Part of the story I did not tell was that I had to take a radio amplifier and get a speaker and make a case for it and I was having problems. And my dear departed father this will explain a lot about my psyche came down to the basement as he saw me trying to put this together and he said you're so clumsy you couldn't stick your finger up your ass with two hands. Who said that? Thanks, dad.

2:41:06 - Leo Laporte
My father, my own father. That's not nice, no.

2:41:12 - Mike Elgan
Well, I mean, mean he probably was tired of the sound of bongos, maybe, right yeah, that's true.

2:41:19 - Leo Laporte
You know what. You have to forgive any parent. Electronic bongos who has suffered through electronic bongos? We gave henry, when he was about eight, a toy saxophone to play the same song over and over and over again. Big mistake, big mistake, very annoying. Eventually threw it in the closet and, of course, it hit the button and started playing. Ladies and gentlemen, that concludes episode one, dash slash, stroke 805. The first episode of our rebranded this week in google now intelligent machines. And uh, I think very often, uh, as we do this show, we will feel like the dumb machines trying to figure out these intelligent machines, but that's part of part of the fun. Mike elgin, thank you so much for spending, uh, your late night in sicily with us.

2:42:13 - Jeff Jarvis
I don't know what time.

2:42:14 - Mike Elgan
Make your dinner like geez it's 2 am, it's 2 am, but, but I'm, but we, we are here in sicily, we don't have an experience here, uh, for this trip, and so we're actually trying to cling to uh pacific time.

2:42:29 - Leo Laporte
Oh, that's smart oh okay, so it's not even dinner time, yet it's not even. Oh, that's smart. Oh okay, so it's not even dinner time, yet it's not even five. That's right, that's right, he's a cool.

2:42:37 - Jeff Jarvis
European staying up at night Dinner at 3 am.

2:42:39 - Mike Elgan
Like the rest of us, that's right, it's the life, it's the life.

2:42:44 - Leo Laporte
So you're coming back soon, is that why?

2:42:47 - Mike Elgan
Yeah, we're coming back in a couple of weeks, um and uh, and then we're gonna be in california for a couple weeks and then it's off to oaxaca. So yeah, so nice always a pleasure.

2:42:57 - Leo Laporte
Give my love to amira. It's great to see you. Thank you, will do, will do and thank you, jeff jarvis. Professor of journalist, professor emeritus of journalistic innovation at the craig newmark graduate school of journalism at the city university, you know I finally memorized that after you left, now teaching at Montclair state university in New Jersey and at the state university of New York, stony Brook, did you decide whether to do a class or not, have you?

2:43:28 - Jeff Jarvis
I don't know. I've got time to decide. That's not helping with there. There's a new degree program I'm working on on some other things.

2:43:34 - Leo Laporte
That's nice and you're writing a new book, so yep, staying busy. Jeff's latest is the web. We weave excellent book, highly recommended. Also the Gutenberg parenthesis in paperback. Ah, all available at jeffjarviscom and magazine, let's not forget. Thank you Jeff, thank you Mike, thanks to all of you for joining us and a real thanks to our club members, because club makes this show possible.

One of the reasons twigs survived so long without any advertising, it's because of you. Advertising, of course, helps a lot, but it doesn't cover all of our expenses, and your membership only seven dollars a month in club twit makes a huge difference. If you're not yet a member, twittv slash club twit you get ad free versions of the shows. You get access to our great discord. You get access uh to special shows we don't put out anywhere else, including tomorrow. 1 pm pacific, 4 pm Pacific, 4 pm Eastern, 2100 UTC.

Chris Marquardt our little photo segment we do every month. Chris will review submissions for the Luminous theme and he likes to talk about photography news too. So if you're a photo bug like Chris and I are he's a real photographer and I'm just an amateur please join us tomorrow afternoon in the club, 1 pm Pacific. Actually, we're going to stream it everywhere else as well. We always do that YouTube, twitch, tiktok X, linkedin, facebook, kik, all the usual places so you can watch that, but you have to be there live. 1 pm tomorrow, pacific, 4 pm Eastern. Thanks everybody for being here here. Thanks for your support. We look forward to seeing you next week. Paris will be back for intelligent machines. Bye-bye.