What Happens When AI Outsmarts Mathematicians

00:00:00:07 - 00:00:25:25
Unknown
Coming up next, Jeff Jarvis and I dig into OpenAI solving some unsolved math problems. And what does it actually mean for mathematicians? Also agents breaking free from containment. Again, Sam Altman demoing the next big model for Congress the EU AI act, getting enforcement teeth. Google giving Gemini a humanoid body and LinkedIn's new AI slop button. What's it all mean?

00:00:25:25 - 00:00:41:22
Unknown
We talk about it next on the AI Inside podcast.

00:00:41:24 - 00:01:01:00
Unknown
To another episode of AI inside the show, where we take a look at the AI that is layered throughout the world of technology one day earlier than normal. So surprise, happy birthday! Whatever you want to say. I'm Jason Howell, one of your hosts. I joined, as always by Jeff Jarvis. Good to see you, Jeff. Happy birthday. Yeah. I don't know.

00:01:01:01 - 00:01:21:03
Unknown
Where did that come from? I'm just I'm just. It just seemed like a, like, a surprise thing that you say surprise. Happy birthday. But it has nothing to do with third or fourth four podcasts. Today is is is fraud. He's fraud. Yeah. I'm. Yeah. This happens every once in a while where things come up just as a little kind of background.

00:01:21:04 - 00:01:43:06
Unknown
My my daughter is a swimmer, and she qualified for a pretty big swim meet happening in Irvine, California. And so we suddenly, very suddenly have to leave tomorrow through the rest of the week. But before that swimming order. Yeah, she's doing she's doing amazing. And so normally we would record this on Wednesday. We're doing it one day early.

00:01:43:06 - 00:02:04:14
Unknown
So surprised everybody in their feed. You also have another podcast you do on Wednesdays that recorded I think Early Day. Yeah. So it's a it's a discombobulated summer week. So suddenly you have you get to know what everyone else experiences as a normal Wednesday instead of podcast craziness. Yeah, just a hump day like everybody else. Yes, exactly. Exactly.

00:02:04:16 - 00:02:34:05
Unknown
So yeah. So we're going to do it one day early. Thankfully, you know, the AI gods have have granted us a number of great stories to talk about. I've got some news about our Patreon that I'll talk about here in a little bit. I'll save it, but definitely something to pay attention to their. And if you hear things happening in the background, the daughter my daughter that I told you about, she's making banana bread for for the meat because it's a good way, you know, it's a good food for her with peanut butter and stuff.

00:02:34:05 - 00:02:53:25
Unknown
So if you hear things in the background, that's what's going on. You get to heck yeah, I love banana bread with some chocolate chips in there. And then put a little light layer of peanut butter on top and yeah, yeah. Me I don't know. Do you like banana? I like that, yeah. Do you like zucchini bread. Yeah. Banana bread is better.

00:02:53:26 - 00:03:16:18
Unknown
Yeah I agree banana bread is better. Yeah. But I'll but I'll eat zucchini bread. Carrot cake. Carrot cake. My grandmother made carrot cake. And she made it incredibly, horribly sweet. When I was a little kid, once I said I liked it. So every single time I saw her, she had to throw Jeff. I made carrot cake and good, I couldn't stand it.

00:03:16:20 - 00:03:36:24
Unknown
Yeah, that one time, that one clue. Well, you know, she was paying attention. Or at least she was the first time she was. Maybe she wasn't picking up on the hints later. Yeah. For me, for me, my grandma's. And then we'll actually talk about artificial intelligence. My grandma clued into the fact that I liked potato salad. And so she would make potato salad.

00:03:36:24 - 00:04:00:00
Unknown
For the longest time I loved her potato salad. But then as she got older, her taste buds started to kind of change and her potato salads became very, very onion like, super onion y. But she had so much joy giving me the potato salad that I always loved. So I would just. Yeah. So I wouldn't say anything. You.

00:04:00:02 - 00:04:25:11
Unknown
Put a bunch of mustard in there and and cover it up anyways. Love you grandma, wherever you are. Let's let's start with open AI. Why the heck not? Few things happening in the open AI world. They published I think this was yesterday. Was it yesterday? So basically there's a new model and we've been hearing about this like unreleased new model that isn't public yet necessarily.

00:04:25:12 - 00:04:52:02
Unknown
We already know about soul and there are other things, but these like breakout stories have been mentioning a new model. And that new model is apparently what is it called? Astra. Astra. That's right. So apparently it's called Astra. And Astra has all of these capabilities because it's the new big thing as they all do this. Right. So now OpenAI is kind of like on a press tour to tell you why it's Astra is super amazing.

00:04:52:05 - 00:05:14:08
Unknown
Publish results from an from from that model saying that it solved ten previously unsolved math problems soon. We're not going to have any math problems to solve. These things are just going to solve them all. Really interesting. And the thing is, I can't understand how difficult these are because I can't understand the problems. Yeah, it's impossible to know exactly what that means other than these people said it was a big deal, right?

00:05:14:13 - 00:05:37:25
Unknown
Right. And it's a weird it's a weird field because every field has its mysteries, right? Obviously, biology has a mystery. We want a medicine we want to solve. We want to cure this cancer, whatever. But there are these odd logical, described problems that that whoever comes up with them can't solve. And so they go out there and it's almost as if there's an honor, there's an honor, and coming up with an unsolvable puzzle.

00:05:37:28 - 00:06:02:13
Unknown
Oh, totally on its own, right. Yeah. And then the Second order isn't solving it. And mathematicians are worried. But it's a weird kind of worry. It's not like it's not. I don't know how applicable any of these unsolved problems were in life. Right. So it's not as if a company says, oh, good. Well, we can fire 20 mathematicians now because we know we got this now.

00:06:02:13 - 00:06:22:23
Unknown
What does it apply to you? That's a really good question. So at the end of the day, like, it's interesting to know that these problems exist and that maybe they go they like that they're unproven, which means there's a potential that they either go against common understanding of mathematics, I guess, or I don't know, like what at the end of the day, what?

00:06:22:25 - 00:06:48:08
Unknown
Why is it so important to solve these unsolved things? I mean, I would imagine that some of them are useful in, like, high dimensional sphere packing. New. This is the first problem that solved new upper bounds on sphere packing density down to the cone LCS threshold. I don't understand that bit of that. Get it? But his sphere packing something you need in industry or science or chemistry?

00:06:48:08 - 00:07:14:20
Unknown
I have no idea. Yeah. Binary and spherical codes exponentially improved bounds on the maximum size of binary codes at any prescribed minimum distance, with analogous results for high dimensional spherical codes. I'll stop there, folks. I don't. So yeah, I don't know what the. Does any of this have applicable value in science or industry? I have no idea.

00:07:14:23 - 00:07:33:24
Unknown
What does this mean? Gem. And I explain this to me like I'm a 12 year old and we'll just put that binary and spherical codes in there and see if Gemini knows how to explain this. Imagine sending a message to a friend across a noisy room using only a flashlight that blinks on and off. Oh, okay, that would be binary code.

00:07:33:25 - 00:07:58:10
Unknown
Okay, a spherical code works the exact same way, but instead of using ones and zeros, you place points on the surface of a globe or multi-dimensional sphere. The goal is to pack as many points onto the surface as possible, while keeping a set minimum distance between every single point so they don't overlap or get confused. Okay, for decades, mathematicians had mathematical limits, called bounds, that estimated how many messages or points you could pack together.

00:07:58:14 - 00:08:27:01
Unknown
This breakthrough exponentially improves those bounds, meaning researchers found a way to mathematically prove that you can fit far more distinct error proof signals into these systems than previously thought. Could you next ask it is there any application for this right in industry or science, or just plain light real life? Yeah. Okay. This this podcast has become let's talk to Gemini.

00:08:27:04 - 00:08:57:04
Unknown
But that's okay because sometimes that's what this is good for, right? Like like I see something like this and I'm like, even if I ask like an AI model to explain this to me, it's going to explain it to me as if I understand the entire scope of it. So I'm like, okay, dumb this stuff down. So this says, they directly power technology's used every day 5G, Wi-Fi satellite when a phone sends data through the air, environmental noise, magnetic fields, physical walls try to scramble the signal.

00:08:57:05 - 00:09:26:01
Unknown
So this could be applied to that. Audio and video streaming. Yeah. Okay. Spherical codes represent a signal points on a continuous space. Modems convert digital photos or video streams into geometric arrangements of continuous waves. Finding better bounds for spherical codes allows engineers to pack more data into radio frequencies without your stream buffering or dropping quality. That sounds you.

00:09:26:04 - 00:09:50:00
Unknown
That's a winner. That's a winner. Winner. That sounds pretty applicable to deep space communications. Hard drives. Okay, so yes, this stuff matters. And that yeah, I didn't mean to be philistine here and act as if it doesn't. I just didn't understand. No, totally. And nor nor have I. And I've often wondered that, like these unsolvable mathematical things, are they a big deal because they're unsolvable?

00:09:50:00 - 00:10:15:12
Unknown
Or are they a big deal because solving them would actually open up some sort of breakthrough that actually makes a difference in something in life, you know, versus just being a really curious puzzle to solve, you know, or a curiosity. You know what I mean? So then Gary Marcus growling as ever said that open AI is amazing but vastly oversold.

00:10:15:12 - 00:10:42:04
Unknown
New model Astra okay. And he said that some number of these I think were solvable by what's the what's the anthropic one. Oh mythos. No. The one down from that. Oh fable. Fable five. Yeah. So there's there's poking at this.

00:10:42:06 - 00:11:08:03
Unknown
So fine. So, so we're at a point to where these mathematical, dilemmas or whatever you want to call them, these problems that have never been actually solved by people who are far smarter than anyone else in the planet to solve these sorts of things. Anyways, mathematicians, whatever. We're at the point to where multiple models are doing this relatively easily, so.

00:11:08:06 - 00:11:46:08
Unknown
Right. And it could go through multiple disciplines like this. Yeah. And my view is that this is the better path to take as a specialized path, rather than constantly seeking the generalized artificial general intelligence. Yeah, okay. Go after a bunch of specific things. And this is this is Yann LeCun argument that you will reach those goals faster if you don't waste time trying for the general, that if you put the resources, the same resources behind specific tasks math, science, medicine, whatever, you'll you'll be more efficient, simply put.

00:11:46:11 - 00:12:09:24
Unknown
So again, I think it's important to say Gary Marcus said it is amazing he's giving it his due. But then he said that it suffers from the fallacy of composition, and he puts up a Wikipedia description that arises when one infers that something is true of the whole, from the fact that it is true of some part of the whole.

00:12:09:26 - 00:12:30:00
Unknown
A trivial example might be this tire is made of rubber. Therefore, the vehicle of which it is a part is also made of rubber. So, he says, thinking of a system that is great at a certain kind of math problem is great at all math, great at science, or even great, quite possibly great at everything the AGI is near.

00:12:30:01 - 00:12:51:22
Unknown
Community keeps committing the same logical fallacy over and over. Every time there's an advance, I see the same error. Someone pretends that all cognition is created equally. Totally untrue, he says. Whenever AI achieved success at some form of fancy cognition, they want you to believe that success on all forms of AI is imminent. And I agree with them.

00:12:51:25 - 00:13:24:06
Unknown
Yeah, totally. You have to. You don't have to be accounted of a psychologist to realize that this inference just doesn't follow. So it's the AI boys. They're so weird because they they cut down their own successes by setting this bar of general intelligence, by then claiming doom related powers. They just don't take the win. Wow, this thing did something amazing in math and even mathematicians haven't done in one fell swoop.

00:13:24:06 - 00:13:46:13
Unknown
Cool. Take the win. Stop there. Shut up there. But they can't do it. They've got dead. What they want to do is use it as a as one. Brick in the wall of. This is the most amazing thing in all facets or whatever. As Gary Marcus puts it here, expertise in one domain does not at all guarantee expertise in all or even most domains.

00:13:46:14 - 00:14:19:26
Unknown
Right? Yeah. And that's and that's why on one hand, it's amazing to see a model like you said, like you just said, it's amazing to see a model that can do these things that were, you know, that were unproven by just humans who were incredibly smart doing them. That's an accomplishment. But that doesn't automatically translate to and therefore AGI is right around the corner because nothing that we've seen so far has shown because AGI is that but applied to all facets of everything.

00:14:19:29 - 00:14:41:29
Unknown
And we, you know, through that lens, man, we feel like I feel like that is, you know, that kind of goes into what you have stressed time and time again from your perspective. And I'm still like, jury's still out. On whether AGI eventually happens at one, one time somewhere down the line. That just makes it feel like it's a very far ways down the line, because we're nowhere near that.

00:14:42:00 - 00:15:01:19
Unknown
We're you're good at this one thing, and that's impressive. But doing all things to that, I mean, I work with AI every day to do things. And let me tell you, it is not it is not perfect. No. And I wouldn't expect it to be capable. Yeah, exactly. And there's I'm going to go a little too far here.

00:15:01:22 - 00:15:25:25
Unknown
But I think this also ties into a kind of Nietzsche and Übermensch eugenicist view. Here. We can make human strength. We can make human intelligence. Is this effort to say, look how powerful we are. Yeah. And it's so smart we can exceed humanity. We can make the better human. It's the wrong goal. It's the wrong goal. It's. It's a tool.

00:15:25:28 - 00:15:52:14
Unknown
I'm editing manuscripts right now for my new book series, intelligence, AI and humanity, and I'm reading a manuscript by a Swedish author. I'll talk about more when we buy the book, but it's really interesting because he talks about these. Learning used to be about friction and about.

00:15:52:17 - 00:16:21:02
Unknown
The return of the tool in the hand, and he uses the example of his grandfather cutting down trees and knowing the sense of the ax, and when it was dull, or when his was just whatever. Right. And, and his grandfather learned from that interaction. And, and this author comes around and says, very interestingly, when it comes up with, with, with answers before we ask the question, when it takes all the friction out of things, we lose sight does really well.

00:16:21:04 - 00:16:44:08
Unknown
Yeah, it does exactly. And we lose something along the way. We lose learning. We we we lose discovery. We agree. We lose other things. So, you know, I think if you if you give yourself over entirely to that the ease of that discovery. Yeah. Whereas if I get back to the point here, if I were a mathematician, mathematician, and I'm sure as hell not, I never took calculus.

00:16:44:11 - 00:16:49:00
Unknown
Neither did I, really.

00:16:49:02 - 00:17:10:00
Unknown
Why do you think I went into journalism? All right. I would hope that I could look at this and say, oh my God, what all can I accomplish now? What are the front? What are the new problems I can come up with? What are the new solutions that I can find? I think it should be a very exciting time for a mathematician, but they're freaked.

00:17:10:02 - 00:17:29:25
Unknown
And are they freaked because they're because of a, like, a self-preservation thing, like. Oh, but, you know, kind of like they're taking, like, a different spin on they're taking our job sort of thing. Like. Yeah, I think it's that's my job to figure that out. Now you just made it super simple. It's I think I would class it more in this case is identity.

00:17:29:28 - 00:17:55:26
Unknown
Yeah. What classes a true mathematician is they can understand this stuff that others can't. Yep. And they can puzzle on something for years. And they can make one discovery in life that sets them apart because they are that damn smart and they're smarter than the rest of us. And that's what defined them in our world. And God bless them, because they are damn smart and they can do things that others can't even understand.

00:17:56:01 - 00:18:35:14
Unknown
It's like, you know, quantum physicists. I can't understand what they do at all. And I think that's that's part of it. They worked in an aura that was beyond the comprehension of most people. And here comes this machine that just solves this stuff. Yeah, this is interesting as I think about this, because when you think about other industries, when you think about creative industries, authorship and writing or, you know, writing articles or written word making, music, art and stuff, there's a subjective quality to what someone creates and what AI creates.

00:18:35:19 - 00:18:54:17
Unknown
You know, there's a lot of people that would look at what what an AI creates as a piece of art and would be like, no, I know that that's AI, that's not art, that's that's slop or whatever. There's a subjective quality to what's good versus bad in mathematics. It's either right or wrong, right? And if the AI gets it right, it is getting it right.

00:18:54:17 - 00:19:21:05
Unknown
It's not just getting it subjectively right. And so that, like existential piece for a mathematician is probably a little stronger for a mathematician because it's like, no, literally, this is what I do and there is an end. There is a solution. One solution to this problem that I would normally be the person to figure out, or would want to be the person to figure out this thing, just figured it out in a matter of weeks or whatever.

00:19:21:07 - 00:19:46:02
Unknown
Yeah, but I think every field is going to have to look at this and say, good, we can accomplish more with but with this as a tool, not as a replacement. Right? Right. And who knows how many other suggestions that came up with for other proofs that it got wrong? Yeah, well that's true. You don't. Yeah. Open AI is they didn't show the work here right.

00:19:46:05 - 00:20:22:08
Unknown
Yeah. They're they're not releasing that that's for sure. Yeah. Interesting. Well, that's, that's fascinating. Yeah it is, it is. And you know, the open AI model is like we said, it's not the only one that's doing these mathematical, you know, discoveries. So I think we just we continue hearing more and more about this. And I will be curious to see, like, how do mathematicians then, how how does their role in all of this kind of transition and transform, do they use these as tools or do they stick with their I think we have no choice but to.

00:20:22:11 - 00:20:43:17
Unknown
Yeah, but the other question becomes, from an educational perspective, how do you teach mathematicians now? Right. The challenge that you have out there, this, this, this move that you try to reach suddenly got a lot closer. And the process of getting there, the process of thinking this through.

00:20:43:19 - 00:21:02:13
Unknown
Becomes unnecessary in some cases. So how do you. But we kind of went through this already with a calculator, right. Yeah. Like, well at a much simpler level. Yes. At a much simpler level. But I mean, at the time before a calculator existed, that seemed a lot like cheating. I have to imagine, like, oh, shoot, like, what does the mathematician even do anymore?

00:21:02:14 - 00:21:20:28
Unknown
They don't have to do the hard work. They just punch in a bunch of numbers on a little digital D-pad, you know, a pad, and it gives them or, you know, so I'm old enough young Jason. I'm old enough that I used to have to figure out how to learn to do a slide rule. Did you have you ever dealt with a slide rule?

00:21:21:00 - 00:21:49:05
Unknown
Let me. What is a slide rule? Yeah, exactly. Oh yeah. Okay, I do remember I remember seeing yeah maybe I use them but no I did not. Not much of my education was ever built around a slide rule. No. Yeah. Not not that I'm aware of. So that was replaced by by HP calculators and okay. Logarithms and square roots and all this stuff.

00:21:49:07 - 00:22:14:17
Unknown
Yeah. And you got the answer. Yeah. I never understood how the slide rule worked. I understood that it got you the answer right. But I you know, I couldn't figure out what was behind the the math that made it work. So using it didn't teach me anything really. Yeah, right. Yeah. Interesting. Yeah. Well, there we go. There. What else is there?

00:22:14:17 - 00:22:34:19
Unknown
OpenAI hugging face. I mean, there's a little bit about hugging a little bit more. It went in some other places and some other stuff. And then anthropic got jealous and said, well, ours got out too. That's right. I don't even know that I put that in here. Oh, you did, you did though. Yeah. And also also saying, hey, we're smart to see we're part of the party.

00:22:34:20 - 00:23:01:11
Unknown
We're too smart for our own good also, right? Our agents are powerful. It's cool. Yeah. Okay. Well, there's that, let's see here. What else do we have? Oh, the EU, the EU doing the AI act thing, which I guess I was just showing, which, you know, I guess this is shifting gears a little bit. I think I had it in the same block, but I don't remember why.

00:23:01:12 - 00:23:29:17
Unknown
But EU act, AI act becoming fully enforceable on a couple of days ago, apparently so. The European Commission now has direct enforcement powers over general purpose AI providers. So you have this happening in the EU. We also have this kind of this storm brewing in the US around this. So this is largely becoming de facto the the direction that we're moving in.

00:23:29:17 - 00:24:02:24
Unknown
Anthropic OpenAI, Google named as targets for its oversight for its expanded oversight anyways, specifically in this case. And, yeah, regardless of where they're headquartered, by the way, just because it's the EU doesn't mean that it only applies to, you know, European companies or whatever. Any of the companies, including anthropic, OpenAI and Google can be fined up to €15 million, 3% of global annual annual turnover, whichever happens to be larger, apparently.

00:24:02:26 - 00:24:23:16
Unknown
And so a lot of this is about labeling AI, and I predict this is going to end up like the damned cookie click. We have to make whatever we go to the inside boy. Yes, everything's going to be labeled AI. We got to acknowledge this label that AI was somehow involved because everybody's going to cover their rump by, say, did anybody hear you spell check?

00:24:23:17 - 00:24:47:23
Unknown
Okay. We're all we're going to say everything that was help with AI. That's the crazy thing is like when you really go back to the the root of it, like, yeah, even spell check would be classified or qualified. Maybe, maybe they put in a, you know, some sort of a specification to say, right. But I'm thinking the lawyers involved, the lawyers are going to say if there's a fine involved and liability involved, cover yourself, play it safe, cover yourself, play.

00:24:47:25 - 00:25:20:22
Unknown
Say it's all AI. Yeah. And it's going to be. If everything's AI, then nothing's AI. Exactly, exactly right. The EU rules also require, you know, speaking of labeling, labeling AI generated content that appears to be realistic. So if you, you know, if you happen to be looking at an image or whatever, and it's representing reality in some way, shape or form, providers would have to embed machine able, machine readable markers for that.

00:25:20:28 - 00:25:44:01
Unknown
And, you know, visible disclosures, all that kind of stuff. And of course, a bunch of fines and, and everything attached to that. So watermarking essentially for your music generated or your images or whatever, you know. Never mind that those things can be stripped out pretty easily, but I, I don't know if, you know, they're saying you got to make that harder.

00:25:44:01 - 00:26:13:29
Unknown
And if they are like, how how do you make that harder? I don't know. But so there you go. Now from Europe we go to China, now to China. The China story with AI just continues to rumble along. Last week, of course, we were talking a lot about Kimmy K3, the whole open weights versus closed weight thing. And now we have Alibaba releasing Quinn 3.8 max.

00:26:14:00 - 00:26:41:06
Unknown
This is their largest model ever apparently. You know, a bunch of numbers that I could read, but a lot of people, the eyes would laze over, but they're saying that it's competitive performance against the top US models in coding multimodality engineering benchmarks. And not only is it very, very capable, not only is it open, but well, I guess that's the thing.

00:26:41:06 - 00:27:00:20
Unknown
They're open sourcing it essentially. So it's very capable. The weights are going to be publicly released this week if they aren't already. I didn't check the news this morning because I've had a crazy morning, so maybe they've already done that. But basically it's their first max class model to be open sourced. And yeah, we're seeing a bit of a pattern here, aren't we?

00:27:00:23 - 00:27:24:11
Unknown
Yeah we are. And at the same time, deep seek this models and lowered its prices immensely. And I think there's a there's just a race to the bottom. There's a price war going on here. Yeah. Which is interesting. And in response OpenAI also lowered some of its prices. So I think we see a race toward the bottom here.

00:27:24:17 - 00:27:52:20
Unknown
On pricing. I watched a video on TikTok. I didn't put it in the rundown from a Marxist analyst explaining the Chinese move in AI from a Marxist perspective. They see and it goes into stuff that I don't get. Well, I'd never fully rocked all of Marxism. I was going to say I'm very clueless on on this aspect of of life and things like.

00:27:52:23 - 00:28:26:01
Unknown
It's one of those terms that I hear and sometimes I situationally understand what it is. But so the part that throws me is, is, is excess creating excess value in labor versus machines and how it's captured and who benefits from it, and capitalism versus Marxism and so on and so forth. But the point of making AI as inexpensive as possible is, in this analysis, moving toward a next phase of capitalism, that it's not it's not as simple as we're going to get.

00:28:26:01 - 00:29:04:22
Unknown
Those Americans grew those Yanks. It's part of a larger philosophical basis. And I didn't think that I was going to find a way to shoehorn this story in. But now I'm going to. So Alex Karp, we've mentioned in the last few weeks, has a philosophy to the bottom of the rundown, a philosophy of telling companies, we've talked about this before, that they're they're fools for using hosted frontier models that they're handing over, not just money, but also their alpha, their corporate perspective, their data to these companies.

00:29:04:22 - 00:29:30:10
Unknown
And it's a mistake. So we wrote a letter to the shareholders. Phenomenal results, doubled their profit a 93% growth rate over the over a year ago in revenue. That is just just huge results we generated. So it was it was let me see here one point. We generated 1.9 billion in revenue for the second quarter of the year.

00:29:30:12 - 00:30:03:17
Unknown
Yeah, another record in our company's two decades of growth, 93% growth over the prior year. We generated 1.1 billion in profit in the second quarter of the year. And we have now generated more profit in a single quarter than we did in total revenue in the same period the year before. So it's going great gangbusters. And what he's going on from his braggadocio is to then going on again and beating big companies over the head.

00:30:03:19 - 00:30:23:17
Unknown
Our results represent a subset of the economic value that our software has created for customers. We do not get paid for clicks or tokens or chats. The gamification of the most significant development in modern economic history seems to us misplaced. The usage of a platform may hit at its value, but is by no means dispositive. Now this is what gets me.

00:30:23:17 - 00:30:54:04
Unknown
There are Marxist overtones and undertones to our business. He's the ultimate capitalist. Alice Karp, the head of of Palantir, and he is talking about Marxism. And last week we talked about how Dean ball at OpenAI was decrying open white models as as AI communism. And Karp was coming along and saying, but I wrote a post last week, say, based on what Jason and I had discussed the week before, saying, yeah, I mean, this doesn't sound so bad.

00:30:54:05 - 00:31:24:10
Unknown
So there's a phenomenal little, little sideshow here coming along. Just as we have a democratic socialist in the US, we have China playing inscrutable economic games with AI. Time to read Marx. Yeah, apparently so. Time to get more educated about it. Yeah, he had said in his I don't. Is it this post that he's he says that OpenAI and anthropic are trying to drug addict us to a future that they believe they want to control.

00:31:24:12 - 00:31:43:23
Unknown
Yeah. You know, there's there's some truth to that. I mean, I've definitely thought of that as I've used more of these, these platforms just like, you know, like I, I hate to admit this, but the other day I ended up giving anthropic $200 instead of $100 because I ran into a wall and it was a wall that I was willing to to give more to.

00:31:43:24 - 00:32:03:10
Unknown
And then I did it, and I got the thing on the other end that I was working for or working for. It was working for me for, and then I it was like, okay, was that worth it? What did he mean? No. What do you think is my usage just got, like, going up now. Like, the more I use these things, the more reliant you become on them, right?

00:32:03:11 - 00:32:26:24
Unknown
And it is like a drug addict, like. So I predict the next thing that's going to happen to you is what we watch Leo Laporte doing is you're going to get a more powerful computer and try to run local models. I mean, I'm seriously considering that, although that sounds like from a technical perspective, I'm like, gosh. Am I opening up a can of worms that I can actually, you know, that's my fear.

00:32:27:00 - 00:32:47:02
Unknown
But yeah. Yeah. But I think you could go down the street and have him teach you. Yeah. That's true. That's probably enjoy that in the same town. Yeah. So. So he just said, we talked about yesterday when we recorded an earlier one earlier episode there. He just bought a DJ spark o d. Yeah I just said that.

00:32:47:05 - 00:33:15:06
Unknown
Now see his news here first. Sorry, Leo. But because he believes strongly in local models. Local because of control, because of a lack of of vulnerability and cost and cost. Yeah. Just because this local, does that mean I mean, if it's truly local, if it's truly not online, then you're not paying anything. But it's not entirely not online because your agents have to, you know what I mean?

00:33:15:06 - 00:33:33:24
Unknown
Like, yeah, there's some cost associated. It's just not as much cost. I think you can have have the saddle go to the most powerful models that are too big for you to run locally, only for the tasks that they can't. Right. And so that's the art of it, I think. And that's what Palantir is doing for its clients.

00:33:33:28 - 00:33:55:14
Unknown
To an extent. It's saying, yeah, we'll use these models. They're using Mondo models. They're using the biggest things they can get their own weight. But part of the efficiency there is to choose which way to go, right? Know which is what the hosted models do as well. Interesting. Keep learning so much. Yeah, I have I have considered that.

00:33:55:14 - 00:34:20:06
Unknown
I know that Leo does. Hermes. Hermes. Hermes. Yes. Agent. That's one. Yes, I think and I think he, you know, the the open brain thing, which I have, I still have to, you know, I love it. They always say, you know Nate natives has said this too. It's just, you know, 45 minutes on a Saturday morning with a cup of coffee and you've got your open round, like believe you.

00:34:20:06 - 00:34:46:02
Unknown
Yeah. I do not believe you. You have you have no you you underestimate my ability to make things way harder than they actually are. But I keep meaning to to explore those things. I keep thinking there's a great business to be had in WordPress, like hosting. You don't want to use anthropic and open AI. You want to use a local model because it'll be cheaper, but you can't figure out how to do all that.

00:34:46:02 - 00:35:08:20
Unknown
And so in a sense, it's a baby step away from the host of models to a different kind of host. Yeah, right. We'll we'll configure this for you. I think I've said this on the show before, way back in the day, we had our journalism students learn how to download WordPress. Yeah. Install it on a server. And now you don't need to do that.

00:35:08:22 - 00:35:30:19
Unknown
No. And so I think there's a business to be had there in making it easier for folks like us. Yeah. So we can use quid. We can use it as other models. We can choose among saddles, I think. I think it's an exciting business to be had. Yeah, indeed. Well, see, we got lots to talk about. We got a lot more to talk about.

00:35:30:19 - 00:35:50:06
Unknown
I do want to take a break, though, and I want to show you. Let's see here. Do do do do do. Let's go to here. This is what I meant to go to. Okay. We're going to take a little Patreon break. Don't tune out because I just want to announce something here. We just passed 100 paid patrons on the Patreon.

00:35:50:07 - 00:36:10:18
Unknown
Thank you. Which, you know, I am. I'm the kind of person that likes a little milestone moment. So, you know, when I see suddenly we're in three digits, like, I see that as a as a chance to celebrate and be like, hey, we must be doing something right, because that number is continuing to go up. And so I thought, like, what can I do to like give back a little bit?

00:36:10:18 - 00:36:38:01
Unknown
And one thing that I've really noticed is we're getting a lot of free subscribers on Patreon. I'm like, what can I do to convince free to convert into paid? Because the reality is that this is a show that really hinges on your support. I fully understand that our podcast that you download like that, you get that publicly accessed podcast has a ton of ads in it, and I feel bad about that.

00:36:38:02 - 00:36:53:13
Unknown
At the same time, this is the only way that we can do this show, right? Like we got to we got to make some money to pay the bills and to pay for what's needed on it. And I'm constantly trying to figure out how I craft a living out of doing what I'm doing and everything. So it's just the reality of it.

00:36:53:13 - 00:37:11:17
Unknown
The best way for you to get out of that ad world is to become a supporter on Patreon. Now, what I've realized is because Patreon gives me the tools to do this, I can run kind of a short term campaign and give you give you a little bit of incentive. So what I've got going right now is a coupon 100 patrons.

00:37:11:17 - 00:37:31:01
Unknown
It's 100 patrons. You get 30% off for the year if you support. Now, what does this actually come out to? If you're thinking about doing like the $5 tier, that's the tier that gets you the AI Inside Daily podcast. You would normally spend on a, on a monthly basis, you'd spend $60 over the course of the year.

00:37:31:05 - 00:37:51:22
Unknown
This wouldn't just give you 30% off of that right now. Normally you would get 20% off if you paid for the year. That's just standard for my for my page. For our Patreon, I'm giving you an additional 30% off of that. So instead of paying $60 per year, if you paid monthly, you pay 3360 and you get the entire year with that discount.

00:37:51:23 - 00:38:18:26
Unknown
Same thing goes for the $7 tier. That's the tier with the ad free episodes included of this particular podcast. So, you know, instead of 84 you'd pay 4700 for. So basically what I'm saying bargain. What I'm saying is if if you like the show and you want to see it stick around and you want to support financially, get rid of ads and you want to deal, you know, to to get in basically half off is what it really comes out to.

00:38:18:29 - 00:38:36:01
Unknown
For the next two weeks, I'm going to be running the coupon. So 100 patrons, that's 100 patrons that gets you the deal if you're free. If you haven't signed up yet, you can support for the entire year and you don't pay that much. And we just really appreciate it. So consider that I'm also holding and you're welcome to join.

00:38:36:01 - 00:38:55:23
Unknown
Jeff, I didn't ask you about this, but I'm holding an Amy and ask me anything at the end of the month is kind of like a thank you for anyone that wants to do this. And so for all patrons. So anyways, that's. Yeah, it's my pleasure. This is what I'm about to say is easy for me to say because I'm retired and the money matters to Jason.

00:38:55:23 - 00:39:13:25
Unknown
It matters to be able to put this out. But it also matters that you show support. It matters a lot that you like the show enough to say that you care about it, that that means the world. And so we're very, very grateful for that as well. Incredibly grateful. My thought when I was putting this together was like, oh, we just passed 100.

00:39:13:25 - 00:39:33:29
Unknown
Can we get to 120? So sure, I'll set it. I'll set a target. Like if we can get to 120 paid patrons, I think the deal is good enough that like, it allows you to support and not like go go bankrupt to the process. But anything does help. So anyways, that's that's that's the plug. I don't need to wax on about this anymore.

00:39:34:02 - 00:39:53:06
Unknown
Patreon.com Inside Show coupon is 100 patrons for 100 patrons. And you know, we got the next two weeks, so you'll probably hear me talk about this on next week's show. And thank you to all the 100. Thank you, thank you, thank you 100%. Thank you to everyone who has supported us for now, two and a half years, almost.

00:39:53:07 - 00:40:12:03
Unknown
Wow. We're coming up on two and a half years to this podcast. It's kind of cool. Yeah, that's pretty awesome. So thank you patrons. All right. Let's take a break and then we will get to more news on the other side, including Google's Android OS for humanoid robots. That's how I like to describe this story coming up.

00:40:12:05 - 00:40:34:22
Unknown
All right, so I said Google's Android OS for humanoid robots and that that's how I like to describe it. That's not actually how I chose to describe it. That's how Demis, Google's DeepMind CEO, described it in something that I read last week when I was reading about this. But Google DeepMind released Gemini Robotics, too. It's a vision language action model.

00:40:34:22 - 00:41:06:00
Unknown
So in other words, it's it's kind of like an operating system driven by their AI models for giving robots full body control. And they've demonstrated this on some some robot prototypes or releases after Apollo two, humanoid, which has five fingered hands. So it can do all sorts of things like sorting and picking up and, you know, of course, walking and crouching and stretching and doing all sorts of things.

00:41:06:06 - 00:41:31:16
Unknown
Where is I know I had a video somewhere, maybe I didn't link to the video. There's a video I put a link in. Let's see here. Oh, Gemini Robotics two. Here we go, here we go. I'm not going to play any audio, but you can kind of see what the what the actual robots can do. Those cute, cute humanoid robots picking up a watering canister very slowly and water.

00:41:31:21 - 00:41:56:00
Unknown
The other demo has happened in there is that they instruct the robot. They use the the dog robot and say there's a video down there, I think, and and it says, go get the popcorn from the snack area. So if you if you scroll down on the Introducing Gemini Robotics to the Google link, not this one. Oh, on the Google link, the Google, which is.

00:41:56:04 - 00:42:14:00
Unknown
Oh I see. Scroll down. No scroll down. Yeah. They're scrolling right here. So two guys sit there and they instruct the dog robot to go to the to the snack area. And they just put out some snacks on a table and all it is and get the popcorn. So it has to go. It has to know where to go as to find it.

00:42:14:00 - 00:42:33:03
Unknown
It has to get there. It has to read the word popcorn. It has to be able to grab the bag and bring it back. Oh, look at that thing. That's a Boston Dynamics robot, is it? Thank you. Yes. The big dog or whatever. So now it's facing for those of you on video, it's facing the, you know, a dozen snacks.

00:42:33:05 - 00:42:53:13
Unknown
The popcorn is more prominent than anything else but fine detected target. Now it's reaching out and it's it's picking it up with its claw hand, which is both cute and scary at the same time. It kind of looks like a pincer, right? It totally looks like a pincer. It's like it hasn't been mouth. Yeah, and so it goes.

00:42:53:16 - 00:43:16:12
Unknown
They had to speed it up here to be more interesting. It goes back and then he has to kind of drop dog drop. Let go of the popcorn. You crushed a couple of kernels. Bad robot bad. But there it is. So it's very cool. So it operated autonomously we believe. And that's neat as far as we know. Yeah.

00:43:16:13 - 00:43:34:04
Unknown
Yeah super cool. And this obviously is where world models become. There you go. Incredibly important. Is it not about language alone. Yes. You can recognize popcorn, but everything else is about understanding how the world operates. And if it had, if it had to go over.

00:43:34:06 - 00:44:04:28
Unknown
Obstacles or found logic and how things operate, then world models matter. So it's got to be cool to work in a lab like that and work with robots like, yeah, you know, or like designing robots and getting them to do these things that really just it feels so science fiction. I love in my local mall where I was taking my senior mall walks when I was reading, there was a large space because so stores are closing.

00:44:04:29 - 00:44:33:15
Unknown
They gave over space to local high schools for their robotics teams, and I think it's a really cool way for young students to get involved with technology to understand it's applicable. I think it's I think it's very, very cool. Yeah. I mean, one of the schools that we toured for, for one of my daughters not too long ago, they had a robotics, you know, they had a kind of a lab that and they primarily used it for robotics and stuff and more and more.

00:44:33:16 - 00:44:59:01
Unknown
This is just yeah, this is going to become something that that isn't just like kind of that obscure thing. It becomes more normalized because the technology is getting there. The, the, the, the, the, I want to say intellectual capability, but the processing capability of the AI models inside of these embodiments like just really, really cool. And everybody thinks, oh, the dog is creepy.

00:44:59:01 - 00:45:20:29
Unknown
Well, not if it brings you popcorn, damn it. Yeah, yeah. And maybe, who knows, maybe there's a speaker on it when it comes up to you, it goes. Yeah, yeah. Is that. Is that cuter or I don't know, that might just be creepy in of itself, but but I really liked how Demis kind of compared this or likened it to an android, you know, like what Android OS was for smartphones.

00:45:21:00 - 00:45:40:13
Unknown
He sees this as an OS layer for robotics. And let's not forget that Nvidia has a huge amount of technology around this and operating systems just for robotics. So this is a new area of competition as well. Yeah. And then Google has their what was the what was the AI thing that they've got. I played around with it at Google.

00:45:40:13 - 00:46:13:26
Unknown
I o that like creates world models and like live in real time. Yeah that one I never remember Google brands. Yeah there's so many of them. But you know that you know potentially using some of that information for training I imagine systems like this, you know, on different environments and stuff. So I mean the video showed the, the, the humanoid had a zip lock and was dexterous enough to, to zip the zip with screw unscrewed a light bulb, unscrewed a light bulb.

00:46:13:27 - 00:46:26:17
Unknown
Yeah, totally. That's pretty amazing. Yeah, it's pretty impressive. It is. It is. Also I should add, over the weekend I watched Terminator two, so it's also great.

00:46:26:19 - 00:46:49:04
Unknown
But it's such a great movie. Oh my God, that movie is so good. It's just such a great movie. Did you watch with their kids? No. So, okay, completely unrelated to everything, I we have this projector because I'm starting to kind of develop AI workshops, and I've got a couple scheduled for later this month in the, in the Bay area.

00:46:49:06 - 00:47:07:15
Unknown
And so I figured, like actually it would be really good for me to have a projector so that I can, you know, take a screen and project it large so that everybody can see and everything. So I use that as an excuse to get a nice projector. And then over the weekend the house is empty. And I was like, oh, I've got that projector.

00:47:07:16 - 00:47:25:03
Unknown
I'm going to hook that up in my living room and black it all out and, you know, end up with a massive screen. I made it like a movie movie night where I watched Terminator and I watched tenet, and it was. It looks so good. Oh, it was so much fun, I love that. Did you have to take pictures down off the wall?

00:47:25:05 - 00:47:44:06
Unknown
No, because the projector came with a floor standing screen that you could put up. Oh, my. And so it was like, you know, it was like 100 inch diagonal screen or whatever. That's cool. So it was large. So I just kind of put it up. And then when I was done, I took it down. Yeah, yeah. Did you make Jason.

00:47:44:06 - 00:48:07:24
Unknown
Did you make popcorn? You know, I meant to, but I ate ice cream instead. Oh, okay. I'm not disappointed in you, but. Okay. You need a robot to bring you the popcorn. Yes, exactly, exactly. Apple having a bit of a rough time. This is a story that keeps coming up around bug bounty programs. So it's not isolated to just just Apple.

00:48:07:24 - 00:48:35:27
Unknown
But Apple's bug bounty program is being overwhelmed by AI generated vulnerability reports Llms, as we know. Make it really easy. Or it seems. Anyways, I've never done it, so I don't know how easy it is, but seems that they make it really easy to discover what appear to be bugs and to, you know, file bug reports. And a lot of these companies offer money for for discoveries that are impactful.

00:48:35:27 - 00:49:00:19
Unknown
So people probably using these models just see it as an opportunity to, you know, inundate them with these in the hopes they make something. They're getting pinged with slop ones, ones that aren't legit. Yes. So that's causing an expense for them to have to go through it. They're limiting the number of bug bounty claims that you can file, which the one hand makes sense and the other hand probably cuts off some because people will stop doing it.

00:49:00:21 - 00:49:34:04
Unknown
But what occurred to me when I read this JSON is at some point, any responsible company is going to put its own software through this. These same. Oh yeah, that's a good point. And I think it cuts off theoretically. It cuts off our way, cuts down towards zero, the bug bounty market. Because if everybody's using the same powerful tool to find it, and the company is using a tool that's at least that powerful to find them, and they eliminate most of them, at that point, the software is more secure and there's fewer bugs to be found, and there's fewer bounties to pay.

00:49:34:07 - 00:49:47:09
Unknown
That's a good point. I mean, I imagine they're already doing that to some degree. There should be. Yeah. So there was this should be they should be going back over their entire.

00:49:47:11 - 00:50:08:10
Unknown
Code base or whatever. Yeah. What they have. Yep. Yeah. Yeah I'm sure that they are doing that to some degree. And and because like why why pay that random person that ran an LLM on your stuff to discover the thing? Why not just have your own person internally do that. Yeah. But I mean there's a lot of money to be made.

00:50:08:10 - 00:50:34:07
Unknown
Their apples paid more than $35 million to since the program launched 800 researchers, you know, in there. So that cuts down a lot because that's a lot of people for that amount of money. But the maximum bounty that you can get is $2 million, could exceed $5 million in bonuses. So there's some real money there. If you if you are able to tackle or discover a pretty major vulnerability for Apple.

00:50:34:07 - 00:51:02:17
Unknown
And so there's a lot of incentive there. So they get overwhelmed by the burden of these reports. Do you remember the friend pendant? Yep. Like I saw it and I was like, God, I vaguely remember it. It's still around. It's apparently it's still around. So this was the the AI pendant that you wore around your neck, and it was like, you know, like a little, like, circular thing that was, you know, to be your friend.

00:51:02:17 - 00:51:25:24
Unknown
It would follow you around and it would text message you so you wouldn't hear it or talk with it. It would pick up the environment, and then it would send you text messages on your phone as if it was your friend, which is, I guess, why they called it friend. Apparently they paid $1.8 million for the URL when they had $2.5 million invested in the company to get friend.

00:51:25:26 - 00:51:50:03
Unknown
Com so not a not a very smart way to spend your money there at the top, but I guess they're still around. They now have friend 2.0 coming out. And this time I guess it does actually talk back to you. So if what you wanted was an AI friend that you wear all the time, and it doesn't just send you text messages but actually talks to you, you got to pay $249 because the price is doubled for that.

00:51:50:05 - 00:52:09:01
Unknown
I just don't know about these devices. No, I don't either. I watched a video. I can't find it now. Watch the video that they were putting out. And it was just depressing as hell. There's a there's a really dark Velten shown to this company. People are all depressed and they don't have anybody else to talk to, and they talk to this dependent.

00:52:09:04 - 00:52:38:09
Unknown
And yeah, the pictures are all mournful. They are. No one's happy. Like, no, everybody is just depressed that their only friend is there is a lanyard on their neck. Yeah, yeah, I guess I'd be depressed about that too. It's hard to imagine what would make me buy this. Yeah. Like, I mean, I can sort of. Oh.

00:52:38:10 - 00:52:55:27
Unknown
Oh, by the way, the pebble. The pebble ring. Do you remember Pebble? They have their ring. I got a shipping order. I'm going to be getting it any day. You're kidding me. And it's. And it's the thing that records, I mean, very short audio snippets, right? You know, it also doesn't have a battery that you charge. It only lasts a certain amount of hours.

00:52:55:27 - 00:53:20:25
Unknown
And then it's. How long ago did you order that? God. Probably like eight months ago. Nine months ago or something. But I suddenly got a delivery order on it, so I'm gonna get it. So I'm. But but see, something like that I can understand because, like, when I'm out and about and I'm moving around like, yes, I can do it with my watch, but I get these very fleeting thoughts of like, oh, I need to remember to do this, you know, remember to blah, blah, blah, or maybe do it in my watch or whatever, but something like this, like, yeah, it does that.

00:53:20:27 - 00:53:41:18
Unknown
Or it might summarize a meeting or whatever, which so many of those things do now, but it's, but it's meant to be more like, yeah, like it like a companion or something. And I just that just I get it for some people, some people, it might actually feel a need. And I don't want to be grudge any one of that, but it just I don't get it for me.

00:53:41:21 - 00:53:51:16
Unknown
But do I trust this guy to fill that need? Well, yeah, that's a good point. Yeah. Yeah, I don't know.

00:53:51:18 - 00:54:18:29
Unknown
I don't know, 30 day retention, a $10 a month subscription. Oh okay. So 30 day retention. If you don't pay for the monthly subscription, $10 gets you more AI memory with the pendant. Still only iPhone. So that's another reason why we don't care. Because we're right. But yeah, 15 Android people are happier. They don't need this. Yes, we're just happier in general.

00:54:19:01 - 00:54:47:11
Unknown
Yeah, we don't need that thing around our next, LinkedIn slot button. Don't don't apparently come into LinkedIn. I mean, LinkedIn does have a lot of what a lot of AI written slop. If it does, Facebook's worse, I think. Yeah, that's true. I've seen a lot of it on Facebook to the flavor on LinkedIn is is very much like very influencer.

00:54:47:13 - 00:55:14:21
Unknown
Very. Yeah. Just very like I know all the secrets and here here's what they are. But you can often detect that it was probably AI that was writing it. And so apparently LinkedIn is bringing a new AI slop button of some sort. So that users can flag a post that seems like AI slop. Now, what's interesting here is that LinkedIn is actually using AI slop as a terminology.

00:55:14:21 - 00:55:41:11
Unknown
They're not corporatized it or softening it or anything like that. And LinkedIn, owned by Microsoft investor in OpenAI, using the term slop in kind of a degrading, I mean, in the degrading way that AI slop has become. AI slop has kind of become a like a bad word, right? Oh, yeah. It kind of has the sting of a bad word.

00:55:41:11 - 00:56:02:11
Unknown
It's like, oh, God, don't don't call it slop. I'm offended. You know? That hurts. I was trying. So when I read the story at first it wasn't available on my LinkedIn, but it now is. So they were rolling it out. And of course, I haven't seen anything that looks sloppy since then because I want to try it out, but I don't want to miss Mark anything.

00:56:02:11 - 00:56:21:07
Unknown
That would be bad. Yeah, totally. And so I'm looking what am I looking for is if you pull up one post. Oh, see? Seems like ai slop. Yeah. Okay. On the on the three dot command on the right. That's an option I think. I don't think I'm showing anything. There you go. Yes. You can see it seems like ai slop.

00:56:21:07 - 00:56:49:27
Unknown
I would not put that on Kevin Purdy's post, by the way. Kevin, Kevin's a good dude. Which is not to say that anyone that uses AI to to generate text is not a good person, but, you know, like, do I use AI to generate text? Absolutely I do, but then I do a very heavy, heavy like edit pass and I put it into me, you know, versus like taking the output.

00:56:49:27 - 00:57:12:10
Unknown
And I feel like that's what people react to, you know, like, like we've said, there are they are tools to be used and sometimes they are a tool that takes me from blank page to something, but I don't then just publish that something. I go through top to bottom and I make sure that everything feels me, you know, and is my, my the way I choose to convey that point or whatever.

00:57:12:10 - 00:57:29:29
Unknown
And I think buttons like this are really about the things that you're just like, oh, there's all the tells, all the tells are there, you know, I hate to tells. I cut it to the opposite. I write it because I'm a writer, but then I use it to judge me. Yeah, yeah, well, it was nothing wrong with that too.

00:57:29:29 - 00:57:51:03
Unknown
That's I mean, that's. Yeah, that's a good way to use it. All right. So it's the three three times I see it more on again on Facebook than I do on LinkedIn. But it's this certain length that you hit where it's much longer than any human. Yes. And they're using it for advertising now where especially medical advertising where it's some or air quotes medical.

00:57:51:07 - 00:58:07:09
Unknown
You know, I found this great thing for snoring and it has to tell this story that goes on for a total it's like, seriously, it's still going. I'm not getting to the effing point and I don't know why they think that works. I can't imagine people reading all the through that stuff, but I guess it must. I guess people do.

00:58:07:10 - 00:58:26:17
Unknown
Yeah. Oh geez. And they stay engaged with the post. And that's probably sending a signal to the platform, like read it top to bottom. And therefore, you know what I mean, how those algorithms work. And it was engaged and all that. Oh. So nightmare. What's up buddy? Thank you for the super chat. Once again, the $10 super chat appreciate that.

00:58:26:17 - 00:58:50:27
Unknown
Says I'd argue a lot of the human generated content on LinkedIn is slop two. Based on some of the chasms I've observed between the qualifications, some list, and what an interview then reveals. That's interesting. Yeah. Because you, you can portray a lot in written form. But then when you're talking on like a podcast or whatever, that's when do you actually know what you're talking about.

00:58:50:27 - 00:59:20:06
Unknown
Or did you, you know, kind of generate your way to knowledge I don't know. Yeah. Interesting point. Thank you for. Thank you for your support. Always. Let's see here. Curious on your take on this. So we talked about I think we talked about this last week, Elon Musk's AI, I think it was like in our headlines, was pushing back on a Minnesota ban on new apps.

00:59:20:07 - 00:59:45:24
Unknown
Right. And I think my take last week was like, oh my God. Because of course he is. Whatever. A judge has denied Elon Musk's XAI request to do that, to put that pause on the ban. And what I've started to notice is I've started to notice some interesting kind of like support for for Elon Musk perspective on here.

00:59:45:27 - 01:00:11:23
Unknown
If you play the devil's advocate, what people are noticing is that it's not and they're not supporting him because they think new to fire are defensible, let's say. But because Minnesota essentially wrote what seems to be a flawed law, there's all sorts of other things kind of wrapped up in this. And yes, new DeFi apps, they should, you know, they they should be cracked down on.

01:00:11:24 - 01:00:39:09
Unknown
They're not a good thing in and of themselves. But along with this, there's no consent requirement, no intent requirements, strict liability to $500,000 per image. A definition of intimate parts that apparently is borrowed from the state's sexual assault statutes that technically covers shirtless men, people in swimsuits. There's no satire exemption. So, for instance, South Park's deepfake of Trump naked in the desert would qualify, would violate that.

01:00:39:09 - 01:00:59:04
Unknown
And even Techdirt is basically saying, like, their headline is this is Mike love is writing, says the worst person you know, just filed a good First Amendment lawsuit against a very badly drafted new to fire ban. And so, yeah, so it was interesting. I wanted to write it off immediately because I was like, what are you talking about?

01:00:59:05 - 01:01:16:07
Unknown
Like, why are you supporting nudity apps? But the more I read about it, the more I was like, oh, actually, that's it's interesting. So yeah, just goes to show it's good to know all the ins and outs. But what are your thoughts on this? Yeah, I think there is a First Amendment issue here. I think that's the problem of writing legislation to the specific technology.

01:01:16:15 - 01:01:36:27
Unknown
Not only will it get outdated, but it doesn't deal with the higher level of principle. My favorite example of this that I use all the time is, is the privacy legislation in the males is written to the technology of the letter, and the letter cannot be opened unless with a warrant. But your email can be and your text can be.

01:01:36:28 - 01:01:56:16
Unknown
And other forms of communication can be. Because the principle of privacy was associated with the technology of the postal service. And so you've got to come up with this at a principled level. That's the problem. So then at a principle level.

01:01:56:18 - 01:02:23:03
Unknown
There's a couple of questions here. One is simply the use of my image. Do I have. And privacy law comes in great measure from the unauthorized use of a woman's image or face on bags of flour. Oh wow. Yeah. That's where privacy because it was a use of her, of her image without her permission. And that's where the the trail of American privacy law kind of starts.

01:02:23:05 - 01:02:52:11
Unknown
I wrote about this in my book that no one bought called Public Parts. And so one principle is, do you have a right of image? Can someone use you? As you just mentioned, though, there's a list of things. There's satire, there's other things, and there's public if you're a public person. So there's a conflict there between my right to control my image and somebody else is right to be able to speak, or if I'm simply in public and they take a picture of a street and I'm there.

01:02:52:11 - 01:03:12:11
Unknown
I have long said there is no privacy in public, so you can do that. So if somebody takes a picture of me in public and then they manipulate it in some way, is that their right to speech? So these are not easy issues. The second part of this is obviously the sexual part of this is, is, you know, to get really, really awful in this.

01:03:12:13 - 01:03:44:06
Unknown
And and I'm not going to defend this for a second, but just, just to go to the true devil's advocate here, if you deal with child sexual material that is fully fictional, made up, we as a society find that abhorrent as we should. But and there's argument that that it can cause other problems, but it didn't. If it is fully fictional, if it is fully made up, does that in fact victimize any existing person?

01:03:44:09 - 01:04:04:13
Unknown
It doesn't. Right. So there's that fictional line two that you had. So with this new stuff, if you're not exposing someone's body, but you're making believe that you're exposing their body, is that a line? I can certainly see it. It's used by kids against each other and it's awful. Yeah, but but the line there again isn't the technology.

01:04:04:13 - 01:04:27:15
Unknown
It's more of a of a line of defining harassment or bullying. Right. Intention. Yeah, exactly. So you've got to write this stuff really well. I understand the desire to and I think my reflex would be to say, yeah, I get rid of this crap. Elon is disgusting. But be careful how you write it. Yeah. Well said. Well said.

01:04:27:19 - 01:04:50:21
Unknown
Yeah. So that was I didn't expect myself to to give this story a second take and I'm happy I did and looked a little deeper into it. And that was really, well, well laid out. Let's take a break. And on the other side of the break, we'll we have a few speed round items to round things out. We'll be back here in a moment, so don't go anywhere.

01:04:50:23 - 01:05:22:04
Unknown
All right Gemini Spark which is Google's kind of their their cloud agent system, you know with MCC and all that kind of stuff can now browse the web. It uses your desktop Chrome. It uses your it can access your logged in accounts, your saved passwords. So basically their agent which can be operated in the cloud, that's kind of one of the one of the cool things when Gemini sparked a launched that it could do that things like Claude Court could not do.

01:05:22:04 - 01:05:43:09
Unknown
Now cloud coworker does have cloud capability but now has more capability. So you know it can it can do more things. Here's what I've noticed about spark though. Have you used it at all? Do you have? I haven't I'm not sure I have access to it. So I set up a handful of spark agents early on when I first got access on it.

01:05:43:09 - 01:06:04:11
Unknown
And, you know, they're kind of like, like one of them is is like, monitor for what? What trends are happening on YouTube around AI coverage right now. And give me, give me like a regular report on on that and give me like three ideas that I could approach because at the time I was like, well, maybe I'll do some videos around AI and you know what?

01:06:04:11 - 01:06:23:26
Unknown
What are things that I could connect into a trend that's happening, you know, see, see what that does? And then I kind of I checked it for a while and then I kind of lost interest in it. And I maybe this is just a symbol of of my own irresponsibility. I haven't gone back into spark yet to say I don't need that anymore.

01:06:23:28 - 01:06:42:24
Unknown
And so now it's an agent running every couple of days in the cloud because it's in the cloud. So it's like out of sight, out of mind sort of thing. I do get notified about it, but, you know, notifications sometimes you, you see a notification so many times that you stop seeing it. You know what I mean? Yep, yep.

01:06:42:26 - 01:07:03:10
Unknown
And so I've realized I've done that with spark is that it's cool that it has these things, but I kind of like quickly fall out of the habit of checking it. And then I realize weeks later that, oh, it's kind of like I left the the water on and it's been dripping for two weeks. And I realized that the other day I was like, oh, shoot, I feel kind of bad about that.

01:07:03:11 - 01:07:30:08
Unknown
Like, that's wasteful. I need to turn that off. Guess what? Guess what? Yes. I can't use spark. Guess what? Of course you can't. Of course you can't. Workspace. I'm paying you Google and I can't use spark. I, I think I think the weird, the weird news is when you can use something. Yes, exactly. It's news. When you can use something that sucks.

01:07:30:08 - 01:07:35:20
Unknown
I'm sorry. That's lame. Google.

01:07:35:22 - 01:08:00:28
Unknown
Yes, AI Pro also got expansion, but that probably doesn't apply to you either. Let's be honest, Google had launched a feature in Google Earth that let users generate AI satellite imagery by typing in descriptions and basically said, like Google Maps is your playground nano banana integrated? No way anyone's going to do something that we agree with. People are wonderful.

01:08:01:00 - 01:08:27:02
Unknown
Why would anyone do anything that that challenges and makes things, you know, not, you know, kind of scary or whatever. And sure enough, they pulled it the next day because people were going in there and doing things like Iran's Kharg Island engulfed in flames, flooding the US Capitol. Bomb damage at hospitals, you know, that sort of stuff. So, yeah, that doesn't exist anymore.

01:08:27:03 - 01:08:50:05
Unknown
I mean, when I heard about that feature, I was like, really? That sounds that sounds like a bad idea, but okay, so apparently it was. Yeah, you should have known. You should have known. Yeah, yeah. And they probably do. No to some degree. But like, how could they not know like. Right. What's the I don't understand the point of it really.

01:08:50:05 - 01:09:19:00
Unknown
If people say you're, you're taking something that has credibility and authority, satellite images that are really important and used for all kinds of used for journalistic investigations around war crimes and all kinds of important stuff. And then you're saying, okay, screw with this. Like, that's such an interesting question. Not that you couldn't take a screenshot and do something you could, but to to enable it to that extent, you're giving people an idea that you shouldn't have given them.

01:09:19:03 - 01:09:46:10
Unknown
Yeah. You're you're in. Yeah. It's enticing to the people who love to see the world burn, to literally make images of the world burning. So yeah. Thinking Machines Lab released Inkling Small. It's a smaller open weights model. They'd released a larger one, I think, a couple of weeks ago, and they say it's really great open weights. You know, they're doing the open weights thing.

01:09:46:10 - 01:10:09:28
Unknown
And so I got to give them credit for that thinking machines okay. You go. And then let's see here finally for a little little laugh at the end. So Sam Altman's driving his kids in the car. And he, and he tweets about how he found this wonderful way to use ChatGPT. He's driving the kids in the car. He has ChatGPT make a podcast for his kids.

01:10:10:00 - 01:10:28:23
Unknown
The the the ratio was hilarious. It was just great. Nobody could stand it. They shouldn't have tolerated this from Sam. You should know better, Sam. And people went on them like crazy. I don't know if I can find it.

01:10:28:26 - 01:10:41:26
Unknown
At the. The baking timer is going off right now. Can you? My daughter is apparently in her room. Has no idea.

01:10:41:28 - 01:11:07:18
Unknown
Here she comes, turning off the timer in the kitchen, I love it. Yeah, so that's been really good though. Anyways, you don't want to burn. You do not want to know. You don't. You don't want it to burn. So Brooke Hamelin, who's a well-known publicist in Silicon Valley, I do mean this with kindness. But like maybe before you post publicly, you should have some trusted people in your world who can guide you to make different choices.

01:11:07:23 - 01:11:31:26
Unknown
People not afraid to push back on you. People who might not be too close to the thing that's blinding you. Folks, why don't you just talk to your kids? Yeah, just talk to your kids. It's wild how disconnected these tech people are from normal human behavior. Or you could be an actual parent. Instead of having AI summarize your kids for you, you absolute sociopath.

01:11:31:28 - 01:11:55:00
Unknown
I feel bad for Sam Altman. He probably thought this was some feel good use case, and had no idea how creepy and weird it is. Yeah. Can we please get some normal people running these tech companies? I think that's probably the case. Yeah. So went on and on and on and on. In light of OpenAI agents doing all sorts of weird things and everything, it was like, oh, well, here's an easy win.

01:11:55:00 - 01:12:25:26
Unknown
Here's here's the way that we can kind of, you know, show ourselves as on the good side of, oh, no, never mind. Yeah, yeah, yeah. Okay. I guess you can do that. I mean, you could your credit, Sam Altman. That is possible. You can do that. You're right. I think that's it. We reached the end of this episode of AI Inside Hot Type two weeks out about ish 2019, and I just did a deal.

01:12:25:26 - 01:12:38:20
Unknown
I'm just doing a deal with a bookstore. Not. I can't put it up yet because I don't have it done yet. Where? If you want an autographed copy where some folks want, if you order from this local bookstore of mine in new Jersey that I will point you to, they will fulfill it from there, which is really cool.

01:12:38:20 - 01:13:01:24
Unknown
So awesome. That's super cool. It's getting close hot type, getting hotter and hotter by the day. Jeff Jarvis. Com preorder. Now you can, you know, have it shipped right on the day that it's released. And then all of your other amazing books maybe not the book that you mentioned earlier that you said nobody read. But that was a long time ago.

01:13:01:29 - 01:13:31:06
Unknown
Somebody read it. It was just a long time ago. Right? Yeah. Right. Yeah, yeah. And then we also mentioned earlier in the show intelligence, AI and humanity coming sometime in the near future. Yep. Early next year. Brewery pod tuneup. Com. If you want my help with your podcasts and or just want to talk things through on your ideas, visions, ambitions, all those kinds of things, I've been doing it for a long time and I will.

01:13:31:08 - 01:13:57:00
Unknown
I will be happy to talk to you about your podcast and see if there's any way that we can work together. Pod tune up com AI Inside Show for all of the information about this show that we do hear all of your reviews, our little bios, everything can be found there. And then let's see here. That's right. Should also mention one last time Patreon.com AI inside show.

01:13:57:02 - 01:14:17:14
Unknown
That is the the Patreon where if you go there and you are not currently a patron and you sign up for one of our paid tiers with the code 100 patrons, you'll get an additional 30% off the year on top of the 20% discount that you normally get when you pay for a year. So thank you. You get the AI Inside Daily podcast, you get ad free shows.

01:14:17:14 - 01:14:42:14
Unknown
You get a discord community that is growing. We've got more voices in the discord community, so it's some pretty cool stuff happening there. I'm throwing a bonus patron, ask me anything. At the end of August. You get a sticker if you like stickers, your name mentioned on the show, you get your name mentioned on the show, and then the the top tier is not part of the sale just because it's the exclusive tier, it's the executive producer tier.

01:14:42:14 - 01:15:04:23
Unknown
And I want my executive producers to feel special. So if you want to, you can become an executive producer like Doctor Do. Jeffrey Marikina, Radio Asheville, one of 3.7 Dante Saint James Bond, Derek Jason knife or Jason Brady, Anthony Dallas, Mark. Starker, Carsten. They get their names read out at the end of every single episode. So at this point, y'all can have it memorized, right?

01:15:04:26 - 01:15:24:15
Unknown
So there you go. And finally, a big thank you to Victor Bogner and Daniel Croft for video support behind the scenes. And, yeah, every single week. All right, Jeff, thank you for the day early I appreciate it. Great hanging out with you. Great seeing everybody. We'll see you next time on AI inside. Take care. Bye.