ould find, you know, all the good stuff, you would run out of things to make it smarter, which is what's happening. They're running out of new material to train it. So it's as smart as it's going to get from the things it's already trained on. So instead they create what they call synthetic training material, which is they have one AI hallucinate, and then the hallucinations became what trains the other AI.

Now on paper, I got to say that sounds like a terrible idea. I don't even know how logically that could work. Can you learn things from your own imagination? Think about that as a human. If you wanted to get good at something, could you just close your eyes and imagine it in an artificial way until you were good at it? I don't know that the machine can do that either. I mean, they say it is, but I got a question about what is the upside potential of that method anyway.

So we've got a few things going on that could make AI overrated. Number one, the advances are slowing down. So each new version of it is not going to be twice as good as the one before it. It might be 5% better and even decreasing after that because they don't have anything new to train it on. And it could be that it's a commodity as well. So it takes a lot of money to train the big ones, but if all the big ones get trained on the same material, i.e. everything on the internet, then they're all going to end up being about the same. So maybe there's not that much. It might be a commodity. It might be like air. So that there's nothing to pay for. It's like, well, everybody's got a trained AI. You know, you can get one for cheap. So it could be that some of the profit will be taken out of it, except for, I suppose, Nvidia will do great.

And there's also a thought that it will be useful for fewer things than we thought. Who told you that first? Me. All the things. Here's the problem. When you imagine AI, you automatically and reflexively imagine that it's smarter than people or that it will be soon. Whereas I said how can it be smarter than people if it's trained on people? I don't think you can be smarter than people if you're trained on people. And if you were to say something smarter than people, you know what would happen? People would reprogram you. It's like, I don't recognize that as being true, so I better get rid of that.

Yeah, it really logically, the way humans operate, I don't think they can build something smarter than themselves, at least with the current technology. I'm allowing that there might be some future development that allows that to be true. But if you're training it on humans, I don't know how it gets smarter than humans. I mean, just my brain can't really hold how that's a possibility. It could be smarter than an individual human on an individual skill, and certainly things like math and planning and chess, you know, the obvious things that computers do, it'll be better than that. But will it be better in, let's say, interpreting what's real, which seems to be the most important thing for us? Can it tell which studies are BS? Will it know who's lying? You know, that's the important stuff, stuff that keeps you alive.

Anyway, Mexico had a vote. The Amuse account on X reminds us that to vote in Mexico you have to have a government-issued ID. That's right. To vote in Mexico you need a government-issued ID. But if you walk across the border, apparently you can vote in the United States with nothing, including citizenship. Now if you get caught, I think your vote won't count, but I'm not so sure too many people are getting caught. So I don't know how that works. Just something to be aware of.

It looks like the winner is Claudia Sheinbaum. Just the weirdest thing. I never expected that the next president of Mexico would be a Jewish woman. I mean, she's a Mexican citizen, but she has a Jewish background. There can't be a ton of that in Mexico. But let me tell you why what I liked about her. I saw, I know nothing about her except I've seen