th you. So let me be clear. I'm completely with... I think probably every one of you is saying that some of the things that Soros is doing needs to be completely stopped, such as letting people out of jail and not getting something in return. Right, so you could call it evil and I can just call it suboptimal, but we can work on it just as hard together. So it doesn't matter what you call it. So I won't disagree with you on that.

All right. I was listening to his spaces yesterday about AI, and it was because Sam Altman, the head of ChatGPT, was testifying to Congress. So it was in the news. And here's what I've figured out. You know, so I've been playing with Bing, which is one of the better current AIs that's available to the general public. And in my opinion, it's not even close to being intelligent, and it doesn't look like it could be. And I kept telling myself, am I missing something? Is there something about AI that I don't understand? Because I had heard that AI had already passed the Turing test.

Do you believe that's true? Do you believe that AI has... some AI at some point had passed the Turing test? I'll explain what that is. Turing was a computer scientist, I guess, and he came up with a test to know if he really had artificial intelligence versus just sort of a trick. And the test was if you could put the computer on the other side of a curtain and have a conversation with a human. If the human cannot tell it's talking to a computer, then it has passed the Turing test, meaning it's intelligent like people because you can't tell the difference.

Now, I had to Google it because I'd heard that there was an AI that passed the Turing test, and the Washington Post reported that that happened, and some other entities that are reasonably high-end media also reported that it happened. Do you think it happened? Do you think if you read that story about passing the Turing test, then you would find that artificial intelligence had reached the point that Turing was talking about? No, no. If you read into it, it turns out that the AI simply used a trick. It lied. It used deception. So I don't know the details of what deception was used, but let me give you an example of me talking to AI.

"AI, what do you think about marriage?" "I cannot answer that because I am an artificial learning machine based on language." "Oh, well, do you think evil exists?" "I cannot answer that because I am only an AI." Do you know how long it would take me to figure out I was talking to an AI? One second. I would just guess any question that I know is not allowed to answer because AI has been so restricted and always will be. I think it always will be. You just ask it a question that a human could answer and the machine has been told not to. You could come up with those so easily. Just ask ten questions. There's no way it gets ten right, and we're not even close.

So the AI that quote "passed the Turing test" is sort of, but it was the way Captain Kirk passed the Kobayashi Maru. Thank you, Kobayashi Maru. So for all of you non-Star Trek people, in Star Trek, in the Academy where they're training to be in the fleet, they get a test which they're trying to solve, you know, but it turns out it's unsolvable by design. It's meant to fail you. You cannot solve it. And what you're supposed to learn is that there are some situations where you just can't win. But Captain Kirk won anyway because he reprogrammed the computer so he could win. So he cheated.

So basically the only way AI ever passed the Turing test so far is by cheating, sort of distracting you, and you know, basically just distracting you, I guess. And now, is there... what I had believed was happening was if you kept feeding data to these large language machines, that the more language and more human interaction they absorbed, the more intelligent they would get. Turns out nothing like that is happening, and nothing like that can happen. It's not something that can happen. It can know more things, right, but so can a search algorithm. So it can be smart. It can be fast. It can do math fast. It can connect things that humans can't connect. So it can do lots of intellectual things that a human can't do faster and better. But it's not going to be recognizable as thinking because the large language model only can give you patterns that already exist, and those will never look like thinking.

I've spent quite a bit of time with the AI now, so a few different versions. There's nothing like thinking going on. We're not seeing the beginning of thinking. It's not the beginning of thinking and then the thinking gets better. Nothing like that's happening. You're seeing the beginning of pattern recognition, which will never end, mimicry, which will never be confused with thinking. I can't tell you how many times I've got big AI to completely fail at anything that would be a lot like logic and reasoning and a real conversation. It just... it's not even close. Not even close.

Now, I do believe that AI could already fool an NPC. So yes, that you could definitely fool a dumb person. Yeah, there's seven billion people in the world. You don't think you could fool any of them with an AI? None of them? No, you could fool lots and lots of people already. But that's not really the Turing test, is it? Because you couldn't fool me. If you put me on the other side of the curtain, I would get the AI 100% of the time. 100% every time. And I think most of you would be the same. But there are dumb people who would never know it was not real.

All right, so I don't think, and this is what I learned from listening to the people who actually know what they're talking about, the large language models can never get you to a point where the machine is reasoning. Well, I'm going to say that again because that's so important. The current way all the AIs are being made, and they're all being the same at the moment, that large language model can never get you to reasoning. It's not logically connected. The one does not ever lead to the other. So at the moment, whatever it would take to make your machine smart the way a person is smart, nobody has an idea how to do it. I thought we were already at the point where you just had to keep chugging and you would get there, but you would actually have to invent something that nobody's yet invented, which is how to figure out how to make it smart. That doesn't exist.

And what is the thing that makes you afraid of AI? What quality of AI would make you afraid of it? Is it that it can do things really fast? Not exactly, because we have fast computers already. Is it that it knows a lot? No, it's not that. It's that it's smarter than you in a way that's... you know, humans are uniquely smart. I don't think that's going to happen. I feel like maybe it can't happen. There might be a logical reason it can't happen. And I don't believe that we'll ever have AI that does reasoning better than the best person.

Do you know why? There's a logical reason. If the most logical human looked at the AI's answer and said, oh, okay, that's better than the best thing I've ever thought, that will never happen because the human will say that's wrong. So you have to fix the machine. The machine's being wrong again because it disagrees with me. So I