Predicting human text can require reasoning beyond the writer

Transcript

From the transcript The Diary Of A CEO publishes with the episode.

1. 1:09:03

   Can you do it as if I'm 16 years old? So the way that a modern AI is made is there is no one programming it. There is no one typing in if this, then that. We're not sort of like writing the code. What happens is you collect an enormous number of computer chips into a huge data center that has basically a trillion numbers inside those computers that you basically start out randomized. Right. And you hook them up in a pretty simple way that involves addition, multiplication, and setting the number to zero if it was negative.

2. 1:09:35

   So it's very simple math operations that are hooking this all up. And you're basically going to put words in the top, and you're going to get numbers out the bottom, and you're going to interpret those numbers at the bottom as a ranked list of words. That's, that's, it's basically the AI's guess of which word is, is here. So you put in like once upon a blank and you're hoping that the word time will come out, but it doesn't because you just have a trillion random numbers hooked up with simple math.

3. 1:09:57

   But here's the trick. You can go to every one of those trillion numbers and you can tune it up a little and you can see, does that make the word time go up or down the list? And you can tune it down a little and see, does that make the word time go down the list? And you set it whatever direction makes the word time go higher up the list. You do this to a trillion numbers a trillion times for basically every word of text ever digitized.

4. 1:10:19

   It's not quite that much. They filter it. But you basically do this to a trillion numbers a trillion times, and then the machine's talking. And we're like, well, how about that? No one really knows quite why. The things that humans code is the thing that runs each of those trillion numbers and tunes it and sees whether the right word goes up and down the list. But... We don't know how it's working in there.

5. 1:10:42

   Then, and that's how it worked up until 2024. In 2024, they started adding another layer where you then train it on basically 100 million hard problems. And you don't just have the AI like produce an answer to the problem. You have it produced like a book worth of text about how it's going to solve the problem. And then you use that book worth of text to sort of try and figure out the problem. Or maybe an essay worth of text, depending how you're doing it.

6. 1:11:04

   So you have it produced this text about like, you know, they call it reasoning about the problem. We could argue all day about whether it's true reasoning. That's just what it's called in the field. They produce this reasoning about the problem and then produce the answer from there. You have them, you train them to solve 100 million of these hard problems. help them predict all of that text in the first phase and solve all those problems in the second phase.

7. 1:11:29

   And this is called a large language model. We probably should have stopped calling them large language models when we started doing the reasoning and the problem solving. One of the things I want to hear your explanation as a muggle, like I am, is it sounds like it's like a word machine. And then, you know, you made it like a problem machine. And I go, okay, so I can solve problems over here and it's a word machine.

8. 1:11:46

   What's the risk of this? Yeah, so let's take the word machine part first. Predicting words that humans wrote often requires solving a harder problem than the human who wrote them. So suppose that you go and inject a drug in a rat. And you're like, you know, it's like you write down the chemical nature of the drug. You inject it into the rat. You see that the rat dies. And so you're like, when I put that drug into the rat, the rat died.

9. 1:12:13

   Now suppose you're training an AI and the AI sees the chemical nature of the drug. It sees when I put that drug into the rat, the rat blank dies. The human who wrote it down gets to just look at what happened to the rat. The AI predicting what was written does not get to just look at the rat. So training AIs to predict human text is training them to be potentially smarter than the humans. Because they need to be able to answer these questions.

10. 1:12:41

    They need to be able to predict. They need to be able to fill in the blanks where humans were writing down what they saw. And just so I understand technologically, there is no knowledge they have, though. Each time, and there are ways of mitigating these, each time it is effectively rereading. But because of training, it gets more accurate at certain things. I mean, somehow as you tune the knobs, somehow it's getting information in there and we don't know how.

11. 1:13:06

    So it's much easier than that. We're humans who have a brain. Brains are made of neurons. Then we try to copy that on a computer. We simplify it, but we create a neural network. So we're making artificial brains. Just like with human brains, with cognitive science, we don't really understand how you function, how you learn, where in your brain certain memories are stored. We have some glimpses of understanding this neuron fires the new CFAs.

12. 1:13:32

    But there is no complete picture. And so a lot of times you can't get intuitive understanding of what's going on, then you just think about it as artificial persons. It's not exact mapping, but it helps. So if you send a child through 12 years of education, they get lots of problems to look at, and then they graduate and become a little better at solving problems. This is what we're trying to replicate here.

13. 1:13:54

    People complain that it takes a lot of money to train those very, you know, intense process. You forget that it takes 20 years to train a human. And they are not general superintelligences. They are very narrow. We're lucky if they graduate with a bachelor's. So a lot of it is exactly the same. Can we make safe humans, for example?