Back in America

  📍 📍 I love the idea that we're producing truth. This is about truth. We live in a world that is full of non-truth and fake information and politicians who lie to us on a regular basis and don't even know how to tell the truth and that truth that turns out is beautiful.

and here I live in a world that is based solely on a single truth, that there is a best way to proceed.

Today I have the pleasure to speak with Dr. John Edwards, the International Correspondence Chess Federation, grand Master, and the 32nd World Correspondence chess champion. He won the title in 2022. Is the only third American to hold it and was named 2023 US Chess Grandma, master of the Year.

John used to work in tech at Princeton. He has written over a Dohan books like Teach Yourself Visually Chess and the Chess Base Complete, and he breaks down how to combine big databases and AI to play with precision.

It might be surprising for our listener to understand that you, chess against another human being assisted.

By a computer, which is now an ai, right? Yes. And in the server that we saw downstairs, you've got a databases full of previous moves. Basically all the moves have been analyzed, which mean that the computer knows what's happening on your screen and can give you references to other games, et cetera. So where does the human come into play

there?

Well, not really in the opening. Because the databases are such that the opening, which can last up until move 30 is done.

there

is a database that is maintained by the Chinese, um, largest opening database in the world.

And they basically, if you put the move in. It will say, this move is good, or this move is less good than that, or This move is terrible and it will tell you why. And this is, this is useful all the way through to about move 20. today if I want to learn an opening, it's not it might strike your listeners that learning a complex opening must take years. Actually, it can be done in minutes now less than, seconds, frankly, if, if the servers are powerful enough.

The end of the game's been solved. We're up to seven now. After all of this time, we have solved the game for any seven units that you put on the board. It will tell you if you're winning and if so, how? work at Bell Labs. So we're familiar with what they did. They, the graphical interfaces with mice, um, Unix c plus plus. This was Dennis Ritchie. This was Brian Kernighan. They cared more about chess than they did about all those amazing things that they wound up working on that they're famous for. And they had this notion that the o, the opening was being solved. True?

Mm-hmm.

If we solve the end of the game, put five pieces on the board and the computer will tell you this is a win or a draw. Six pieces, it's exponential. We're up to seven now. After all of this time, we have solved the game for any seven units that you put on the board. It will tell you if you're winning and if so, how?

And if it's a draw and if then how?

So that leaves a conversation in the middle, the opening's done, the end game's done. There's this shrinking conversation in the middle of the board that the neural net has to deal with. It is my contention, that this middle game is not so small and it's not really shrinking. It's still substantial, and there are class of positions. That involve fixed pawn structures in which the maneuvering can go on for 50, 60, a hundred, 200 moves in order to simply 30 moves, in my case, to win a tempo, to win a single move, to get the pawn to move forward. And then we start this over again.

And that leaves for rich exploration in the game.

Do you think there is an element of human psychology?

Of course. But not the way people might imagine. I can't bluff anybody, so that's out. But can I steer the game into positions with fixed structures such that the method that I just described.

Can work if I do select an extremely complex line with fantastically involved combinations, it's going to be a draw because the machine is good at that and it will guide me through and there's no room at all for human intervention at all. Rather, it's this long slow plotting maneuvering of spending 35 moves to win a tempo where humans excel and the machines.

Of wonder what you're doing.

you started playing 50 years ago, I believe. Yes. What hooked you at first and what kept you coming back to all those game throughout the years?

It's very comforting. So as a young person, a typical male, I had ADHD. Which meant that I was flying in every direction all at once. But chess was grounding. It had provided action in a fixed space, and it settled me down more to the point when I was in fourth grade, I got mononucleosis.

Which back then meant that I would be in bed for six months if I was lucky, more likely a year. And my father asked how can I make this? Less onerous. And I said, could you get me a chess book? We had had a visitor at the house and he'd played chess and asked, I had a chess board. I wasn't good.

And he beat me and he beat me again, and he beat me again and he beat me again. And I didn't like that. But I was also curious as to how it was that somebody could be good at this. So I asked my dad to buy me a chess book. The book was Great. Moments in Modern Chess by Ruben. Fine, I remember it distinctly. And I was in bed. I was reading the book, and four or five days later, he came in and said, how are you doing with that chess book? And I said, well, I finished it. It's a 300 page book with lots of games. And he said, you finished it? I said, yeah, uh, you wanna quiz me? And he started to ask me questions about it and I had imbibed the book.

And after that he started to buy me a book a week.

There's a very good parenting point here, is that when your child expresses an interest in something, however unusual it might be, you support the interest. They were tough lessons because he had no patience. Yeah. But I took to those lessons very rapidly. With the result, I also started playing in tournaments. I started winning money, which impressed my parents. Impressed. Me too. How old were you at that time? By the time I was winning money, I was 14.

14, but I'd been taught correctly.

Yeah.

And I had done exactly what you're supposed to do without realizing that this is what you should do. I had played over lots of games. I had studied the game the right way. Yeah. It's, if you wanna build bridges, you should look at how a hundred or 200 bridges are built.

You look at how it's been done. And that's exactly what I was doing. I was looking at how the best players in the world had played. And you can't help after a fashion to start. If you only see good moves, you start making good moves yourself.

---So moving on, you worked at Princeton University in technology, right? Yes. For many years, 24 before returning to full-time chess. That's right. And, and I wonder how did that work? If it did or not, shape the way you think about intelligence, both human oh and artificial.

---

The life at Princeton, 24 years as first as assistant VP for computing and later on as coordinator of institutional communication.

it occurred to me as we were doing that, that we were creating scholarly environment in every discipline. The key for working with Thoreau was to take a machine, put all of Thoreau into the machine, put everything written about Thoreau, put everything about his life and times so that a student when working on Thoreau doesn't have to spend all their time in the library looking for this material.

Let's give it to them. May I tell one quick story that may Sure. I don't wanna use the names, but we went to see the chairman of the department of the classics. This is a fellow who had specialized in ancient Greek and ancient Latin.

Mm-hmm. Spent his entire life looking for the derivations of various words. He had written scholarly articles about the 14 uses of a, a single word throughout the literature, which is not growing. Ancient Greek is ancient Greek. Ancient Latin is not growing. Um, and he was working on another about a phrase.

Here we came in with Abacus, which was the brainchild of David Packard, of Hewlett Packard Vein.

Mm-hmm.

And put it down and he did his search and lo and behold, everything that he had worked on came, popped up to the top, except that he had missed one. He didn't get embarrassed. He was fascinated that he had missed one.

He was putting in more searches. He was 55 years old at the time. He spent the last 10 years of his professional life turning out more papers. In those 10 years that he had in the 40 years of his, in 30 years of his professional life, never once crediting Abacus for helping him do the research, but he became amazingly productive.

I already had a scholars environment for chess.

And so the key there was to extend this into every discipline I could at Princeton.

And by the time I left we had really done that.

Mm-hmm.

With phenomenal results.

do you think you would be where you are today without your experience at Princeton?

We're all creatures of our environment and experiences and the rest. Chess actually was one of the most useful things that ever happened.

I rarely ever credited Princeton. But what I was attempting to do was literally to give everybody access to the same kind of environment that I already had, which pleased me no end because I could accomplish more in an hour.

With those tools than I had ever been able. I was putting in seven or eight hours a day when I was 14 and 15. I'd come home from school and the schoolwork was trivial. I just wanted to get to my chest. And so as I was getting good grades, my parents didn't care. I wanted to extend the excitement of that, and I could accomplish more now in an hour with these tools than I could in seven hours.

As a kid, if you're playing through a game, you have to set up the board and you play through each line. If you go through a variation. I may have to go back to the beginning and set up the pieces manually. Now everything's on screen.

is a reference to what happened in the World Championship against a fellow named OS pov. Where this is actually the most important part of why I think you're here and why we're having this conversation. We arrived at a position in which the neural net then and since does not really understand what's happening here.

I realized there was a pawn on a square B six. And as a human being, I said, well, I've played through enough games.

If the pawn were not on B six, but if the pawn were on B three three moves from B six to B five B five to B four B four to B three, if the pawn were on B three, I'd be winning and I'd be winning because my king would go to this square, my rooks would go over here.

from B five to B four took 38 moves. Now I had to get the pond from B four to B three, and I knew exactly how to do it. But the method required far more than 50 moves. It could not be done more quickly than that.

And so the game, in fact, tragically ended in a, I still won the world title. Yeah. But I have to tell you, I was, I had worked harder at that one position than anything in my life. I think the major takeaway is that a year and a half, two years before the end of that game, I understood what I needed to do in a way that the neural net did not.

I was able to work out the method in a way that the neural net provided almost no, no meaningful help. I saw all of that. Now, does that mean I'm thinking ahead far further than the neural net? No, I'm thinking in a different way.

Chess has a funny rule. If you go 50 moves without a pawn move or a capture, the game ends in a draw.

We don't think about what move to play that's trying, strikes everybody as highly unusual. We think instead about where our pieces belong.

So what

do you think still separate the creative player from the perfect machine? Oh, tremendous amount of experience on the player side, but I mean, the machine has got the entire humanity of experience. Machine

calculates remarkably well, but it doesn't, it has a hole and a human being steers every game now into that hole.

That hole is. Is, uh, pushing every position into, into those fixed pawn structures with long term maneuver. There is, that is the only place where the hole exists. There is nothing else where where the machine doesn't completely excel perfectly. Where it's horizon is not so large that it just overpowers the game.

I do believe that it's possible occasionally to win. Somebody dies. That's what a horrible thing. Why would I get involved in a, in a sport where I'm waiting around for six months or a year for someone to die? And I think actually, frankly, the rules should be changed so that if someone dies, their games are simply voided unless they legitimately won or lost, um, a game.

Um, there are games where people, humans make clerical mistakes. I was supposed to play the move F four and instead stupidly, I played F three and I hit send, and I'm a moron for having allowed that to happen. But, but occasionally that will happen. Um, the third I mentioned, there were three, and the third is the most interesting, and that is that. Through our cleverness and there is a significant amount of cleverness involved.

We steer the position into one in which that long-term maneuvering and a fixed structure can occur.

There are different end game types, for example. Middle game types, end game types that are conducive to wins. I showed you one downstairs there was a position involved. Both sides had a rook, and both sides have bishops of opposite color. It's a famous end game.

It was popularized by Bent Larson in the 1970s and eighties.

I've been at the, and I really feel this, I've been at the forefront of AI for three and a half years.

AI is a better accountant. AI is certainly a better programmer. It will program something quickly and correctly. There'll be no errors. the message for everyone else is watch out. . They should get the most general liberal arts education and they should learn how to think. Um, and they should be prepared, frankly, to understand this environment, to work with this environment in meaningful ways. They're not going to outthink it.

They're probably not going to be able to do that. They need to learn how best to work with it. And the most interesting aspect of this is a paper that was done by a fellow at Rutgers, who was looking at the human computer interface and wanted to know if humans brought anything at all to the party anymore.

Now can't we just turn the whole bloody thing over to a neural net and be done? And I'd like to think that a human being brings something to the party, at least for the next year or two, or three or four. But it's becoming clear to me that it requires a manic human being to make a difference.

So, 📍 question about the computer you build and and your database I read that you are 📍 using open sources. Database right what are those machine downstairs?

So the 📍 machine itself, the big server, is running chessBase is proprietary and runs its own, has its own databases that it provides, but it also allows you to access open source databases.

the single most important database is of correspondence chess games that is maintained by the International Correspondence Chest Federation on a monthly basis.

When I, I trained my own neural net neural nets contain two parts. Part one are the rules and the rules don't change. Everybody's neural net contains that first, that same first part. Mm-hmm. The second part is the experience based and the experience base should be based upon the highest quality games, and so I have fed it the games of strong correspondence players. Imagine Chat, GBT having trained solely on the works of Nobel Laureates.

Correspondence players are at the forefront of chess. The theoretical work is ours. We are setting the stage in the opening work and in middle game ideas that the best players in the world are emulating.

Yeah, I was learning that classic chess player

are studying what you guys are doing. Yes,

They care about what we do. We tend to be, we don't talk about it ever, but we're working with some of these folks and when we find things they want to know about it and they want to know about it quickly.

And the best, the most fun example, I Beat , the reigning Russian correspondence, chess champion, a fellow named Lebanon with a magnificent new idea, in a variation of the Sicilian, in a variation that got I and I couldn't talk about the game until it was finished. Okay. That's an incredible frustration.

Yeah. This is a beautiful game and a beautiful idea. Um. Right around the time what? As soon as the game ended, I forwarded the game to Peter Nielsen, who was Magnus Carlson's second, and I said, I think you'll find this interesting.

the Russians are in a terrible spot because they don't have the capital and they can't afford the servers.

in order to compete meaningfully at the world championship level two skills are required. You've gotta be a really good chess player, and that's hard enough. Those skills are rare. Not everybody's a great chess player, but it's not enough.

You also have to be technically proficient.

So are we thinking the Chinese? No, there's not a Chinese competitor in the world Championship. So Germans are the Europeans? Yeah. Primarily the Germans. Yeah, Dutch.

Okay.

And they're all, they're all reasonably good friends. I, once I won, when I won the world championship, we traveled to, I picked up my medals, uh, in Amsterdam.

And I got to meet an awful lot of the folks. I got to sit down with some of those who wanted to do it, but weren't technically proficient and wanted to know. That was the question. The chess skills were there. Yeah, but, but what did you do? How do I build my server and, and what tools and, and how do I initialize this and, and how do I work with the neural net correctly?

Those are the questions. The books that I'm surrounded by, all these chess books. But these are not the books that we're reading. The books that we're reading are on the top shelf up here. They're on how to program neural nets.

So my last question is, what is America to you? What is

America to me? Currently, boy, an enormous disappointment. An enormous embarrassment. I'm embarrassed by this leadership, which is.

Devoid of the one quality that I care most about as a human being, and that's empathy. I'm not seeing it, and I, I miss it. I do. Can America

recover?

I think that's a phenomenally interesting question that should be fed to a neural net. 'cause I don't think I'm smart enough to know the answer, but the reality is, it's, the answer is quite possibly no.

You know, it's, it's so much easier to tear down a building than it is to build it. The east wing of the White House is a perfect metaphor for what we're seeing. It got torn down in two days. All, all that history is gone. It got torn down in two days, and if we decide that we want to build it or whatever good luck.

I don't, I was, I ran for office in that interim period when I wasn't playing chess and I became mayor of my town. I was on the committee for six years. A certain frustration of having to do things that are of political nature. You need a consensus, which means that you have to settle on the conclusion that no one of us has a monopoly on wisdom.

And I think that's what our system of government was meant to encourage. I see. I see no such conversations occurring and I'm frightened out of my, out of my mind that it was so easy to make the transition 📍 away from a healthy democracy or maybe not so healthy, but away from a democracy which had its frustrations, but which was absolutely a democracy to something that doesn't appear democratic.

John Edwards, thank you so much for your time today. You're

very

welcome.