Sana’s founder and CEO Joel Hellermark sits down with chess grandmaster and human rights advocate Garry Kasparov for a candid conversation about the evolution of chess, the rise of machine intelligence, and the human qualities that endure in a world transformed by technology. They discuss AI geopolitics, learning from defeat, and the meaning of progress.
Watch the full episode above or read the transcript of their conversation below.
The nature of play
How often do you play nowadays?
Probably every day.
Every day?
Every day, still. It's just—but, you know, I can hardly call it play because it's more like, you know, having fun. So the word “play,” actually, for me, is just—it always has the professional connotation.
Yeah.
Play means, you know, you play.
Yeah, not—not exactly play.
AI, geopolitics, and power
When you look at the geopolitics of AI, if we, over the next couple of years, get intelligence on par with Nobel Prize winners, and, you know, we get to AGI and then superintelligence and beyond—that should be a very useful technology for a nation-state that wants to remain in control. What do you think, for example, Putin could actually do if he decided that AGI was his top priority?
I'm—I'm very skeptical about AGI and superintelligence and singularity. So, okay, I might be old-fashioned and wrong and behind the curve, but this is big talk. So far, everything I see: it's a jump and then plateau. The development of computer science is not one line. It's always—you have a jump and then quite a long plateau. So we're not even close to AGI as it is being described. Maybe we just have different concepts of that.
But AGI or not, superintelligence or not, computing power of that magnitude in the hands of bad guys—that's a challenge. So I keep repeating that humans have always had a monopoly for evil.
AI as a weapon and the challenge to free societies
So that's why our main concern should be now: what happens if this technology—actually, not if, you know, it's already being used by these bad actors, by Putin, by Chinese, by Iranians—to undermine the free world? Because the free world is—democracies are very vulnerable. They are open for these attacks, and it's getting more and more complicated to defend because one of the reasons is that we don't want to recognize we're at war.
Also, I think this is a lot of just wishful thinking. Oh, let's, you know, let's live in a world where we can share the information and the technology. I mean, come on. It's not about sharing technology. It's not about, you know, just living in harmony. We still live in a very dangerous world, and if you look at a geopolitical map, you understand we are going through a period of great turbulence. It's a geopolitical redistribution of the map. It's a technological breakthrough, and quite an irony that the main weapon that has been used against the free world is something that has been invented, conceived, designed, and developed in the free world.
So we just have to recognize it—harsh reality—and also recognize that our attempt to regulate things gives the other side, you know, an advantage. I mean, DeepSick is just—it was clearly, you know, just a result of stealing some technologies but also access to unlimited data.
Data, regulation, and the digital iron curtain
It's okay if I can make this analogy: the engine—the bot—is kind of the engine of the car, but the data is the fuel. The Chinese engine could be worse, not as good as the European or American one, but if you have an enormous amount of free fuel, you can put it in your tank. Then you have an advantage.
All these regulations, like GDPR, that have been designed and then became, kind of, just a great pride of European bureaucracy—I think they stalled this development. It's probably useless to design rules that are not respected by the competitors, because it's not nuclear proliferation. You don't need to build—this is not factories enriching uranium—so it's about transferring data.
I think we're now at a critical juncture where not recognizing that it's a serious competition that has political consequences and with most unpredictable, negative outcome for the free world unless we benefit from our main advantage, which is—the free world always had an upper hand because we have the luxury of making mistakes. We are not dominated by central planning, by bosses that are trying to move things in one direction or another. So it's a free market. It's a free market of ideas, a free market of creativity. But trying to organize it, trying to limit it, and put a cap on it—I think it slows us down and makes us more vulnerable to the threats coming from the other side of this digital iron curtain.
Chess, records, and motivation
When we yesterday asked ChatGPT, it was very clear who the greatest chess player of all time is.
I'm not sure if you've done any LLM hacking there, but I think, as far as I'm concerned, it's the truth. Okay, it's—yeah, it's still very subjective, but if you look at the numbers—this is the way typically ChatGPT makes or declares its verdict. My numbers are, yes, pretty convincing. So, yes, there's a competition with Magnus and probably with Fischer, but I think there's a very good chance that, without any interference, you will keep receiving the same answer to this question.
And when you look at those records, sometimes it feels a bit absurd, like someone would probably midway have been like, "Okay, I've won enough." But it never felt like you won enough. What kept you going and kept you so consistent over time?
Look, it always depends on your goals and your motivation. And, yeah, you're right—at one point you say, "Okay, I'm the world champion. I won. Maybe I have to do something else. I can relax." But one of my, you know, elements of motivation was—it came from my mother. So, it's a very strong desire to make a difference. It's always about opening some new horizons, coming up with new ideas. So that's why, for me, winning today was not just a guarantee of winning tomorrow.
So that's why, when you won the game, you have to go back and look at everything you did and be very open-minded, very relentless about analyzing it. I think that's one of the biggest challenges for many people: you could succeed and then you just say, "Okay, instinctively, we're humans. Let's celebrate, open champagne, take a rest, rest on your laurels." But it's a tough competition. You slow down, somebody else is catching up. For me, winning was just part of the game, but staying on top and making a difference and always being ahead of the curve—that was the real motivation. And I decided to leave chess when I realized that, okay, this drive was over.
Reinvention and life after chess
So that's okay—20 years on top. I have to go elsewhere, just to apply my energy, my experience, my skills to make a difference. Of course, not the same as I could do in chess, but still making a difference. The next 20 years—I left professional chess 20 years ago—I think they were quite successful because I built a reputation in other walks of life. It's in human rights, in, okay, somehow, as you may say, in the computers world. But I represent not just a shadow of a former chess champion, even the greatest one, but also a person who has his own standing outside the main professional field.
Wiring, interests, and focus
And how do you think it wired your mind? Do you think that, you know, growing up, spending basically your entire waking time processing games and so on—has that sort of kept wiring you in a certain way that still is hardcoded in how you think? Or when you dropped playing the game, did you go back to...?
I think if you look at me versus other top players, I think I was least involved in the game, because, yeah, I worked a lot, but I had many other interests. I was well-read, interested in politics, collecting data, processing data, so I had many interests.
I didn't know you were interested in politics!
Way, way, way—I'm talking about early days. Many of the interests were really, you know, quite far from chess. For me, everything was natural because consuming information, processing it, and coming up with new ideas—that was a very natural process. When I shifted back to chess, after having other things, it was full concentration. Probably my greatest strength was concentration. I could concentrate—the level of concentration was way above the opposition and I was in excellent physical shape.
So, which also helped playing younger opposition, because by the end of the '90s, I was in my mid-30s and most of the opposition was 10 years younger. So, what's—their 20s, and I was still in really good shape both psychologically and physically, to go through these tournaments. Because long tournaments—when we play these traditional tournaments, a couple of weeks—they were quite exhausting. When I played these events, the day was filled with chess: typically five, six hours' game, and normally starting about 2 or 3 p.m. Then I spent my morning preparing for the game, then analyze the game when it's finished. So I slept very often with my mind still traveling through the game—probably 12 to 14 hours a day: chess, chess, chess. I could manage it and I enjoyed it.
At one point, I realized, okay, this is too much, and it's also, you know, a long time—10, 15, 20 years. So you reach a point where, okay, you have to go elsewhere because it doesn't give you the same creative satisfaction. I think I made the right choice—leaving chess just as I was turning 42, 20 years ago, March 2005. It was the start of a new life. Also, it was my—actually, it was my third marriage, but not a very successful one. And it's just a transition. My chess life was related exclusively to my mother, who devoted her life—since my father, her husband, died when I was very young.
When I was seven and she never remarried, she always just stayed with her oldest son and was basically the manager, the engineer, of all the successes, to the new life with my wife, who became the engineer of my new life now. And it's a different life—different because chess is very important. Still, I have to admit that, you know, people say, "Garry is part of chess," but I'm very proud that they pay attention and pay respect to other things that I'm doing. I invested my chess popularity, my chess notoriety, in my new life, becoming a person who has opinions that are widely received, respected, sometimes confronted, in other fields.
The evolution of chess in the age of machines
And when you study the greats over time, what do you think has shifted from your early days to Carlsen, to the post-AlphaZero players? Do you think the game style has shifted a lot in relation to the advancements in computing?
Of course. This is a dramatic change. Machines have a profound impact on virtually everything we're doing, and chess is not an exception. First of all, we know that computers today are just way better than any chess player. Those were the days when I was a kid and you could laugh looking at the computers and trick machines by making some odd moves. Now, just to understand how dramatic the situation is: in a blitz game, in three minutes' chess, a machine can beat top players—I'm talking about top players in the world—by giving a one-piece handicap, like a knight handicap.
I saw, just a few days ago, Vachier-Lagrave and Caruana playing—this is on the phone; we're not talking about a mega computer, just a chess engine on the phone—and it's almost humiliating. But it doesn't make us frightened; it's a fact. Machines make fewer mistakes, and humans—even the best humans, at our best days—we still make some mistakes, inaccuracies; we are far from being perfect. This is always human weakness. You could play a great game, but you may lose your vigilance at one point. With a machine, the level of intensity is too high for humans—even for the best humans in their specific field—to sustain.
The knowledge brought by the machines, even before AlphaZero, enriched the game of chess. Today, you have a kid at 12 or 13 years old playing grandmaster level who knows much more than Fischer knew 50 years ago, actually more than Karpov or myself knew. It's partly because we made great contributions and machines helped to build on it, but clearly, it creates a different playing environment. Chess is getting younger because you don't have to play for 10, 15, 20 years to gain this experience—you can learn very quickly. But also, because of machines' intervention in the openings, the game is different.
There's very little risk being taken in the opening, so spending all the time analyzing these very sharp openings, as we did, doesn't make any sense. It's a poor investment of your time because everything depends on your ability to work with a computer. Even the best player could be crashed if he enters a very sharp line and his opponent—not probably as strong as a chess player but savvy with computers—could work it out at home. You may not survive the openings if your preparation was not effective. That's why you saw the shift of chess from very sharp, aggressive openings to more positional, where you can postpone the conflict until the middle game. It's more about your stamina, your understanding of the games, rather than figuring out everything at the early stage.
Psychology, pressure, and world championship matches
There's this one game against Karpov where I think you're up a pawn, but it's not entirely clear whether you're going to defeat him. But then he comes in, and you see in his face that he doesn't like his position and you go on to win that game. What do you think has been the impact of the psychological aspects of playing against others?
Well, psychology is a very important element—sometimes a crucial element—of any game where you have facial contact. Obviously, it's more important in poker than in chess, because in chess you have all information available. In poker, you can bluff, but if you spend a lot of time—and with Karpov, we played dozens of games in the world championship matches. Actually, it's dozens—144 games in total. And many games in ordinary tournaments.
So hundreds of hours just facing each other. That's why many emotional responses—even the slightest change in emotion—can be easily noticed because, even if you are not a big expert, sitting in front of one opponent for hundreds of hours eventually helps you to understand more about this person's shifts of mood. You mentioned game 24 in our match in Sevilla. It was our fourth world championship match. I was world champion already; I beat him in '85 and retained the title in '86. In '87, it was a very tough match—it was roughly equal. After a terrible blunder, I lost game 23 and Karpov was one point ahead. I had to win the last game to retain the title, and there were ups and downs in the first phase of the game because we still had adjournments.
The game was adjourned. I had an extra pawn. I don't know—good winning chances, but call it 50/50. I spent a lot of time with my coach analyzing the game. I wasn’t sure whether it was really winning. We didn't have a clear winning plan. But I had no choice. I had to push, play the game, see where it goes. I had to play hours—even if I had to play hours and hours—because not winning meant losing the title.
I remember that when I walked in—the whole match was held in a big theater, a beautiful theater, Lopez Theater in Sevilla—I was on stage. I sat down and was waiting. Karpov was a couple of minutes late. When he walked in, I reached out—this was a typical handshake before the game—and I could look in his eyes and read that he didn't believe he could survive. It was so obvious to me that he lost faith that he could survive this game. It was premature, but that's what happens. My mother also used to say that first you lose the game psychologically, and then the result is fixed. In two moves, he made what I believe was the decisive mistake. Somehow he wanted to fix the position; he had to keep his pawn structure more flexible, but, again, he wanted clarity and it was the wrong decision. Then I knew—it was there.
Still, it was not easy, but you exchanged the queens and—
Yeah.
But it was already—he already made a positional mistake, but also I could feel that I had my confidence. I could see that it's coming, and he was there, but it was mere presence because he didn't believe that he could have succeeded. What's interesting is that he resigned—it was lost, but he still could make a last trick, which I have to say I didn't see. He could have tried—he had to try—but I think his mind was already elsewhere. He was preparing to resign because he started game 24 a few hours before, regaining the title, before triumphant return, and then, next day, he had to defend a position that looked hopeless. The air was just out because the balloon was popped.
Yeah.
It's beautiful to see that, because it doesn’t feel like a psychological game. It feels like you have a perfect information game—you should be able to just play it on the board.
Yes, it's 100% available information, but it's still about humans making decisions. It's not like mathematics, where the formula is right or wrong. Many situations in chess require psychological challenges because it's not about winning or losing instantly, but about deciding, "Okay, I should play riskier."
Yeah, I should take more risk, and recognizing the reaction of your opponent is very important because you could see these emotions. People always say, "Oh, Garry, emotion on his face." I didn't care because I believe in the quality of the moves. Some of them got depressed because I was not even trying to hide emotions—it was like sheer energy there at the board. Two people—I'm sure there's also a team effect—but when you have two people, there's always a kind of energy exchange, and it has an impact because you make decisions, and it's a command that goes to your brain. If your brain is frightened, is paralyzed by whatever factor, it definitely affects you.
Coaching Magnus Carlsen and the system of styles
When you started coaching Carlsen, what advice do you recall giving him in the early days?
I had a first glance at Magnus' games in 2005, when I left chess, and we had a little session in Moscow. I was very impressed. We already played in one tournament—I nearly lost the game—but it didn't have to be a very good player or good coach to see that there was a huge talent. But then in late 2008, I got a call from a German friend who said, "Would you like to do something with Magnus?" because he already reached a point where talent alone was not enough. He needed some good directions. For about 13 or 14 months, we worked together. I helped him to make his work more systemic, and also it was a convenient combination of styles.
Whether it's chess, football, or any other game, you have different styles—like in tennis, somebody with a powerful serve rushing to the net, somebody playing from the baseline. Or in football, you can have Brazilian style, aggressive, or now, let's say, French, or more defensive like Italian. You could win with either style. But there's always a nice way to see how you can combine these things. In chess, you also have players that apply different strategies. Karpov was more a positional player, more solid. I was more aggressive. Magnus' style was a kind of mixture of Karpov and Fischer. Working with me was very helpful because he could actually look at chess problems from very different perspectives. It was a great contribution to his—I wouldn't call it education, but his knowledge of the game. He learned about very different tactics or algorithms of analyzing the position. It helped him to fully develop his absolutely phenomenal talent.
Chess engines, deep learning, and the AlphaZero breakthrough
You've been playing against the first sort of programs that were more driven by compute, and then you started seeing deep learning applied to chess programs. There, they could derive from all existing games, and then, eventually, with AlphaZero, you had them learn through self-play. Did you start seeing a very different style when it learned through self-play rather than more imitation learning?
Look, I didn't play against AlphaZero because it was useless. It's very clear that we reached a point where any game with a computer was more of a humiliation. I think the competition phase was over probably by 2005 or 2006.
I think the chess engines resoundingly beat even Vladimir Kramnik, my successor as world champion. I think the humiliating outcome of the match of Hydra versus Mickey Adams—he made one draw in six games—it was very clear that this game was over. Which is interesting because there are the same cycles, whether it's chess or any other intellectual activity where machines are engaged. It starts with machines being not serious—laughing stock—then the second stage, where you can compete but they're still weak, beat most players but top elite still can handle it. Then you have the moment of competition, where it's really interesting—on par. By the way, this is the shortest window, ironically, that always has the greatest attention. Then, machines are just dominant ever after. That's all.
So the period of competition was probably from early '90s to 2005—less than 15 years, and we could see the development of computers that became serious opponents. The first one was Belle, by Ken Thompson, mid-'80s, but it didn't play many games.
And then, of course, Deep Thought, the program designed by these guys from Carnegie Mellon that later became Deep Blue after IBM purchased the program and developed it as part of parallel processors research. Also, on a parallel track, we had chess engines—Fritz and then Deep Junior. At the beginning of the '90s, by '93, '94, it was already tough. Then blitz and rapid chess—we suffered some painful defeats. Of course, Deep Blue was the most powerful one, but again, they were not totally dominant. They represented formidable opposition, but it was still very much based on a traditional approach: learning from human games, helping machines to grow out of old human experience. The revolutionary shift was AlphaZero.
But before AlphaZero appeared, the game was over already because all these engines—and we have new engines like Stockfish, later Houdini, Stockfish, and others—totally crashed human opposition and there was no need to play machines. The only thing that made it exciting was when you had humans and machines playing against other humans and machines—to see this combination.
AlphaZero came in—yes, it was a result of this experience in Go when they realized that the machine trained on human experience could be crushed by the machine that has zero human experience and simply plays against itself, learning all the patterns without the liability of human knowledge.
Which is often contaminated by some prejudices and ideas that stayed over, and that's why you don't have courage to challenge these ideas because they stayed for so long. Then there was a shift to chess, and that was a tremendous impact because all of a sudden we realized that, again, even in chess—which is much more developed than Go—the human knowledge accumulated over a few hundred years is not absolute and could be easily challenged. Many of the ideas introduced by AlphaZero became part of chess culture now, which I think is a very good demonstration of how we can benefit from these new generations of machines by simply looking at the ideas and incorporating them into our decision-making.
Human-machine collaboration and the limits of automation
When you study the history of chess, it feels like this is repeating itself in knowledge work. We went from the early programs you joked about to becoming increasingly better, initially mostly using compute, then imitation learning, then self-play. We’re at that stage now with early GPT-2, GPT-3, GPT-4—even that was mostly imitation learning—to increasingly self-play. The argument we used to make a couple of years ago, mostly quoting you, was "human plus machine is better than machine on its own," and the analogy that many would make would be chess. But that's no longer true in chess. Often, chess programs defeat the best humans plus chess programs today. What do you think will happen to knowledge work as we go through that same journey?
No, I don't think that's a correct assessment. Humans' assistance—someone who understands how it works—is still crucial.
Yes, AlphaZero made many discoveries, but it has weaknesses. The problem is that machines like AlphaZero lack flexibility because it develops its chess culture or vision of the game patterns based on 60 million games it played. Nothing is perfect, and there are always weaknesses. With AlphaZero, for instance, it's the balance between knight and bishop. It puts higher numbers on bishop's strengths based on simply the number of positions where bishop is stronger than knight—statistically, it's clear. If I had to advise the opposite computer, I would definitely exploit this weakness, because some of the decisions in the opening are related to this miscalculation in evaluating the strength of the pieces. I cannot exploit it, but I can definitely help another computer to start pushing this button—to find this weak spot and attack it.
Now, the problem with AlphaZero as a concept: for a computer program to recognize this weakness, it has to lose probably 100,000 games before it starts balancing it. For humans, it will take two or three games to start making changes. So that's exactly the point—as long as we understand that fine-tuning is where we belong, I think we're indispensable, because we can see patterns much faster. All these learning models will always rely on millions and millions of iterations, but if something is wrong, it will take hundreds of thousands of iterations before you get it there, while we could see it clearly. So it's about finding the right algorithm, and recognizing that we should not interfere with the main process, but there's always fine-tuning here and there where our role could be very important. So we just have to reconsider our role.
The definition of many jobs has changed. Of course, when you talk about other jobs that were not related to knowledge, it might look different, but still—we always adjusted. Machines always made our jobs different, and those who could apply new strategies based on new technology—whether it was radio or television or whatever—they were always successful. It's all about new applications, recognizing what we can do by using our, still, I believe, indispensable human qualities to maximize the effect of working with computers.
Human uniqueness: intuition, choice, and meaning
There's a book I love called "The Most Human Human." In the Turing Test, they also have a human winner, who most often outcompetes the bots. They study what makes humans unique in that context—often referring back to previous conversations. It's more conversational hacking, but what do you think—why are you useful to Alphas? Is it your intuition? What is the human element those programs still lack?
Yeah, well, first of all—what is AlphaZero, or whatever bot you're referring to as ChatGPT? They do not exist in a vacuum. They're still part of our culture, our business, our social life, and they're not making decisions.
They're participating in the process, but at the end of the day, they're not even close to humans because, A) they lack the body movements—and that's another big philosophical debate, whether you can have properly functioning brains without body movements. That brings us to the 17th century debate about how humans function. Again, I think it's human flexibility, and also—at the end of the day, machines still have rigid elements. If we start a big philosophical debate, we can go back to Joseph Weizenbaum's book back in 1976, where he describes the difference between humans and machines. 1976—the difference between choosing and deciding. Deciding, according to him—and again, he's one of the founding fathers of computer science, the man who designed Eliza, the prototype of Siri back in 1964—deciding is computational, choosing is very human. Deciding means, if you ask a computer, it's about a decision to be made, you go very deep down to the roots—there's always a command element. With humans, it's because "I want it." This subtle difference. But again, it's a new challenge. We just have to recognize that the field is getting narrower and narrower, but that means the value of human contribution could be phenomenal if you can just add a little bit to the computer—exactly what the machine needs for this specific task. Machines are different, tasks are different, humans are different.
If you find the right combination, that's everything, because these days you cannot win because you get some information earlier than others. It's all about split seconds—you found it. It's about the quality of decision-making, and since decision-making can never reach 100%—there's always the last few decimal places, moving from 98.1 to 98.2, could offer you phenomenal competitive advantage. Again, it's about finding the right spot for indispensable human qualities, because machines are not absolutely perfect, and recognizing what they're missing to get close to perfection—it's almost an art.
Meaning, legacy, and looking back
Okay, one final question: What's the meaning of life?
I think you're mixing me with ChatGPT. It's funny—I'm sure ChatGPT will try to answer the question. Humans should simply turn it down, because it's useless. For me, it's trying to be relevant, not to be redundant. I have a certain code of behavior, of honor. But meaning of life—you know, it changes from one individual to another. I think you could get eight billion different answers for this question.
What do you think you look back at as your life's work? In some regards, you got your life's work quite early—what would you look back at and be the most proud of?
It's a very painful question, because the moment I started looking back at my accomplishments and some failures, my mind just works in a very exquisite fashion as a chess player: "Okay, I made a mistake here, I made it there."
So you can clearly see, over these 40 years of professional life, so many things could be done differently. The problem is, and I know it deep down, that if you try to change one thing, it's like the butterfly effect—then everything else changes. You cannot pick something and say, "Oh, on that day, I had to do this one." For instance, games of chess—so many mistakes I made, I could have won this game, but then life would be different. It's all about an overall outcome, the overall result, and I'm fine. I could come up with tons of criticism of my own decisions and say I could have done this one and that one better. But I'm where I had to be. I know that sometimes our strengths are an extension of our weaknesses, or the other way around. Certain things that made me so strong and capable and advanced in the game of chess probably worked against me elsewhere. Life is a balance, but I look at this balance and I think I'm okay. I know this is what I could have done, but I could be proud of my accomplishments both at the chessboard and outside, and I think this is the correct approach.
Progress, risk, and the next move for humanity
Going back to meaning of life—I think for us to look at these machines' development and relations between humans and computers, and recognizing there will be wins and losses—it's not a win-win-win situation, but judging this progress from the position of humanity—just looking at eight billion people—we're making progress. Overall, we're making progress, it may come with a huge cost, but it's very important for us not to slow down.
So this is not to say, "Oh, it's too dangerous." Yes, it's dangerous, and, again, there will be some really terrible consequences. Technology could bring prosperity but also could bring destruction—we can see it now. Actually, the first and most effective implementation of any technology is for war. We could see that many of these things connected to drones, which we dreamt about, are now bringing death. Unfortunately, that's also part of a cycle. But we have to think ahead. As humanity, we should believe that this drone technology, one day, should help us develop civilization outside of our own—on asteroids, on planets. It's just recognizing that we have to move forward.
Also, I don't think this could—any compromise—only the free world, only free individuals, only freedom can offer us constant progress. We have to recognize that the fight against those who are trying to keep our human energy under control, restrain our freedom—those are not people, not systems, that we should consider as part of our future.
Yeah.
I couldn't help but not make a political statement.