🔒 An Unexpected AI Conversation (About Chess) — with my spouse JJ Lang

Amelia HrubyWelcome to the Off The Grid Clubhouse. Thank you for being a paid supporter of the show. Let's get off the grid and into this super secret episode. Hello, and welcome to the Off the Grid Clubhouse. This is our private podcast and newsletter for paid supporters of Off the Grid Leaving Social Media.

I'm your host, Amelia Hruby. And right now on the public feed, we are in the middle of a series all about AI. I've been talking to some really amazing people like Mel Mitchell Jackson and Casey Zabala about their relationships to AI, particularly through the lens of AI sobriety and AI and spirituality. And coming up this week, I'll be talking to Ayana Zaire Cotton about their relationship to AI, particularly as a black feminist creator thinking about the impact that AI data centers have on their local community in rural Virginia. I am really enjoying the series, and I hope that you are too.

And here in the clubhouse, I wanted to bring you a very special conversation about AI. So today, I have invited my spouse and partner, JJ, onto the show to talk about AI from a very, like, niche and unique perspective. JJ is a chess professional, as I like to put it, which means that JJ works professionally in the chess space, works for the US Chess Federation, and is also an advanced amateur player, maybe I'll say. They're not a professional chess player, but they do play at a very high level, and they work in chess. And in the space of chess, they've actually had AI and machine learning much more normalized for much longer than I think it has been in the broader culture.

And so I wanted to have a conversation with JJ about this for you, dear Clubhouse listeners. This is not something I would do on the main feed because it feels personal and private because it's literally my private life. JJ and I have conversations about AI and chess all the time in our home, and it feels vulnerable to bring it to the show. But also, I've just found them a really interesting interlocutor in this, And they're approaching AI from just a different perspective that I see in online business. So in this episode, you're gonna get to meet JJ, my spouse.

There's a little bit of a personal behind the scenes moment there. And then we're gonna talk about AI and chess and creativity. JJ is also a writer and a teacher, and so that's gonna come up a lot in their relationship to AI and to these LLMs and GPTs and ABCs, etcetera. So with all of that said, hi, JJ. Welcome to the podcast.

JJ LangLongtime listener, first time caller. Happy to be here.

Amelia HrubyI'm thrilled to have you. I love that you called yourself a long time listener because I know you don't like listen to the edited version of the show that much, but you definitely hear me record almost every episode in our tiny little house.

JJ LangYeah. Maybe long time hearer is more appropriate.

Amelia HrubyYes. Long time hearer, first time caller. I like that. Okay. So, JH, could you start by just telling Clubhouse listeners a little bit about your background in chess and in writing, and maybe how you first encountered chess engines and AI?

JJ LangSure. So hello, listeners. My name is JJ. They, them, pronouns, he, him is fine. And my background for this, we can say that I've been playing chess competitively since middle school, and it was a big part of my life in middle and high school.

And I kinda walked away with from it in college and during grad school, but got back into it as I was sort of transitioning out of grad school and got really into playing, had a lot of fun with it, and was doing some writing on a personal blog and eventually for a publication about chess and stumbled kinda backwards into a job where I get to write about chess all day. And that's been a really interesting experience to get to work in a field that I'm passionate about, and also one that, I don't know, you do what you love and you never get a day off in your life. Sometimes feels like my experience, but it it it but when I step back, it's very cool what I get to do. And for those who don't play chess, don't worry. I'm speaking to you.

Amelia HrubyMe included.

JJ LangAnd for those who want to learn how to play chess, remind me, and at the end, I can give some tips on where to start, because it's honestly the barrier to entry is not very high. The barrier to mastery is very high, but easy to learn, fun to play. But for longer than I've been playing chess, computers have been better at chess than the best humans in the world. So my entire experience in chess has been one where there exist computer programs. At first, they were kind of expensive.

At first, they were kind of localized. Now the best ones are free, and you can get them on your phone. But there were computer programs that could beat the best players in the world. And if you were playing a game of chess and you wanted to know what the best move was or whether a move was good or bad, you could ask a computer and it would tell you and it would tell you with more accuracy and precision than human on the planet could. In that way, I think there being this omnipresent AI claiming to know things and tell you things whenever you want to know them has been a part of my life for as long as I've been a chess player.

But it's also been very different. And specifically, what computers can do in chess is very limited. They can tell you the best chess move. But also and I think if we're okay with taking this little detour, I think it's worth talking about how the the difference between the old chess computers and the neural net revolution.

Amelia HrubyYeah. Let's get into it.

JJ LangOkay. Sweet.

Amelia HrubyI love being nerdy about this. Please tell us more.

JJ LangSo 1997 was when IBM's computer Deep Blue beat then world champion Gary Kasparov in AHS match. That was the turning point of, okay, the best computers are better than the best humans. But what Deep Blue was running was sophisticated algorithm on a very high powered processor. I'm not a computer person, so I probably used both those words wrong. But the way that I think about it is even though a computer beat a human, the program the computer was running was a program that was written by humans.

It was written to tell the computer how to evaluate positions. It was told, well, first that a win is better than a draw, a draw is better than a loss. But it was also told that a bishop is usually better than a knight unless it has these sorts of features on the board, and then the knight could be better, and an algorithm that's way more complicated than that. But then the computer is running that evaluating process for thousands, if not millions, of positions per second, predicting which ones are optimal from the current board state, and then running its algorithm on all of those positions to figure out which ones are going to produce the best state for the computer and the worst for the opponent. So even though the computer is making a judgment about chess games, it's only making that judgment because humans have told it how to judge.

If we told it something different, it would give different outputs. It's all about what the humans are telling the computer. What the computer can do is process really, really quickly. And that's what computer engines were for the nineties, for before the nineties, and for most of the February. And interestingly enough, the kind of chess they played didn't look anything like human chess.

Computer chess is a term that was mostly derogatory to talk about moves that lacked any sort of passion or creativity, but had some long forcing sequence behind it that would make it work. But this idea, this forceful connotation of making it work is what a lot of people think of. Clunky might be a good word for it. Very frustrating and hard to beat, but not very inspiring.

Amelia HrubyOkay. So computer chess is created by clunkers, and therefore, it feels clunky. And this reminds me, I think, of, you know, the research and writing I did about algorithms in my book. Right? Like, it seems to me like what you're saying is that originally chess engines or computer chess is run by algorithms because algorithms are written and programmed by people.

They're like a sequence of instructions given to a computer. And they can be highly complex, but like humans have authored them. So then I'm guessing you're gonna tell us that now things have changed, and there are chess engines that are not just running instructions from humans. Is that true?

JJ LangSo enter a little company called DeepMind.

Amelia HrubyAh, yes.

JJ LangAbout twenty years later, I think the for the findings of the research were announced in 2017, maybe the publication was 2018. But DeepMind announced that it had created, and Google was involved somehow, first AlphaGo, which was a program that could play the game Go, and then AlphaZero, which was a chess playing computer. And that these computers could beat the strongest traditional chess computers in the world. But they were neural networks who were not given an algorithm. The neural network was given instructions on what the rules of chess were and what the objective in a game of chess is, what counts as a win.

And it was given an algorithm on how to learn. And it would play itself about a million times at a very high speed over eight hours. And of those games, as it's playing those games, it's developing its own strategies, its own, if you will, algorithm of what's good or bad based on what does or doesn't work. And the result of that four hours, eight hours, if you wanna be flashy, or the result of that million games of trial and error, if you wanna be more accurate, was a computer that was not told by humans how to play chess, or what the strategy behind chess was, and developed its own strategy of how to play that allowed it to perform at a higher level than the strongest chess computers in the world despite having far less processing power. It was evaluating fewer positions, and it was evaluating them slower.

And what it was doing was playing chess in a way that would beat those computers anyways. And as an added bonus, it plays in a way that's recognizably human. It's speculative. It looks creative. Kasparov is quoted in a book about the neural net revolution as saying that not only was he thrilled to see these new engines that could beat his nemesis Deep Blue, but they could beat Deep Blue by playing like Kasparov.

Amelia HrubyYeah. I think this is so interesting and notable. I remember when I was researching and learning about AI companies that I read Parmi Olsen's book Supremacy, and that's where I learned that DeepMind started as a company building a chess engine. The founder of DeepMind, whose name I never remember, maybe you know it.

JJ LangStarts with a d. I can tell you that he was a chess prodigy. He was a junior champion in England. Was it like Demis?

Amelia HrubyYeah. Demis Hruby. I don't know how to pronounce this, but he started this company in 2010, which is five years before OpenAI was founded, I believe. And their first project was chess. And so this is one of the reasons I like talking to you about this because some of the origins of contemporary AI, which as you previewed, DeepMind was bought by Google.

Their programming or networks or whatever language I should use, like, now fuel or are the engine behind Gemini. And so, like, I think it's really interesting that chess for so long has been a seed of both algorithms and neural networks which become AI.

JJ LangTotally. And the way that it's revolutionized the chess space, you could really tell. So around 2018 when those first papers are coming out, the first players to really use the neural net engines and use them well were revolutionizing how the game was played from the early moves onward. That