Patch #17: Cheat codes vs. the state of play
The Chess Game by Emile Vloors via Wikimedia Commons
Even after he lost to Deep Blue in a full length chess match for the first time in 1997, Gary Kasparov argued that while machines had advantages when it came to calculation, they lacked imagination. Twenty years later, reflecting on his experiences in Science, Kasparov clarified his position further. He pointed out that even the AlphaZero system — which learned from itself and adopted a “dynamic, open style” like his own — did not end “the historical role of chess as a laboratory of cognition.” Kasparov memorably argued for chess as a drosophila of reasoning: a way to better understand the mysteries of human thought. Despite the ability of machines to dominate the game, the ability of humans to learn from the machines meant the game’s relevance was, if anything, enhanced.
The evolution in Kasparov’s thinking mirrors a broader evolution in societal orientation toward technological tools. If we want to better understand the future of how we’ll collaborate with machines, we could do worse than looking at how we use them in game environments.
Chess, for example, remains a canary in the coal mine. Take the cheating scandal that rocked the chess world last September, when top player Magnus Carlsen resigned from a game after a single move. Carlsen believed his opponent Hans Niemann was cheating by using machine analysis to calculate the odds of his moves — a practice Niemann later admitted to in some online play.
Central to this debate is both the prevalence of cheating in online games and the rise of “thinking in bets” aided by machine analysis. As two-time women’s chess champion and author Jennifer Shahade pointed out in a Financial Times article on the subject last September, “competitors and their teams do not just seek out the best positions — they look for those in which their opponents are most likely to err.”1
These shifts acknowledge, as Kasparov did, that when it comes to raw compute, machines have considerable advantages that translate especially well to the rule-based constraints we find in games like chess, Go, and countless others. Yet instead of seeing chess, poker, and video games as “solved problems” riddled with machine-assisted cheating, their popularity continues to surge. Chess, for example, is booming in a way that it hasn’t since the 1970s. We could chalk this up the pandemic and the excellent Netflix series The Queen’s Gambit, but the trend shows no sign of ebbing. Why is that?
We might point to the increasingly lucrative ecosystem around some of these games or the financial and social incentives for individuals to play. We could also argue that the broader trends of isolation and tech-mediated engagement make gaming a more appealing option for those who seek out connection than more passive forms of interaction. Both of these arguments are frequently discussed. Less examined is the experiential allure of solving complex problems in real time — either in groups or individually. What if the experience of cognition gets, if anything, more engaging the more access to knowledge we have?
I recently finished reading the Ian M. Banks novel The Player of Games. The protagonist, Jernau Morat Gurgeh, is the best game-player in the future utopian society the Culture. In the novel, a drone induces a somewhat bored Gurgeh to cheat, which sets off a journey to a distant galaxy to participate in a less evolved society’s “Azad” tournament. Azad is a complex, multi-part game followed by all participants in the opposing Empire — with high-stakes consequences for winners and losers. Players gamble their body parts on games, and the tournament Gurgeh participates in even selects who can be emperor.
One memorable passage about the final match in the tournament lyrically describes the game between Gurgeh and his alien opponent, Nicosar, the would-be emperor of the Empire:
[Gurgeh] didn’t talk to Nicosar, but they conversed, they carried out the most exquisitely textured exchange of mood and feeling through those pieces which they moved and were moved by; a song, a dance, a perfect poem. People filled the game-room now, engrossed in the fabulously perplexing work taking shape before them; trying to read that poem, see deeper into this moving picture, listen to this symphony, touch this living sculpture, and so understand it.2
What I love about this description is its acknowledgement that the apotheosis of play is an intellectual dance between two humans that becomes the point in and of itself. I think this is partly what Kasparov is getting at in the ideas mentioned above: if we focus too much on the win/lose outcomes of machine-assisted performance, we miss the richer mysteries and nuance that exist in the state of play. Beyond the financial ramifications, this is what really offends those who love chess and poker about those who cheat by relying on machine intelligence during games to win.
Notably, Banks does not argue machines should be left behind entirely. For example, Gurgeh consults nightly with a super-intelligent “Mind” on his ship, road-testing strategies and possibilities to sharpen his play. To use a phrase sometimes suggested in the current vernacular, Gurgeh treats the ship’s Mind as a “sparring partner.” But in his most masterful game, he leaves the Mind and its suggestions behind. To quote the novel, “Even the ship couldn’t work out what was happening on that board.”
Perhaps this is overly romantic. Player of Games was written in 1988, when Kasparov’s defeat was still almost a decade away. The state of play may be appealing to those who achieve mastery, or at least desire to achieve mastery, but it’s certainly possible I’m overestimating its allure. And surely, outside the confines of games in use cases of the sort we find in work or life, people may just want to solve the problem and move on so they can spend more time doing what they enjoy. New studies like this one from MIT on professional writing and the much-discussed GitHub Copilot study point to eye-popping productivity gains to the tune of 40-50% and a raft of other benefits from those utilizing AI-based tools.3 If we focus purely on outcomes, it may well be that the machines have it.
But when it comes to complex challenges like making art or solving the hardest problems, we can neither dismiss the experience of the state of play as a consideration nor assume that odds-based decision-making will inherently triumph. You can run the numbers and ask for input, but if you want to make something that deeply engages other humans or breaks new ground, it might well be that at some point you need to leave pure reason behind.
With the creation of art, for example, we know that machines can make things that look like fully realized works, but it remains an open question whether they can produce something that truly moves audiences in the way of our greatest human practitioners. When recently asked a question in an interview about whether this might be possible, scifi author Neal Stephenson responded:
My theory is that when we experience art — whether it’s a video game or a Da Vinci painting or a movie — we’re taking in a huge number of micro decisions that were made by the artists for particular reasons. In that way, we’re communicating with those artists, and that is really important. Something generated by AI might seem comparable to something produced by a human, which is why people are so excited. But you’re not having that awareness of commuting with the creator. Remove that and it’s hollow and uninteresting.4
It’s not too much of a stretch to think of the human to human relationship that Stephenson suggests between creator and audience as a similar state of play as the relationship between two players in a game.
There is surely a role for both cheat codes and the state of play. In the midst of something of a discourse tsunami triggered by a new wave of AI tools, it’s sometimes easy to forget that human to human interactions remain a primary source of mystery and, I would argue, the most rewarding sources of engagement. While singularity enthusiasts suggest that the line between humans and machines will fall, I’m skeptical that a rationally optimized world would ever truly satisfy its participants. These tools will allow us to interrogate our approaches and solve increasingly complex problems, but they will never replace the “fabulously perplexing,” but ultimately rewarding pursuit W.B. Yeats called “the fascination of what’s difficult.”
This week’s recommendations:
Reading: ChatGPT is nothing like a human by Elizabeth Weil in New York Magazine
Listening: Not Your Muse by Celeste
Music credits for article audio:
Opening Theme: “Friendly Evil Gangsta Synth Hip Hop” by mesostic via Wikimedia Commons
Closing Theme: “Hopes” by Kevin MacLeod via Wikimedia Commons
“The Complexities of Cheating in Chess,” Jennifer Shahade, Financial Times, September 30th, 2022
Ian M. Banks, The Player of Games, Macmillan, 1988 p.349
I learned about these studies via Ethan Mollick.
“Metaverse creator Neal Stephenson on the Future of Virtual Reality,” Financial Times, February 23, 2022
Hey, I discussed this piece in my weekly review: https://novum.substack.com/p/weekly-retrospective-5
Thanks, Anton. Appreciate the mention and seeing your further thoughts on the matter.