Blogs
Why Computers Didn't Ruin Chess (And Why They Might Yet)

Why Computers Didn't Ruin Chess (And Why They Might Yet)

Avatar of Saulimedes
Saulimedes
| 0

In 1997, IBM's Deep Blue defeated Garry Kasparov, the world chess champion, in a six-game match. Newsweek's cover declared it "The Brain's Last Stand." Commentators announced the death of chess. Why would anyone play a game that machines had solved?

Twenty-eight years later, more people play chess than at any point in human history. Viewership is up. Prize funds are up. A Netflix show about chess became one of the most-watched series of 2020. Magnus Carlsen has more Instagram followers than most professional athletes.

What happened?

The short answer: everyone was wrong about what chess is for.

The slightly longer answer is more interesting, and involves thinking about the difference between "solving" a game and "experiencing" it.

THE DEATH THAT WASN'T

The prediction made sense at the time. Chess had been positioned for centuries as the ultimate test of human intelligence. "Chess is the gymnasium of the mind," said Pascal. "Chess is mental torture," said Kasparov himself. The game's prestige rested on its difficulty, its depth, its resistance to mastery.

If a machine could beat the best human, what was the point?

This logic assumed that chess's value lay in being unsolved. That we played to find the objectively best moves. That the goal was optimization.

But watch how people actually engage with chess and you'll notice something: almost nobody cares about optimal play. They care about drama. They care about stories. They care about watching humans struggle against other humans, making brave decisions under pressure, sometimes brilliantly and sometimes catastrophically.

Deep Blue's victory didn't eliminate any of that. It just added a new character to the narrative: the machine lurking in the background, silently evaluating every move, an omniscient narrator we can consult after the fact.

THE CENTAUR INTERLUDE

For a brief period, something genuinely strange happened.

In the early 2000s, a format called "Advanced Chess" or "Freestyle Chess" emerged. Human players could consult computer engines during the game. The idea was to combine human intuition with machine calculation.

The results were fascinating. In 2005, a Freestyle tournament was won not by a grandmaster with a supercomputer, but by two amateur players using three ordinary computers and superior "process." They knew how to ask their machines the right questions, when to trust them, when to override them.

Kasparov, who had pioneered the format, noted that "a weak human + machine + better process was superior to a strong computer alone and, remarkably, superior to a strong human + machine + inferior process."

This was the "centaur" era. Half-human, half-machine hybrids playing the strongest chess ever seen.

Then it ended.

By the mid-2010s, pure AI surpassed centaurs. The machines no longer needed human guidance. The humans were just adding noise.

But here's the thing: nobody watches computer-versus-computer chess. It exists. You can find it. It's technically the highest level of chess ever played. And it's completely unwatchable. Two optimizers grinding toward drawn endgames, making moves no human would consider for reasons no human can articulate.

The centaurs are gone. The pure machines are boring. What remains, thriving more than ever, is humans playing humans, with machines commenting from the sidelines like omniscient Greek gods, occasionally noting that the mortal has blundered.

THE ENGINE OVERLAY

Modern chess broadcasting looks like this: two humans play, while a sidebar displays the computer's evaluation. A little bar shows who's "actually" winning according to the machine. Numbers like "+1.3" or "-0.7" float next to every position.

The humans don't see this during the game. The audience does.

This creates a peculiar dramatic tension. You watch a grandmaster think for twenty minutes and play a move. The engine bar jumps. The commentators gasp. The grandmaster, still ignorant, looks pleased with themselves. The audience knows something the protagonist doesn't.

This is Greek tragedy structure. We know Oedipus is marrying his mother. He doesn't. The dramatic irony creates tension impossible in a world where the protagonist knows everything.

Chess, accidentally, evolved a storytelling format that depends on human limitation. If the players could see the engine evaluation, the drama would collapse. They'd just play the recommended move. The game would become karaoke.

Instead, we watch humans navigate uncertainty while the gods watch from Olympus, knowing the true score.

THE ALIEN TEACHER

Here's where it gets genuinely weird.

In 2017, DeepMind released AlphaZero, a chess engine trained through self-play rather than human games. It learned chess from scratch, knowing only the rules, and within hours was beating the strongest traditional engines.

More interesting than its victories was its style. AlphaZero played moves that human grandmasters found beautiful. Long-term piece sacrifices. Positional play that accepted short-term disadvantage for structural compensation humans couldn't quite articulate.

Chess players began studying AlphaZero games not just for tactical ideas but for strategic concepts. The machine, having learned nothing from human history, had independently discovered principles that humans found aesthetically pleasing.

Why? One possibility: there's something objectively beautiful about good chess, and both humans and sufficiently advanced machines converge on it. Another possibility: we're pattern-matching machines ourselves, and AlphaZero's patterns happen to tickle the same neural circuits that respond to "elegance."

Either way, human grandmasters now spend hours studying alien intelligence, trying to reverse-engineer insights from a mind that doesn't think in concepts, doesn't have intentions, and cannot explain itself.

Magnus Carlsen, the current best human, has talked about deliberately playing "human moves" rather than "engine moves." The distinction is telling. Engine moves are objectively stronger but don't make sense to human cognition. Human moves are slightly inferior but flow from recognizable ideas.

Carlsen is intentionally handicapping himself to stay comprehensible. He's choosing narrative over optimization.

THE COMING GENERATION

Here's where the "might yet" comes in.

Current top players learned chess from humans, then later incorporated computer analysis. They have intuitions built from human games, supplemented by machine insights.

The next generation is different. Kids now learn chess from engines from the beginning. They don't have pre-engine intuitions. They haven't internalized "human" principles that engines have since overturned.

Early reports suggest these players make different kinds of mistakes than previous generations. They're better at tactical calculation but sometimes miss strategic ideas that older players find obvious. They've absorbed the alien's perspective more deeply.

Whether this produces better chess, worse chess, or just different chess, nobody knows yet. But it's possible that the "human" chess we've been watching, the kind that generates drama and narrative and beauty, is a transitional form.

Future chess might be played by humans who think more like machines than like their grandparents. The drama of watching humans struggle with uncertainty might fade as players develop intuitions that align more closely with optimal play.

At which point, maybe chess does become boring. Not because machines beat us, but because we became too much like them.

Or maybe not. Maybe there's no endpoint to this process, just endless evolution. Maybe each generation's chess is beautiful to itself and alien to its predecessors. Maybe beauty is always local, always contested, always in flux.

I don't know. Neither does anyone else.

WHAT CHESS IS FOR

The Kasparov-Deep Blue match was framed as a battle for human dignity. Man versus machine. Brain versus silicon.

That framing was wrong, but not in the way critics usually suggest.

The real question wasn't whether humans could beat machines. It was whether chess was about winning in the first place.

If chess is about winning, machines have won. Permanently and irreversibly. No human will ever beat a top engine again. The competition is over.

If chess is about something else - the experience of struggle, the narrative of conflict, the aesthetic pleasure of patterns emerging from rules, the social bonding of shared competition, the meditative absorption of deep concentration - then machines are irrelevant to its purpose.

The fact that chess thrived after 1997 suggests it was always about something else. The "mental gymnasium" framing was marketing, or self-flattery, or a useful fiction that motivated practice.

The real thing was always underneath, waiting for the fiction to be stripped away.

THE PARALLELS

I can't help noticing that this pattern appears elsewhere.

Photography was supposed to kill painting. Why paint a portrait when a camera could capture perfect likeness? Instead, painting evolved, finding purposes photography couldn't serve.

Recorded music was supposed to kill live performance. Instead, live concerts became more popular than ever, offering something recordings couldn't.

Calculators were supposed to make arithmetic education pointless. Instead, we discovered that understanding numbers matters for reasons beyond computation.

The pattern: when machines master the "objective" component of a human activity, we discover the activity was never really about that component. The thing we valued was hiding underneath, parasitic on the task but not identical to it.

Chess players thought they valued finding good moves. They actually valued the experience of searching for good moves, against opponents doing the same, with uncertain outcomes and emotional stakes.

Machines can find moves better. They cannot have the experience.

THE BUBBLE AND THE BASELINE

Here's something worth noticing: chess AI is almost thirty years old now. It's mature. Settled. We understand roughly what it can and can't do. The hype cycle completed long ago.

Current AI - the large language models, the image generators, the things everyone's excited or terrified about - is not there yet. Not even close.

We're in the bubble phase. The phase where every prediction is extreme, where companies promise artificial general intelligence next quarter, where critics warn of imminent extinction, where neither side has much evidence because the technology is too new to have a track record.

Chess engines spent decades being refined, understood, integrated into human practice. We learned what they're good for (tactical calculation, endgame tablebases, opening preparation) and what they're not (explaining ideas, teaching concepts, generating interest). The human-machine relationship evolved through trial and error over years.

Current AI models are, by comparison, surface-level. They can produce remarkably fluent text. They can pass tests. They can simulate expertise convincingly enough to fool casual observers. But the deep integration, the mature understanding of capabilities and limitations, the evolved human-machine workflows - none of that exists yet.

Chess gives us a preview of what that maturation might look like. Not replacement, but integration. Not competition, but collaboration. Not the death of human relevance, but the discovery of what human relevance actually meant all along.

Or maybe not. Maybe language and reasoning are different enough from chess that the pattern won't repeat. Maybe the current AI is genuinely different in kind, not just degree. Maybe the bubble will pop, or maybe it'll inflate forever.

I don't know. But I notice that the people most confident about AI's trajectory are usually the ones who haven't watched a similar technology mature over decades. Chess players have. They've seen the whole cycle: hype, crisis, adaptation, integration, new equilibrium.

That experience doesn't tell you what will happen with current AI. But it does suggest that confident predictions are probably wrong, that humans are more adaptable than they expect, and that the interesting questions aren't "will machines win?" but "what will we discover the activity was actually about?"

WHY NOW IS THE TIME

Here's the thing about living through a technological transition: it's the best possible time to learn.

When chess engines were new, the players who engaged with them early gained advantages that compounded for decades. They developed intuitions about when to trust the machine and when to trust themselves. They learned the new workflows before the workflows became mandatory.

We're in that window now for a dozen different fields. The tools are powerful enough to be useful, immature enough to be forgiving, and new enough that expertise hasn't ossified into credential requirements.

Twenty years from now, "knowing how to work with AI" will be like "knowing how to use a computer" - assumed, invisible, unremarkable. Right now it's a frontier skill.

The centaur era in chess was brief. The window where human+machine beat both human alone and machine alone lasted maybe a decade. Then it closed. The humans who learned centaur chess during that window still carry insights the next generation won't have.

Current AI might follow the same pattern. There might be a brief window where human+AI produces something neither can alone. Learning to operate in that window, before it closes, seems valuable.

Or the window might stay open. Or there might be no window at all. Or the whole thing might collapse into another AI winter and we'll laugh about the chatbots like we laugh about 1990s virtual reality.

I don't know. But I know that learning is cheap, ignorance is expensive, and certainty is foolish.

The people who learned chess deeply before engines, then adapted to engines, then watched the next generation grow up with engines - they've seen something most people haven't. They understand technological transition from the inside.

That understanding is available to anyone willing to pay attention during the current transition. Not by predicting the future, but by staying engaged while the future unfolds.

This is a good time to learn things. Any things. The connections between fields that seemed separate are becoming visible. The tools for exploring are more accessible than ever. The old credentials matter less than they used to. The new credentials haven't solidified yet.

Windows like this don't stay open forever.

FINAL POSITION

Deep Blue is in a museum now. The specific hardware that beat Kasparov is a historical artifact, less powerful than your phone.

Kasparov is still alive, still commenting on chess, still occasionally playing. He lost the battle and won the war. The game he championed is more popular than ever.

The machines that were supposed to end chess have become its supporting infrastructure. They analyze games, train players, evaluate positions for broadcasts, generate puzzles, catch cheaters. They're the stage crew, not the stars.

Whether this equilibrium holds forever, I don't know. The next generation might play a chess so alien that current fans don't recognize it. The drama might drain away as humans converge toward machine play.

Or the drama might evolve, finding new forms we can't anticipate. Maybe the interest will shift from "who found the best move" to "who stayed human longest" or "who produced the most beautiful loss" or something we don't have words for yet.

The only prediction I'm confident in: the confident predictions will be wrong. They always are.

Chess survived its supposed death. It might yet survive its actual transformation into something unrecognizable. Or it might not.

Sources: