When Garry Kasparov resigned Game 6 against Deep Blue in May 1997, the reaction was swift and existential. Newsweek's cover declared it "The Brain's Last Stand." Commentators spoke of humanity's dethronement. The greatest chess player alive had been beaten by a machine, and surely this meant something profound about our place in the universe.

It did mean something profound—just not what anyone thought at the time.

The wrong lesson

The 1997 panic centered on the idea that Deep Blue's victory represented the beginning of machine superiority in domains requiring intelligence. Chess was the canonical test of genius, the game of grandmasters and prodigies. If a computer could beat Kasparov, the reasoning went, then human cognitive supremacy was finished.

But Deep Blue was not intelligent in any meaningful sense. It was a brute-force calculator, evaluating roughly 200 million positions per second through hand-coded rules written by chess experts. It had no understanding of chess, no ability to learn, no capacity to apply its skills to anything else. It could not play checkers. It could not hold a conversation. It was, in essence, a very fast lookup table with excellent heuristics.

The real lesson was subtler and more important: machines do not need to be intelligent to outperform humans at specific tasks. They need only to be fast, tireless, and well-designed for the problem at hand. This distinction matters enormously, and we keep forgetting it.

The pattern repeats

Every subsequent AI milestone has triggered the same misreading. When IBM's Watson won at Jeopardy in 2011, headlines proclaimed the arrival of thinking machines. Watson turned out to be mediocre at actual medical diagnosis and business applications—good at pattern-matching trivia, poor at reasoning. When AlphaGo defeated Lee Sedol in 2016, the discourse again turned apocalyptic. Go was supposed to be the game computers could never master, too intuitive, too dependent on human judgment. AlphaGo's victory felt like a threshold crossed.

Yet AlphaGo, for all its elegance, was still a narrow system. It could not play poker. It could not write a sentence. It represented a genuine advance in machine learning techniques, but the existential framing obscured the technical reality.

Today's large language models have triggered the most intense version of this cycle yet. They can write essays, debug code, and pass professional exams. The temptation to see them as generally intelligent—or nearly so—is powerful. But they remain, in fundamental ways, pattern-completion engines. They predict likely next tokens based on statistical regularities in training data. They do not understand what they are saying in the way humans understand language. They hallucinate confidently. They cannot reliably count letters in words.

What we should actually worry about

The Kasparov match taught us that machines can dominate narrow domains without possessing general intelligence. This is precisely what makes them economically disruptive. You do not need artificial general intelligence to automate a call center, generate marketing copy, or review legal documents. You need systems that are good enough at specific tasks to be cheaper than humans.

The threat was never that machines would become smarter than us in some cosmic sense. The threat was always that they would become competent enough at particular jobs to reshape labor markets, concentrate economic power, and alter the information environment—all without achieving anything resembling consciousness or understanding.

Kasparov himself came to understand this better than most. He spent years advocating for "advanced chess," where humans and computers collaborate, and arguing that the interesting question was not whether machines could beat us but how we could work alongside them. His framing was more useful than the doomsaying.

Our take

We are still having the wrong conversation about artificial intelligence. We debate whether ChatGPT is sentient while companies quietly use it to eliminate entry-level writing jobs. We worry about superintelligent takeover scenarios while algorithmic systems shape what billions of people see and believe every day. Deep Blue's victory was a warning, but not the one we heard. The machines do not need to think to change everything. They only need to be useful.