Who knows?

In a long and thoughtful essay inspired by his work as a teacher, Julian Waldner also touches on the Turing test and its limitations.

What is fascinating about Turing’s Test is not it’s astuteness as a test for machine intelligence; I am not alone in finding it unsatisfying. The philosopher John Searle famously threw down the gauntlet against Turing’s Test with his Chinese Room thought experiment. Searle imagines a room with a man locked away inside. The man is fed statements in Chinese through a small slot in the locked door. Although the man speaks only English and does not understand a word of Chinese, he has access to a giant book of rules. This book of rules allows the man to infallibly translate the Chinese statements and produce completely flaunt Chinese responses to be fed back out of the room. This procedure allows the man to produce Chinese statements, without understanding a word of Chinese. Searle’s point is that Artificial Intelligence produces the correct response through the mindless manipulation of symbols, without understanding what those symbols mean.

Turing’s test has left something out by design. The Turing Test, like the scientific project as a whole, has simply bracketed the question of understanding or consciousness to focus instead on what can be measured: We can’t touch consciousness or see understanding, but we can distinguish between something that looks like the product of intelligence and something that doesn’t. The fascinating thing about the Turing Test is that it still functions, even seven decades later, as a definition of what Artificial Intelligence is. Whether we are talking about Google Maps giving you a route that works, Claude generating text, a Tesla navigating through a busy street, your iPhone recognizing your face, or Spotify recommending songs—the goal is always to produce the kind of outputs that intelligent minds would.

Alan Turing himself was a distinctive intellect. Even aside from being gay and suicidal, he was one of those British eccentrics who enrich the world's cultural life now and again. But it's fair to say he had blind spots as well. Left brain types in a number of fields have a tendency to reduce all mental processes to a matter of processing power. Ted Chiang's short story "Understand" depicts a man who is treated for anoxia with a drug that regenerates his neurons, and it winds up turning him superhuman. Entertaining, but dubious. The engineers and investors who have brought us LLMs often insist that if they're not engaging in actual thought now equal to or greater than humans, it's only a matter of time and technological development. But it might just be that robots are one thing and people are another.

As for students and learning, schools have been engaged in a large-scale industrial production of young minds for long enough for Charles Dickens to have noticed. In some ways that seems to have intensified. The mass of students are trained to complete set tasks at best. And doing more than that is hard, no question. No education system is perfect, and arguments about ours will no doubt continue. But at this historical juncture it's crucial to at least preserve individual thought as an ideal.