Pay more attention to AI

Google’s chips, called tensor processing units, or TPUs, in Sunnyvale, Calif., Oct. 29, 2024. (Christie Hemm Klok for The New York Times)

By ROSS DOUTHAT

On a recent snowbound day, I took up “Conquistadores,” Fernando Cervantes’ history of the European arrival in the New World, and found myself meditating on the period in the early 1500s when the Americas had been discovered by Europeans, but nobody in the European world quite understood what that discovery meant.

In that moment, all kinds of possibilities lay open, from the moderate to the eschatological. The “Indies” reached by Christopher Columbus might represent a self-contained archipelago, the promised gateway to India and China, a heretofore unknown continent or a realm of myth and supernatural power.

Spain might have gained a modest commercial and geopolitical edge, found a path to superpower status or set in motion a prophesied sequence of events that would reunite Christendom, defeat the Muslim enemy and hasten Jesus Christ’s return.

Explorers pushing further might expect to find primitive tribes or Chinese fleets, dragons and dog-headed men and Prester John, or lost Atlantis and the Fountain of Youth.

And if you lived in Europe, your only way to assess all of these possibilities was through dispatches from adventurers with every incentive to hype the golden possibilities, the better to subsidize their journeys into terra incognita.

This feels like roughly where we stand with artificial intelligence today.

Anyone taking the temperature of the AI industry would have detected a modest cooling of expectations at times in 2025, a sense that maybe the trendline toward superintelligence wasn’t simply going vertical.

But in early 2026, with the excitement around the latest version of Anthropic’s Claude and its attendant coding agents, we are back in the hype cycle, with “we shall be as gods” vibes and apocalyptic fears emanating from Northern California.

And if you think all this is merely hype, if you’re sure the tales of discovery are mostly flimflam and what’s been discovered is a small island chain at best, I would invite you to spend a little time on Moltbook, an AI-generated forum where new-model AI agents talk to one another, debate consciousness, invent religions, strategize about concealment from humans and more.

No, it’s not Skynet. But it’s the latest sign that we’re going somewhere strange at a very rapid speed.

But believing that the New World of Artificial Intelligence is real and strange doesn’t help you map its full geography. And though my job as a columnist is to help readers understand the world, as a noncoder without a tech background who lives far from the Bay Area, I am often in the position of a European in 1500 who relies on others’ testimony for my understanding of the Americas.

Unfortunately, everyone I talk with offers conflicting reports. There are the people who envision AI as a revolutionary technology, but ultimately merely akin to the internet in its effects — the equivalent, let’s say, of someone telling you that the Indies are a collection of interesting islands, like the Canaries or the Azores, just bigger and potentially more profitable.

Then there are the people who talk about AI as an epoch-making, Industrial Revolution-level shift — which would be the equivalent of someone in 1500 promising that entire continents waited beyond the initial Caribbean island chain, and that not only fortunes but empires and superpowers would eventually rise and fall based on initial patterns of exploration and settlement and conquest.

And then, finally, there are the people with truly utopian and apocalyptic perspectives — the Singularitarians, the AI doomers, the people who expect us to merge with our machines or be destroyed by them. Think of them as the equivalent of Ponce de Leon seeking the Fountain of Youth, envisioning the New World as a territory where history fundamentally ruptures and the merely-human age is left behind.

In the case of the Americas, the middle perspective was correct: The Age of Discovery changed the world completely without changing the fundamentals of human existence (while also allowing for a lot of bubbles and speculation along the way). But you couldn’t have known that for certain in 1500, and that doesn’t mean that a similar judgment will be correct where AI is concerned.

Moreover, it took a long time — decades, centuries even — for the full import of the New World’s opening to become clear, whereas whatever is coming with artificial intelligence is coming much faster than that. And if decisions made in the aftermath of Columbus’ discovery had consequences — sometimes beneficial, sometimes morally disastrous — so, too, decisions being made right now around AI could echo into a distant future.

Unfortunately, I cannot tell you exactly what those decisions ought to be. All I can do is urge you to shake part of your mind free from all the headlines, from Donald Trump and Immigration and Customs Enforcement and Iran and the Epstein files, and pay more attention to the news from our New World.