The brake is under the accelerator

Editions:
🇮🇹 Italiano – 🇬🇧 English – 🇫🇷 Français – 🇦🇷Spanish – 🇭🇰 中文（繁體）

For years, the artificial-intelligence race followed a wonderfully simple commandment: faster, more powerful, and ahead of the fellow next door.

Every new model had to outperform the previous one. Every company had to prove that it was months, weeks or, preferably, fifteen minutes ahead of its competitors. Artificial intelligence became a technological Formula One in which almost everyone was interested in horsepower and remarkably few people seemed to be asking where the brakes were.

Until the drivers themselves began looking at the bend ahead.

Dario Amodei of Anthropic, Sam Altman of OpenAI and Elon Musk, founder of xAI, now converge — with different nuances and while running fiercely competing companies — on an idea that until recently might have sounded almost heretical: perhaps the technological frontier should advance a little more carefully.

There is a certain comedy to this.

It is rather like three gunpowder manufacturers rushing into the town square to announce that society really ought to have a serious conversation about the dangers of matches.

Yet beneath the irony lies an extremely serious problem.

Artificial intelligence is no longer advancing simply because scientists want to discover what machines can do. It is advancing because there are companies, enormous investments, markets, contracts, prestige and international competition. In that environment, voluntarily slowing down may mean watching your rival continue down the track.

That is the real difficulty.

The greatest danger need not be a machine waking up one morning in a foul mood and deciding to conquer humanity — splendid material for Hollywood — but something far more human and therefore perhaps more dangerous:

nobody wants to be the first to brake.

If one company spends an additional six months testing the safety of a system while its competitor uses those same six months to release something more powerful, the market may reward recklessness before responsibility.

An excellent recipe.

The fire brigade arrives after the building has burned down because someone first had to determine whether extinguishing it might hurt the share price.

That is why independent scrutiny matters. Oversight conducted exclusively by those who build, sell and profit from a technology inevitably leaves an awkward question sitting in the middle of the room:

Who watches the watchers?

None of this means that tomorrow morning we should unplug the computer, dust off the typewriter and declare the digital revolution cancelled.

Artificial intelligence offers extraordinary possibilities. The sensible challenge is not to stop its development, but to prevent speed from becoming the only criterion by which we decide where we are going.

Because behind every model there are societies.

There are jobs, schools, universities, journalists, doctors, judges, governments, armed forces, businesses and millions of citizens who will eventually live with the consequences of decisions now being made by an extraordinarily small number of people.

And those people do not own crystal balls either.

The fact that Amodei, Altman and Musk can agree that the speedometer deserves more attention is worth noticing precisely because they are competitors.

But listening to their warning does not mean handing them the whistle, the rulebook and the traffic police as well.

There must be independent oversight. There must be reasonable transparency. There must be outside research. There must be legal accountability. And above all, we may need to preserve something wonderfully old-fashioned:

time to think.

There is a delightful final paradox.

We have spent years building machines capable of thinking ever faster, only to discover that human beings may now need to do precisely the opposite.

Think a little more slowly.

Before pressing Enter again.