1. A NEW ERA

We have entered a new era.

Artificial intelligence has had — and will continue to have — profound repercussions across every domain, without exception, of human activity.

There will always be voices seeking to downplay the AI revolution. There is a market for the Gary Marcuses of this world, who take a perverse pleasure in pointing out the shortcomings of current AI systems built on large language model (LLM) technology. Strangely, when three months later the AI succeeds where it had failed, you never hear Marcus or his ilk recant. They move the goalposts, question the tests, hunt for new deficiencies. Meanwhile, frontier AI models keep shattering benchmark records. While they still lag behind in certain areas — planning and reasoning, in particular — they are evolving very rapidly. Even if progress is jagged, the overall trajectory remains undeniable, pointing toward a powerful and profoundly transformative technology. It is possible that in the coming years a new scientific paradigm will emerge outside of or alongside LLMs; it is equally possible that hyperscaling alone will suffice to achieve artificial general intelligence (AGI), or even artificial superintelligence — an intelligence superior to ours in every respect.

In an essay I published in January, Planned Human Obsolescence, I discussed the significant risks entailed by the pursuit of an AI that would match or surpass human capabilities. What we are witnessing right now is a very dangerous arms race that could lead to an extreme concentration of wealth and power, or even to the extinction of the human species. These scenarios are not merely the ravings of doomsayers; they are considered probable by leading figures in economics and computer science, including Nobel laureate in economics Joseph Stiglitz, Nobel laureate in physics Geoffrey Hinton, and Turing Award recipient Yoshua Bengio.

I take these risks very seriously. That said, my pro-safety stance should not be read as an anti-AI position. We will never again live in a world without advanced AI. One would search in vain for a toggle switch: AI / no AI. Technological advances have always been irreversible, and artificial intelligence will be no exception. All we can do is push collectively toward safe, public-serving AI while fully reckoning with the implications of the revolution in our respective fields. Does being pro-safety in the world of aviation make one “anti-airplane”? I fully intend to board the supersonic aircraft that is artificial intelligence — and at the same time, without contradiction, I insist that the technology be regulated and safe.