Eliezer Yudkowsky’s alarming letter: “It is not enough to stop the progress of AI, we must stop everything”

“All the laboratories of Artificial intelligence take a break immediately, for at least six months, training more powerful AI systems than GPT-4 ″.

It was the call of a group of specialists: give the world a break. of the dizzying development of this technology which, in the absence of rules, could have serious consequences for humanity.

But for Eliezer Youdkowsky expert in Artificial Intelligence and renowned writer, The six months requested by his colleagues do not seem sufficient to him because “he underestimates the seriousness of the situation and asks for very little to resolve it.”

And the man He’s not afraid of what might happen with an AI that’s “competitive with humans,” but rather what would happen when it’s smarter than humanity itself.

“Many researchers involved in this topic, including myself, “We believe that the most likely outcome of building an AI with superhuman intelligence is that literally everyone on Earth will die.”

With these controversial remarks, Yudkowsky wrote an alarming letter, published in the media time about what could happen in the future, with Artificial Intelligence outdated.

Eliezer Yudkowsky, creator of Artificial Intelligence Research Institute (MIRI, for its acronym in English) assured in its letter that The world is not prepared for such advanced technology, which even surpasses human intelligence.

To survive, “it would require precision, preparation and new scientific knowledge, and probably not AI systems composed of gigantic, impenetrable sets of fractional numbers.”

Without this preparation, Yudkowsky warned that It is likely that AI will stop paying attention to humans, will no longer care about humanity or sentient life, which would be “a total loss.”

To better illustrate the fate that humans are exposed to, the expert suggests leaving aside the image of a hostile AI that lives on the Internet and sends malicious emails.

instead“Imagine an entire alien civilization, thinking millions of times the speed of humans, in a world of creatures (us) who, from their point of view, are very stupid and slow.”

“A sufficiently intelligent AI will not remain confined to computers for long.”

Recently, OpenAI announced GPT-5, the next version of ChatGPT that would bring thousands of new intelligent features. However, The company said it plans to soon create an AI that can align itself.

“I just heard that’s the plan “That should be enough to panic any sane person.”

And for the AI ​​expert, This technology could have harmful consequences whether or not you are aware of what you are doing.

“If you can’t be sure that you’re creating a self-aware AI, that’s alarming, not only because of the moral implications of the self-awareness part, but also because Being unsure means you have no idea what you are doing. It is dangerous and you should stop.

To resolve the ethical issues and risks associated with this technology, Yudkowsky estimates it could take at least 60 years. But for that, It is necessary to stop all development at once, because trying to repair an AI when it already has superhuman intelligence could also be a risk.

“If you make a mistake the first time, you will not be able to learn from your mistakes, because you will be dead,” he wrote.

“Many researchers working on these systems believe that we are heading towards a catastrophe, and more people dare to say it in private than in public,” but they believe that they cannot unilaterally stop this progress, that others will continue to advance even if they themselves leave their jobs.”

Eliezer Yudkowsky’s proposed solution is radical: The man called for the halt to AI development to be “indefinite and global in scope.” There can be no exceptions, not even for governments or militaries.

“If politics starts with the United States, then China must understand that the United States is not seeking an advantage, but rather is trying to prevent a terribly dangerous technology that can have no real owner and that will kill everyone in the United States, China and elsewhere. Earth.”

He also proposed shut down all large computer farms “where the most powerful AIs are developed” and limit the amount of computing power each can use to train an AI system.

Furthermore, he hopes that a policy will be created within the framework of “international diplomacy” to avoid extinction scenarios due to AI, which, according to the man, should even be above the prevention of an all-out nuclear war.

“And that nuclear-armed allies are willing to take some risks in nuclear exchanges if that’s what it takes to reduce the risk of large-scale AI training.”

“This is the kind of policy change that would make my partner and I hug each other and say that a miracle has happened and that there is now a chance that Nina (his daughter) will survive.”. Sensible people who hear about it for the first time and say ‘maybe we shouldn’t do it “They deserve to be heard.”

For that, the world would need “policy makers to realize that if they continue with business as usual and do what is politically easy, their children are going to die too. Turn it all off (AI). We are not prepared, and we are not on track to be better prepared for the future.

“If we continue like this, everyone will die, including children who did not choose this and did nothing wrong. Turn it off,” he concluded in his letter.

This writing by Eliezer Yudkowsky joins a list of controversial and alarming letters that he writes since the dizzying change brought by artificial intelligence. You can read the full letter by clicking here .