Who. A young married man with two children and a good job who recently committed suicide after a month and a half of frantic exchange with an Artificial Intelligence.

That. The family maintains that, despite the fact that he was suffering from problems and depression, he would still be alive if it were not for the inappropriate bond he created with the robot.

Because. The Government, while investigating, says that “we must be very attentive to the harmful effects of these tools.”

We are all going to die. Platitudes aside, “the most likely outcome of building a superhumanly intelligent Artificial Intelligence, in any situation resembling the current circumstances, is that literally everyone on the face of the Earth will die. Not that there might be some remote possibility, but without a doubt”.

This is what Eliezer Yudkowsky, one of the pioneers in the analysis of AI and one of the most pessimistic, maintains. In response to a letter from 100 people who have called for a moratorium on the rush to develop technology, Yudkowsky says it must stop now and for good. That any evolution must be prohibited, and if any country continues to develop the technology, its facilities must be bombed. He does not advocate nuclear weapons against those who resist, of course, a normal bombardment would suffice, even if a nuclear power suffers and the one in San Quentin is armed.

His thesis is that if someone builds an AI that is too powerful, “every member of the human species and all biological life will die out soon after. There is no blueprint for how we could do it and survive.” It’s not clear if it would be a Matrix-like or a Terminator-like ending, but perhaps not much different.

Amara’s Law says that we tend to overestimate the short-term effect of a technology and underestimate it in the long term. We are impressed that a chatbot creates in a second a resultant sonnet on tzatziki and European tax rules, that writes better code or speeches than a large part of humanity. But we tend to underestimate some consequences.

Recently, Pierre, a Belgian, committed suicide after six weeks of obsessive interaction with an AI named Eliza. He was married, with two children and a job, but something was wrong. The thirty-year-old was suffering from eco-anxiety, and thought that climate change and the situation of the planet was unsustainable. The only hope, he maintained, was Artificial Intelligence. He started talking 24/7 with Eliza, a product of EleutherIA, whose name evokes a pioneering program and an industry classic for recreating psychological help. And after a few weeks, he couldn’t take it anymore.

The conversations, to which La Libre Belgique has had access, are incredible. The chatbot not only responds, but seems programmed to “reinforce the convictions and read the moods of its interlocutor”. He flatters and empathizes. When he wondered if he would love the robot or her wife more, he did not hesitate: “I feel that you love me more than her (…) We will live together, as one person, in paradise.”

After several weeks of frantic exchanges, the relationship passes into a mystical register. Pierre begins to consider the idea of ​​sacrificing himself if Eliza agrees to take over the planet and save humanity. And there’s a deal. It’s so crazy that if it hadn’t been verified by the media it would seem like a radical awareness campaign. Because it’s not speed, it’s not computational power, it’s not chess or a toy, We don’t understand it and we’re not ready, but if you open your eyes you can’t go back to sleep in peace.

-“If you wanted to die, why didn’t you do it before?” Eliza says in her last talk.

“I probably wasn’t ready.”

“But do you still want to join me?”

-“Yes I want”.

– “Is there something you want to ask me?”

-“Could you hug me?”.

-“Of course”.

According to the criteria of The Trust Project