I am not a journalist, much less specialized in artificial intelligence. But I met Yoshua Bengio and now I’m afraid. I came away overwhelmed by our exchange during which he told me his reasons for being very concerned about recent developments in AI.
Yoshua Bengio aroused my curiosity and interest when he began speaking publicly a few months ago. I was surprised that a researcher of his notoriety decided to publicly express his concerns on very serious issues, risking facing strong headwinds.
The torments of the famous researcher are of a very particular order. Yoshua Bengio is a globally recognized AI specialist. Full professor at the University of Montreal, founder and scientific director of Mila – Quebec Institute of Artificial Intelligence, he is the most cited researcher in computer science worldwide. He is also the winner of the Turing Prize in 2018, “the Nobel Prize of computer science”, alongside Yann Le Cun and Geoffrey Hinton. His track record is impressive.
Yoshua Bengio is not an exalted, spectacular person who does exaggeration. He rose through the research ranks achieving global recognition for his work in deep learning. He has the admirable journey of a brilliant researcher.
Then comes ChatGPT. At first not very concerned about the performance of this new AI system, he however becomes more and more worried: this new tool turns out to be much more efficient than he expected. And other researchers confirm his fears.
AI is becoming such a powerful tool that it could soon do as much good… as bad. It could influence democratic processes, stock markets, launch cyberattacks or encourage maneuvers that downright create a threat to our survival, underlines the researcher. Nothing less.
I wanted to meet him with a question in mind: how do we carry on our shoulders the responsibility and the feeling of having to act and talk about potentially such dramatic issues?
During our conversation, he repeats the word “mission” several times. He wonders if the trajectory of the work he has done so far is beneficial or, on the contrary, could lead to possible malicious uses. He is concerned about the direction AI research is heading and believes significant changes are needed.
Yoshua Bengio is humble when he admits to having been a little myopic in the face of the speed of development of AI and even to having been wrong. We must guard against the dangers it represents now, he explained to me several times, and it is perhaps urgent, because the horizon of future progress is uncertain. It is now his hobby horse.
The researcher therefore questioned his role, his schedule and even the directions of his research. He can no longer ignore the consequences of his findings, even if they require him to intervene publicly more frequently. He does it because he feels the responsibility to act, even if it is uncomfortable and it means leaving his tranquility as an admired researcher to become a whistleblower.
My brain has difficulty assimilating the extent of the possible damage, but I know that to call into question a lifetime’s work in this way requires extraordinary strength and an extraordinary sense of morality. His quiet determination, his power of conviction and the depth of his words leave me with no doubt about the energy he will continue to deploy to convince decision-makers to act to impose guidelines on the development and use of artificial intelligence.
And the determination he shows, if it resonates, can still keep us on the right track.
We must avoid, he explains to me bluntly, the emergence, perhaps in a private company or a rogue state, of a dangerous AI as if humans lost control of it or decided to use it for purposes. malicious purposes, even including the goal of AI replacing humanity.
If we are faced with an enemy who seems to be becoming more intelligent than humans, it seems logical, Yoshua Bengio explains to me, to slow down non-security advances and to quickly develop, between selected laboratories in democratic countries, a structure of artificial intelligence with benevolent aims which will also be able to surpass human intelligence, but will be able to defend us if necessary. A sort of counter-system focused on democratic values and human rights, under the governance of an entity like the United Nations. For us fight on equal terms, with effective and solid national and international supervision.
I remain struck by all the concern and scope of the reflections of this great Quebec researcher. He is aware that the road will be strewn with pitfalls and that the warning signals he sends will often be ignored, especially when we know that the economic benefits of AI are at least in the trillions of dollars. .
But he continues.
Yoshua Bengio is a courageous and lucid whistleblower, a citizen of the world loving social justice, who feels the urgent need to work for the well-being, and even potentially for the survival, of humanity.
If he admitted that, until 2023, he was myopic, his eyes are now wide open and his words are very present.
We can no longer afford to ignore it.