AI pioneer wants to build systems that don't harm humanity

Yoshua Bengio is a stickler for ideas. Winner of the 2018 Turing Award and scientific director of MILA, the Montreal Institute for Artificial Intelligence (AI), this Canadian computer scientist is renowned as one of the pioneers of deep learning, driving the revival of AI over the past fifteen years. He's also known for warning, more recently, about the risks inherent in these technologies, including catastrophic scenarios that could lead to the annihilation of humanity.

In January, he published a vast work that he had coordinated to assess the risks . While the report was balanced, he himself has a more clear-cut opinion, worrying about a possible mass extinction and calling for the precautionary principle to curb current developments. On June 3, he took a new step, no longer content with simply warning. He is launching a new private research laboratory to develop "technical solutions for AI systems that are safe by design." That is, as he explained to Le Monde in a video interview, to create AI "that will not turn against us and that cannot be used to cause harm."

You have 73.38% of this article left to read. The rest is reserved for subscribers.