Geoffrey Hinton, considered one of the founding fathers of artificial intelligence (AI), has warned of its dangers as he leaves his job at giant Google. “I left so I could talk about the dangers of AI without worrying about the impact on Google,” he tweeted after his departure was announced in The New York Times.

Advances in this sector induce “profound risks for society and humanity”, estimates in the American newspaper Geoffrey Hinton who created a foundation dedicated to AI systems. “Look at where we were five years ago and where we are now,” he continues, calling the outlook for the future “scary” as he makes projections based on the progress of the past few years.

“It’s hard to see how to avoid bad actors using it for bad things,” he said. The rapid deployment of an increasingly “general” artificial intelligence (AI), endowed with human cognitive capacities and therefore likely to disrupt many professions, was symbolized by the launch in March by OpenAI of GPT-4, a new, more powerful version of the natural language model that powers ChatGPT.

This generative AI interface has been used by millions of people for several months to write essays, poems or even lines of computer code. This launch also spurred competition in this area.

Geoffrey Hinton also warns, in the New York Times, against the misinformation that could be generated by AI. The expert informed Google of his resignation last month, according to the newspaper. In the tweet confirming his departure, he refutes any desire to criticize the tech giant for this decision. “Google behaved very responsibly,” he wrote.

In March, billionaire Elon Musk – one of the founders of OpenAI, whose board he later left – and hundreds of global experts called for a six-month break from research into AIs more powerful than GPT. -4, citing “major risks to humanity”. Geoffrey Hinton was not among the signatories, but he told the New York Times that scientists shouldn’t ramp up these AIs “until they know if they can control them.”

In 2019, Mr. Hinton had received, along with two other specialists in artificial intelligence, the Turing Prize, the equivalent of the Nobel for computer scientists.