New investment. After Geoffrey Hinton, one of the fathers of artificial intelligence, it is the turn of his counterpart Yoshua Bengio to express concern about the dazzling progress of AIs like ChatGPT and to lament some of his contributions to the field.
With the BBC, the Canadian scientist, a historic figure in the industry, says he fears that malicious actors will use increasingly sophisticated AI. “It could be the military, terrorists, a very angry madman… If we can easily program these AIs to do something bad, it could be very dangerous.”
For the researcher, the risk is greater if AIs become smarter than humans: “If they are smarter than us, it will be difficult to stop these systems or prevent damage.”
“Think about the atomic bomb”
Yoshua Bengio is one of the main signatories of the declaration that calls for AI to be considered a risk as great as pandemics or nuclear war, and that it could cause the “extinction of humanity”.
“It is emotionally difficult for those who are [dans le secteur de l’IA]especially if like me you have built your career, your identity around the idea of contributing something useful to society”, explains the researcher.
Despite everything, it does not say that it is defeated. “It’s not the first time scientists have gone through these kinds of emotions — think World War II and the atomic bomb.”
“It’s exactly like climate change,” sums up the specialist. “We put a lot of CO2 into the atmosphere, we would have done better to avoid it, but let’s see what we can do now.”
But not all specialists agree with Yoshua Bengio’s concerns. Among them, the last of the three “founding fathers” of modern AI: a Frenchman, Yann Le Cun. he he is great On twitter that “superhuman AIs are far from the greatest threat to humanity, largely because they don’t exist yet”, and that “until we have a basic plan to create a dog-level AI (…), for so to speak rendering [l’IA humaine] harmless is premature.”
Source: BFM TV
