Facebook’s chief artificial intelligence scientist Yann LeCun says that Elon Musk is “nuts” to call for regulation on AI at this relatively early stage in its development.
While there have been a number of important AI breakthroughs in the last few years, AI is still fairly nascent. Try having an in-depth conversation with Siri and you’ll soon see. The most sophisticated AI agents can excel at certain narrow tasks, such as games, but they lack common sense and the ability to learn across multiple domains.
Yet Musk has repeatedly warned that AI could spell the end of humanity, saying he thinks the technology has the potential to be more dangerous than nuclear weapons and that it may even spark a third world war. He believes AI should be regulated now before it becomes a danger to all of us.
Asked about Musk calling for AI regulation at Facebook’s New York office last week, LeCun said: “That’s nuts.”
Last year, Facebook CEO Mark also said Musk’s warnings about AI were “pretty irresponsible”. Musk responded by saying that Mark’s “understanding of the subject is limited.”
While LeCun has his concerns about AI, he said Musk is overreacting. “He talked to some people who were a little more optimistic in the business that human-level AI was just around the corner.”
Five years ago, these people were saying human-level AI was five to 10 years away, LeCun said. “Most of us who knew what we were doing knew that was just not going to happen that fast.”
“He [Musk] got a little panicky. I think it’s partly because he talked to some of the slightly more dreamer people at DeepMind who have since become much more…it was before they confronted themselves with the realities of AI. It’s much harder than they thought.
“Then he [Musk] read Nick Bostrom’s book [“Superintelligence”] and put one or two together and said you know, ‘we’re doomed’. He also likes the idea that he’s going to save humanity, right? So that was kind of a perfect.”
Musk was one of the first investors in the DeepMind AI lab, which was acquired by Google for a reported £400 million in 2014.
Musk set up his own $1 billion AI lab called OpenAI in December 2015 with Y Combinator president Sam Altman. “That didn’t make any sense because, you know, ‘I’m going to save humanity from building AI by actually building AI faster than Google and Facebook,” said LeCun. “I don’t know.” Musk is no longer a member of the OpenAI board.
LeCun said OpenAI is a “cute place” doing some “decent work”. He added: “It’s not revolutionary, but it’s not bad.”