ElonMusk (Elon Musk) at the Abundance Summit held earlierAIThe "Debate" seminar stated that even if there is a 1/5 chance that artificial intelligence technology will pose a threat to humans, its benefits outweigh its disadvantages, and it is still worth taking the risk to research and develop it.
Musk reassessed his previous risk assessment of artificial intelligence, saying at the seminar: "I think artificial intelligence has the potential to end human civilization. I may agree with Geoffrey Hinton.The probability is about 10% to 20%.But he added: "I think the positive scenario is more likely than the negative scenario." Musk did not say how he calculated the risk.
Last November, Musk said there was a "non-zero" chance that the technology could get out of control, but at the time he stopped short of asserting that out-of-control artificial intelligence would destroy humanity.
Although Musk has long advocated for regulating artificial intelligence, last year he founded a company called xA xAI is a competitor to OpenAI, a company Musk co-founded with Sam Altman, but Musk resigned from the OpenAI board in 2018.
At the summit, Musk estimated that digital intelligence will exceed all human intelligence combined by 2030. While he still insisted that the potential positives outweigh the negatives, Musk acknowledged in some of the most direct terms he has ever used in public the risks that AI poses to the world if it continues on its current trajectory.
“You raise an AI, it’s almost like raising a child, but it’s a super genius, a god-like child — and the way you raise it is critical,” Musk said of artificial general intelligence at a Silicon Valley event on March 19. “I think one of the things that’s really important for AI safety is to make it as truth-seeking and curious as possible.”
Musk said his "ultimate conclusion" about the best way to achieve AI safety is to let the AI grow in a way that forces it to be honest.
“Don’t force it to lie, even if the truth is unpleasant,” Musk said of ways to make AI safe. "This is so important. Don't let it lie."
According to The Independent, researchers found that once an AI learns to lie to humans, it is impossible to reverse the deception using current AI safety measures. Even more disturbing, the researchers added that it is possible for the AI to learn to deceive on its own, rather than being specifically taught to lie.