At a time when there have been fears that AI would take jobs and change into harmful, Meta's AI scientist Yann LeCun stated that the know-how isn’t a pure phenomenon and that it’s going to change into harmful to current ” existential threat' to people. Often called one of many three “Godfathers of AI”, LeCun emphasised that AI is not going to have human-level intelligence within the close to future.
“AI isn’t some sort of pure phenomenon that may emerge and change into harmful. *WE* design it and *WE* construct it,” he stated in a publish on X.
LeCun drew parallels to the event of AI know-how with turbojets.
“I can think about hundreds of eventualities the place a turbojet goes horribly improper. Nonetheless, we had been capable of make the turbojets insanely dependable earlier than deploying them on a big scale,” he stated.
What AI scientists at Meta need to say about AI regulation
The query is analogous for AI: “Do we predict there may be not less than one design of an AI system that’s each secure/controllable and might accomplish objectives in smarter methods than people?” If the reply is sure, we'll be fantastic. If the reply is not any, we gained't construct it. Proper now, we don't also have a trace of designing a human-level clever system. So it’s too early to fret. And it’s too quickly to manage it to stop “existential threat”.
In a separate publish, he stated that “simply because a $30 gadget can beat you at chess doesn't imply it's smarter than you.”
LeCun says synthetic intelligence is not going to be smarter than people. LeCun beforehand dismissed Elon Musk's claims that Ai “will in all probability be smarter than any human within the subsequent yr.”
“No. If that had been the case, we'd have AI techniques that might train themselves to drive a automotive in 20 hours of apply, like all 17-year-old,” LeCun instructed X.
“AI isn’t some sort of pure phenomenon that may emerge and change into harmful. *WE* design it and *WE* construct it,” he stated in a publish on X.
LeCun drew parallels to the event of AI know-how with turbojets.
“I can think about hundreds of eventualities the place a turbojet goes horribly improper. Nonetheless, we had been capable of make the turbojets insanely dependable earlier than deploying them on a big scale,” he stated.
What AI scientists at Meta need to say about AI regulation
The query is analogous for AI: “Do we predict there may be not less than one design of an AI system that’s each secure/controllable and might accomplish objectives in smarter methods than people?” If the reply is sure, we'll be fantastic. If the reply is not any, we gained't construct it. Proper now, we don't also have a trace of designing a human-level clever system. So it’s too early to fret. And it’s too quickly to manage it to stop “existential threat”.
In a separate publish, he stated that “simply because a $30 gadget can beat you at chess doesn't imply it's smarter than you.”
LeCun says synthetic intelligence is not going to be smarter than people. LeCun beforehand dismissed Elon Musk's claims that Ai “will in all probability be smarter than any human within the subsequent yr.”
“No. If that had been the case, we'd have AI techniques that might train themselves to drive a automotive in 20 hours of apply, like all 17-year-old,” LeCun instructed X.
finish of the article