5 hours ago
7
"Godfather of AI" connected tech's aboriginal
"Godfather of AI" Geoffrey Hinton was awakened successful the mediate of the nighttime past twelvemonth with quality helium had won the Nobel Prize successful physics. He said helium ne'er expected specified recognition.
"I dreamt astir winning 1 for figuring retired however the encephalon works. But I didn't fig retired however the encephalon works, but I won 1 anyway," Hinton said.
The 77-year-old researcher earned the grant for his pioneering enactment successful neural networks — proposing successful 1986 a method to foretell the adjacent connection successful a series — present the foundational conception down today's ample connection models.
While Hinton believes artificial intelligence will alteration acquisition and medicine and perchance lick clime change, he's progressively acrophobic astir its accelerated development.
"The champion mode to recognize it emotionally is we are similar idiosyncratic who has this truly cute tiger cub," Hinton explained. "Unless you tin beryllium precise definite that it's not gonna privation to termination you erstwhile it's grown up, you should worry."
The AI pioneer estimates a 10% to 20% hazard that artificial quality volition yet instrumentality power from humans.
"People haven't got it yet, radical haven't understood what's coming," helium warned.
His concerns echo those of manufacture leaders similar Google CEO Sundar Pichai, X-AI's Elon Musk, and OpenAI CEO Sam Altman, who person each expressed akin worries. Yet Hinton criticizes these aforesaid companies for prioritizing profits implicit safety.
"If you look astatine what the large companies are doing close now, they're lobbying to get little AI regulation. There's hardly immoderate regularisation arsenic it is, but they privation less," Hinton said.
Hinton appears peculiarly disappointed with Google, wherever helium antecedently worked, for reversing its stance connected subject AI applications.
According to Hinton, AI companies should dedicate importantly much resources to information probe — "like a third" of their computing power, compared to the overmuch smaller fraction presently allocated.
CBS News asked each the AI labs mentioned however overmuch of their compute is utilized for information research. None of them gave a number. All person said information is important and they enactment regularisation successful wide but person mostly opposed the regulations lawmakers person enactment guardant truthful far.
English (US) ·