AI could soon become super intelligent, and this may not be good for humans
May 3, 2023—at MIT’s EmTech Digital conference, cognitive psychologist and computer scientist Geoffrey Hinton spoke about how he’s helped build machines that are immortal and the various dangers these machines now pose to humanity: “Smart things can outsmart us,” says Hinton.
After a morning of tech companies touting their AI products, from acceleration of drug discovery to autonomous airplanes, Hinton, who quit his job at Google yesterday, took to a virtual podium as yet another leader in the field raising concerns about AI.
“I used to think that the computer models we were developing weren’t as good as the brain,” says Hinton, adding that over the last few months he’s changed his mind completely and the existential threats have become apparent. AI is getting much smarter than us with increasingly paired down hardware compared to the wetwear of human brains.
Hinton likens AI to a hive mind that can make thousands of copies of itself. Then everything one copy learns, the whole hive learns. With advances in multimodal models that have trained on video (visual and auditory/sensory information) these machines will be much smarter than large language models that just train on language alone.
Compound all this with the fact that these are goal-oriented machines that need control over their autonomy and environment to better achieve their goals, whether the goals are programmed by a human or autonomously generated. Hinton emphasizes the control problem, “If these things get carried away with getting more control, we’re in trouble.”
Hinton feels that Google handled AI responsibly when his former employer was the forerunner in AI. Once OpenAI very publicly burst on the scene and was bought by Microsoft, the race to compete, according to Hinton, superseded safety concerns because in any system with competition, it’s not possible to slow the process of development.
He does find some hope in coming together to responsibly manage AI, since the threat extends to all biological beings. “We are all in the same boat with respect to the existential threats. So we all ought to be able to cooperate on trying to stop it,” says Hinton.