A University of Toronto professor and Nobel Prize winner, often referred to as the “Godfather of AI,” is warning that the technology could develop and “make us irrelevant” without significant research into how to control it.
Geoffrey Hinton, a British-Canadian professor at the University of Toronto, won the Nobel Prize in physics at the beginning of the month along with Princeton University researcher John Hopfield for work on machine learning and artificial neural networks.
Speaking to Global News in a rare interview this week after winning the prestigious international award, Hinton pointed to the positives AI can bring but warned its rapid evolution could be existentially problematic.
“I’m most concerned about the long-term dangers because those are the ones that people think are maybe just science fiction,” he said.
“So, the biggest long-term danger is that, once these artificial intelligences get smarter than we are, they will take control –they’ll make us irrelevant. And that’s quite worrying; and nobody knows how to prevent that for sure, so we need to do lots of research on that right now.”
Hinton’s joint prize was for work using physical to “design artificial neural networks that function as associative memories and find patterns in large data sets,” according to a member of the Nobel committee.
Discoveries Hinton worked on are now used to advance physics in areas like facial recognition and language translation, according to the committee member.
Hinton, who left Google so he could speak more openly about the dangers artificial intelligence poses, said health care was a key area where artificial intelligence can be a huge help.
Where the technology goes in the next decade, he said, is impossible to predict, even for those leading the growth and development of AI.
“I don’t think you can see clearly ten years in the future, things are moving too fast,” he said. “It’s like fog with fog, you can see very clearly for a certain distance and then the wall comes down and after that, you can’t see anything.”
He suggested that looking back at the past decade shows just how fast the technology can develop.
The rapid growth of AI and technology over the past decade has confounded all expectations, Hinton explained, pointing to large language models like ChatGPT or Google’s Gemini as examples.
While the technology still has hiccups and issues with inaccuracy, it can generally write complete and compelling sentences, something the industry could not have predicted 10 years ago, he said.
“These large language models and their ability to deal properly with natural language is amazingly much better than we would have predicted 10 years ago,” he said.
“Nobody thought we would have language models that could answer any question you gave and could be a not-very-good expert at everything — that’s amazing. Ten years ago, people would have very confidently predicted we wouldn’t be there yet.”
— with files from The Canadian Press