Listen to this story
|
Godfather of AI Geoffrey Hinton recently said that we still won’t understand brains at the point where AI is smarter than humans.
Earlier this month, Hinton was presented with the UCD Ulysses Medal for his contributions to the field of deep learning. During his lecture at the college, Hinton spoke about how it wasn’t necessary to emulate the human brain with artificial intelligence. In fact, he said that he stopped believing the neural nets for digital computers were similar to how the human brain operated.
“I stopped believing that if you make them more like brains, they’ll get better. It’s quite possible that we still won’t understand the brain at the point when these things are smarter than us,” he said.
Further, speaking on the differences between AI and the human brain, he said, “I started believing neural nets using backpropagation on digital computers are already somewhat different from us.”
He said that while AI could hold knowledge similarly to a brain, the ability to make copies and share efficiently between models made them more efficient.
Hinton has been vocal about the divergence of AI from human brains, stating that AI is much better than a human brain at learning. However, to do so requires significantly more power than a brain.
“Biological computation is great for evolving because it requires very little energy, but my conclusion is that digital computation is just better,” he had said during a Romanes Lecture he delivered in February. He had also mentioned that this was likely something that was as far as 20 years away.
Following Hinton’s statement, several seem to think that AI, or even ASI, could help explain brains to us. Including Perplexity’s Aravind Srinivas who said,
Some also believed that this could help provide breakthroughs in terms of helping identify and diagnose mental illnesses, as well as find cures for them.
However, as Hinton had stated, AI might become smarter than humans before that happened.
Further, Hinton also joked that he had spent his entire life trying to understand how brains worked using artificial neural networks, which had been a “failure” leading to his current contributions to the field.