"Maybe we could look at some of the things that the two-year-olds do when they’re learning and see if that makes a difference to what the AIs are doing when they're learning," said Alison Gopnik, University of California, Berkeley psychologist, in a recent New York Times podcast interview. Gopnik is also a member of the institution's Artificial Intelligence Research Group, which is working to figure out how to design an AI system that is as smart as a two-year-old.
"It's kind of striking that the very best state-of-the-art systems that we have that are great at playing Go and playing chess and maybe even driving in some circumstances, [but] are terrible at doing the kinds of things that every two-year-old can do," Gopnik added. "If you look at the current models for AI, it's like we're giving these AIs hyper helicopter tiger moms. There's a programmer who's hovering over the AI and saying, 'oh, yeah, yeah, you got that one right. That one's a cat. That one's a dog. That one's another cat. That one's another dog.'
"What you end up with is AI systems that are very, very good at doing the things that they were trained to do and not very good at all at doing something different."
From The New York Times
View Full Article