Communications of the ACM
The Meaning of Alphago, the AI Program That Beat a Go Champ
The victory of Google DeepMind's AlphaGo algorithm over a champion Go player is seen as a resurgence for deep learning from a once-derided programming method to the next great advance for artificial intelligence.
Credit: Lee Jin-man/AP
The victory of Google DeepMind's AlphaGo algorithm over a champion Go player is seen as a resurgence for deep learning from a once-derided programming method to the next great advance for artificial intelligence (AI), and the University of Toronto's Geoffrey Hinton, who pioneered deep learning, foresees significant milestones ahead in an interview.
Hinton says AlphaGo's success illustrates neural networks' ability to think intuitively, and he expects such networks eventually to enable better Web searches as natural-language processing improves, among other things. "My belief is that we're not going to get human-level abilities until we have systems that have the same number of parameters in them as the brain," Hinton notes.
He anticipates it will take more than five years for such systems to come to fruition, and accomplishing this requires much more computation and better hardware to enable proper common-sense reasoning.
Hinton also says anxieties about AI overtaking mankind are mostly misplaced, arguing such scenarios stem from misuse of the technology and not from the technology itself. He acknowledges such issues bear consideration, and says the focus of thinking should not be on crippling AI, but on improving the political system to prevent abuse.
From Maclean's
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found