The comments here are really missing the forest for the trees. The point is: here is one computer Go researcher predicting computer Go dominance would be achievable by 2017. I keep reading in the media that "Go experts thought it would take decades," but the fact is it's hard to find actual quotations from computer Go experts.
You can easily find many professional Go players who didn't think Go programs would surpass humans for decades, but these people aren't engineers. This would be like asking a human off the street for predictions about AGI.
It doesn't really matter whether or not the researcher predicted the exact methods of AlphaGo (he didn't - while AlphaGo uses tree search methods very extensively, just as important were the policy/value nets). Does anyone really believe that the people with predictions about the timeframe of AGI are going to predict the exact methods used in AGI?
>Go programs have improved about one stone per year for the last three decades. Hardware
doublings produce diminishing Elo gains on a scale consistent with accounting for
around half of all progress.
They don't give a clear prediction for when it was expected to exceed the best humans, but AlphaGo was clearly a discontinuous jump in ability.
>After the match, I ask Coulom when a machine will win without a handicap. “I think maybe ten years,” he says. “But I do not like to make predictions.”...
>Even with Monte Carlo, another ten years may prove too optimistic. And while programmers are virtually unanimous in saying computers will eventually top the humans, many in the Go community are skeptical. “The question of whether they’ll get there is an open one,” says Will Lockhart, director of the Go documentary The Surrounding Game. “Those who are familiar with just how strong professionals really are, they’re not so sure.”
Right. The only quotation is "I think maybe ten years," by Remi Coulon. Obviously the community was mixed. But if you believed the way the media has been spinning it, you'd think the computer Go community as a collective thought it would not be possible.
Otherwise, the opinions of human Go players at the time (that were not developers of computer Go programs) aren't relevant, nor is the opinion of a WIRED article writer.
You can easily find many professional Go players who didn't think Go programs would surpass humans for decades, but these people aren't engineers. This would be like asking a human off the street for predictions about AGI.
It doesn't really matter whether or not the researcher predicted the exact methods of AlphaGo (he didn't - while AlphaGo uses tree search methods very extensively, just as important were the policy/value nets). Does anyone really believe that the people with predictions about the timeframe of AGI are going to predict the exact methods used in AGI?