News for hackers. Stuff that matters.
Re:What Happened in the 2010s by CalChris on Wednesday January 01, 2020 @03:05PM (#21930220)
Strongly dis-disagree. Google's AlphaZero and AlphaGo look awfully fearsome. Generalizing from that to other expert endeavors seems inevitable rather than some sort of mania. Moreover, saying that chess and go aren't benchmark hard problems strikes me as a technical No True Scotsman because they were always seen as that before Alpha obliterated them.
Re:What Happened in the 2010s by hakfoo on Wednesday January 01, 2020 @10:00PM (#21932644)
Chess and Go are constrained problem spaces. The space may be huge, but it's still finite. The win condition is unambiguous. The rules can fit on a postcard. There are only a finite number of possible board configurations.

Real-world win conditions tend to be more complex. For non-general AI tasks, this is much less of a problem. "Identify hot dog vs. not-hot-dog accurately 99% of the time" is measurable and verifiable. But a simple goal for a complex general-AI task is likely to build a Paperclip Optimizer. I suspect we don't even have the right language to build a good win condition for a lot of real world problems-- we may end up ignoring "insignificant" factors only to discover they're huge over tine or at scale.

Games also tend to provide near-perfect or perfect information. At a minimum, there's constrained state transitions. I can't pull out my Oyster card in the middle of a poker game, add it to my hand, and declare it's actually a four of clubs. Real problems can have surprise off-the-wall, or outright externally triggered transitions, that completely throw the model off.

"AI, build me a market-beating stock portfolio to cash out 2040-01-01" is a lot harder than "AI, checkmate this king."

Re:What Happened in the 2010s by CalChris on Thursday January 02, 2020 @08:28AM (#21935152)
I was replying to a comment about ML, not AI.