3 ms·
Alpha zero is a generalized AlphaGo, so they are headed in the right direction! AlphaGo -> win this boardgame AlphaZero -> win any boardgame (well, three righ
by iheartpotatoes 8y ago
Alpha zero is a generalized AlphaGo, so they are headed in the right direction!
AlphaGo -> win this boardgame
AlphaZero -> win any boardgame (well, three right now)
AlphaMinus1 -> win any game
AlphaMinus2 -> win anything
AlphaMinus3 -> win winning. so much winning.
But you get my drift, I'm extrapolating hugely off one abstraction step.
- ForHackernews 8y agoYeah maybe. How do you define "win" in terms of general intelligence? Outsmart all top human experts at their field?
- wetpaws 8y agoWhatever the reward function dictates would be a win :)
- ForHackernews 8y agoRight. So we've just successfully moved the goalposts to "write a reward function that objectively evaluates g"
- iheartpotatoes 8y agoImpossible to answer: who "won" politics? Democrats or republicans? Eventually we go from a black-and-white rules (hell, even Go is kind of a negotiation), to using economics as a tool to determine what option satisfies all parties. I don't think humans have solved that yet, but like radiology, perhaps AI can make better compromises?