4 ms·

I thought this was supposed to be a few years away -- i.e beating Lee Sidol was impressive, but Lee wasn't nearly the top player in the world? Wow, this must b

I thought this was supposed to be a few years away -- i.e beating Lee Sidol was impressive, but Lee wasn't nearly the top player in the world? Wow, this must be what it feels like to be living on a parabolic curve...

daveguy 9y ago
Lee Sedol was near the top (#7). He wasn't the very top, that was Ke Jie. And now Ke Jie has been beaten 2/3, maybe 3/3 times. Sedol was beaten soundly -- 4/5 games. So this is an improvement, but not an exponential one. Go is definitely conquered like chess was 1-2 decades ago. https://www.goratings.org/en/ https://www.goratings.org/en/
- stupidcar 9y ago
  Ironic that even though Sedol is not the strongest human player, he may end up being the last human in history to beat the strongest AI.
- thomasahle 9y ago
  > this is an improvement, but not an exponential one. I wonder how you define exponential here. If the old version had a probability 20% of losing against Lee Sedol, and the new one has 5%, then one might call it exponential. Something like losing prob = 2^(2012-current year).
- Deimorz 9y ago
  According to some information that the AlphaGo team released earlier in the week, the current version of AlphaGo can give a three-stone handicap to the version that played Lee Sedol: http://www.usgo.org/news/2017/05/new-version-of-alphago-self-trained-and-much-more-efficient/ http://www.usgo.org/news/2017/05/new-version-of-alphago-self.... I don't know if that qualifies as "exponential", but that's a massive difference in skill at the game. If it's accurate at all, I think there's basically zero chance that a human will ever beat AlphaGo again.
  - cjbprime 9y ago
    Wow, hadn't seen that, that's terrifying.
    - Deimorz 9y ago
      Terrifying in some ways, but I think it's also inspiring. It's pretty amazing to have some evidence that there's still that much "room" left at the top of the game for players to continue improving.
  - BoiledCabbage 9y ago
    And the more interesting note - it gained a three stone handicap of skill with 1 year of practice. If they were to continue at it should we expect a 3 stone improvement in skill every year? Maybe they made some major leaps this time and future changes aren't as large. So a 3 stone improvement every two years. It doesn't take long until the Go being played is fundamentally a different game than what we play. Even if it's on the same board with same rules. What does it look like when Go still isn't solved, but AI is 20 stones better than the best humans? At that point we can no longer even tangibly measure its brilliance. Only ask it to compare its brilliance to other brilliant AIs. What will that even look like through out eyes? What will like look like when an AI is 20 stones better than us at war games? I assume next step for AlphaGo is a real time strategy game like Starcraft, or some other game with incomplete information. 3-5 years from now when mastered it'll be hard to persuade people not to use it in the real world.
waqf 9y ago
AlphaGo had a fundamentally new architecture (nobody had successfully applied convolutional neural networks to Go before, although of course the headline underestimates the expertise and experimentation that's needed to get the details right). So it's not surprising that at the first public announcement they hadn't yet pushed the technology as far as it would go.

daveguy 9y ago

Lee Sedol was near the top (#7). He wasn't the very top, that was Ke Jie. And now Ke Jie has been beaten 2/3, maybe 3/3 times. Sedol was beaten soundly -- 4/5 games. So this is an improvement, but not an exponential one. Go is definitely conquered like chess was 1-2 decades ago. https://www.goratings.org/en/ https://www.goratings.org/en/

stupidcar 9y ago
Ironic that even though Sedol is not the strongest human player, he may end up being the last human in history to beat the strongest AI.
thomasahle 9y ago
> this is an improvement, but not an exponential one. I wonder how you define exponential here. If the old version had a probability 20% of losing against Lee Sedol, and the new one has 5%, then one might call it exponential. Something like losing prob = 2^(2012-current year).
Deimorz 9y ago
According to some information that the AlphaGo team released earlier in the week, the current version of AlphaGo can give a three-stone handicap to the version that played Lee Sedol: http://www.usgo.org/news/2017/05/new-version-of-alphago-self-trained-and-much-more-efficient/ http://www.usgo.org/news/2017/05/new-version-of-alphago-self.... I don't know if that qualifies as "exponential", but that's a massive difference in skill at the game. If it's accurate at all, I think there's basically zero chance that a human will ever beat AlphaGo again.
- cjbprime 9y ago
  Wow, hadn't seen that, that's terrifying.
  - Deimorz 9y ago
    Terrifying in some ways, but I think it's also inspiring. It's pretty amazing to have some evidence that there's still that much "room" left at the top of the game for players to continue improving.
- BoiledCabbage 9y ago
  And the more interesting note - it gained a three stone handicap of skill with 1 year of practice. If they were to continue at it should we expect a 3 stone improvement in skill every year? Maybe they made some major leaps this time and future changes aren't as large. So a 3 stone improvement every two years. It doesn't take long until the Go being played is fundamentally a different game than what we play. Even if it's on the same board with same rules. What does it look like when Go still isn't solved, but AI is 20 stones better than the best humans? At that point we can no longer even tangibly measure its brilliance. Only ask it to compare its brilliance to other brilliant AIs. What will that even look like through out eyes? What will like look like when an AI is 20 stones better than us at war games? I assume next step for AlphaGo is a real time strategy game like Starcraft, or some other game with incomplete information. 3-5 years from now when mastered it'll be hard to persuade people not to use it in the real world.

waqf 9y ago

AlphaGo had a fundamentally new architecture (nobody had successfully applied convolutional neural networks to Go before, although of course the headline underestimates the expertise and experimentation that's needed to get the details right). So it's not surprising that at the first public announcement they hadn't yet pushed the technology as far as it would go.