Hacker News Comments on
Match 2 - Google DeepMind Challenge Match: Lee Sedol vs AlphaGo
DeepMind
·
Youtube
·
8
HN points
·
9
HN comments
- This course is unranked · view top recommended courses
Hacker News Stories and Comments
All the comments and stories posted to Hacker News that reference this video.I've been looking forward to this one basically since the Lee Sedol matches. Excited it's finally here!I started the Lee Sedol matches knowing nothing about Go but the commentators really helped me understand what was so special about the game, and to come to appreciate the poetry in the language that's grown around the game. I still don't know how to play Go well, but now I understand the basics enough to appreciate all that has been done to get AI to the point where I no longer expect any human will ever defeat the top AI in this game again. But only time (and matches!) will tell, and that's what tonight finally brings!
I'll be watching live hoping for moments like this: https://youtu.be/l-GsfyVCBu0?t=4692
https://youtu.be/l-GsfyVCBu0?t=2866
⬐ tazjinThank you, that's exactly what I was looking for!⬐ dwringerThanks for bringing this up. I was recalling more recent articles that did state there were modifications each night between matches. Perhaps those were in error; I would not have commented in that case. It's a shame nobody pointed this out sooner, I could have removed the misinformation.
Seems to be similar for intuition in humans–some moves just feel right, and they're the result of thousands of hours of experience. You can justify it afterwards, but the intuition itself usually comes first.The AI google designed is architected similarly to how the brain works in dual process theory: you have a NN providing intuition like system 1 and a supervisor much like system 2 which double checks system 1
In the video for the second match, a Google employee mentions that a neural net they call the policy net (trained on a large sample of historical games) provides intuitive moves, while another NN evaluates board strength. They apply the policy net to find multiple interesting moves, then continue to apply the net to anticipate the opponents moves to generate a tree of possible moves. It then just settles on which move to make that gives it the best odds of winningStarts at 42:00 https://www.youtube.com/watch?v=l-GsfyVCBu0
During the second game it made a move that was decidedly not human (which helped lead to the win later):
It was during the game on the main stream:
Specifically, move 37 at O10: https://youtu.be/l-GsfyVCBu0?t=1h17m45s
⬐ trompthat link doesn't work for me. but these do:http://gobase.org/replay/?gam=/games/inter/matches/AlphaGo/L...
Here: https://www.youtube.com/watch?v=l-GsfyVCBu0&t=2511
I'm pretty sure I heard them say this.They can see certain stats and high level overview.
But they have no idea what it's thinking.
They know the general pattern of the algorithm. They even explain it.
But the algorithm involves two deep neural networks, and they don't really know what's going on inside them.
One of the developers showed up during the commentary on the second game and talk about this stuff:
⬐ taneqIs it normal for players to leave the board without announcing a break (whatever the correct terminology is)?⬐ Jerry2Does anyone know what DeepMind's software stack looks like? Just based on past work of some of the people working there, I'm guessing most of the code is C++ with some Lua. Anyone know for sure?