AlphaGo vs Lee Sedol: Game 1 – Video Commentary

 
The comments below are from the discussion during the live stream.
 

Related Articles

Comments

  1. Younggil An says:

    Hello everyone, I hope you’re looking forward to the game today!

    Unfortunately, real time comments were causing problems for our server with so many visitors, but you can leave a normal comment if you’d like to discuss the game with us. You’ll have to refresh the page from time to time to see new comments.

  2. Michael says:

    This is so exciting!

    No matter the result, may the power of human ingenuity win πŸ™‚

  3. Very excited to watch this with you all! Hope for a great match (and win for Sedol hopefully)

  4. Zak Smith says:

    Looking forward to an exciting match. I highly doubt the Alpha Go that played Fan Hui could beat Lee Sedol, but 6 months is an eternity in A.I. time. So I’m on the edge of my seat, waiting with baited breath!!!

  5. Younggil what sort of opening would you try against AlphaGo? What sort of opening do you expect Lee Sedol will play?

    • Younggil An says:

      Hi Wasi, that’s hard to expect, but I think Lee will try normal openings which Lee favors.

      He might try to play strange style of moves in the middle game along with kos I assume. πŸ™‚

  6. Zak Smith says:

    Yeah, I’m looking forward to seeing how AlphaGo responds to ko, and whether that is a weakness of this program like it is for every other go program.

  7. Here we goooo!

  8. Sedol trying to make AlphaGo play a moyo game?

    • David Ormerod says:

      Black’s two 3-4 points surprised me a bit. Lee used to play this opening a lot when he was younger, but I don’t think he plays it as often these days.

    • David Ormerod says:

      Move 7 is interesting πŸ™‚

      • Maybe he has prepared some unusual openings for these games?

      • Younggil An says:

        Lee’s opening especially Black 7 is abnormal, and I’ve never seen it.

        I think Lee wants to test AlphaGo’s strength in the opening. Lee might have prepared this opening.

        • David Ormerod says:

          White’s moves in the top right (attachment and extension in the corner) are starting to feel heavy. Crazy Stone often plays this exact corner pattern too.

          • Is Black’s jump normal?

            • David Ormerod says:

              All of Black’s responses in the top right seem normal (and good) to my eyes. Making a tiger’s mouth in the corner helps to prevent White’s group from making a base, and now White is helping Black to consolidate territory on the right side.

              Black 7 is looking like a good move now πŸ™‚

              That being said, it looks like White aims to build power and pincer at the top. I don’t like this way of playing, but computers seem to think it’s good in my limited experience. That makes me wonder if I’m wrong πŸ˜‰

            • Younggil An says:

              Yes, White’s moves in the top right seemed to be a little heavy, but White’s pincer at the top looks nice, so White chose that sequence I think.

              Black’s jump was alright, but Black could also take sente to play at the top first.

  9. David Ormerod says:

    Peeping the Black’s one space jump at the top makes it look like White might be about to cut and fight. Ho ho!

    • Jorik Mandemaker says:

      It feels like AlphaGo is playing more aggresively than in the matches with Fan Hui. But I guess she has no choice since black got all that territory on the right hand side.

  10. Anonymous says:

    The peep looks so cool. I would never come up with such a move^^

    • Younggil An says:

      Right, White’s peep was very well timed, and the opening is already quite interesting.

      Yes, AlphaGo’s play is more aggressive, because Black built territory on the right side, but his top is weak.

  11. Looks like a tough fight ahead

  12. charlie kang says:

    It is so exciting to watch the game… One comment though, most of us are watching this game know how to play the game yet the comments by the commentators seemed that the viewers have no knowledge of the game. I would like to hear more analytic points of view.

  13. Jorik Mandemaker says:

    Younggil, what is your opinion about the variation that the baduk tv analyzers have come up with? The one where white captures the four black stones but black captures three white stones?

  14. I think it is possible that AI will make a shock news today. The first game could always deliver some surprise. Lee could use a strategy to analyse the way the machine reacts then make comeback in later games.

  15. Anonymous says:

    Who is winning?

    • Younggil An says:

      It’s hard to say who’s winning yet.

      The flow of the game seems to be smooth for White so far.

  16. X_X getting complicated…

  17. Josh Hoak says:

    (Move 55) Wow, is this fightable? Terrifying

  18. Zak Smith says:

    I love the commentary so far, but whoever is in charge of the camera goes to the game while the commentator is going through variations. It’s nice to see the actual game when the commentators are making conversation, however when they are doing variations on the magnetic board it becomes impossible to follow their variations.

    Can you please fix this? Thanks!

    • Younggil An says:

      Thanks Zak for your feedback, but we’re not able to fix it.

      The game is still playable for White, but Lee is specialized at this sort of complicated game, so let’s see.

    • David Ormerod says:

      I’ll try letting them know on twitter. Chris said earlier that he’d be checking it at some stage. Apart from that we can’t really control the camera guy. Maybe he doesn’t really understand Go, or maybe he’s Korean and doesn’t understand English.

  19. Any major errors so far?

  20. Jorik Mandemaker says:

    AlphaGo sure is in a fighting mood today with another cut.

  21. Anonymous says:

    Why did black did not take the cutting stone K12?

  22. Michael says:

    Is there Korean live commentary somewhere? Where can we watch it?

    • David Ormerod says:

      We have another stream here: https://gogameguru.com/baduk-tv-live/

      But it’s a paid service we offer, because we have to pay the TV company in Korea to let us stream it. You can subscribe for just one day for $2.70. Other options and pricing are here.

      We’re primarily promoting the above live stream on Youtube instead (for this match) because it’s free.

  23. Lee looking more stressed than usual?

  24. David Ormerod says:

    White doesn’t really have much certain territory yet. I prefer Black’s position slightly.

  25. Younggil An says:

    Yes, Black is not bad, and White suddenly played in the top left center, and the game seems to be slightly better for Black now.

  26. Anonymous says:

    What do you think about white G13?
    IsnΒ΄t that a bit too slow?

    • Younggil An says:

      That looks slack, because Black could play one more move to make a double approach.

      White’s cross-cut looks skillful, but Black is alright.

  27. To my weak eyes, white’s group on the right doesn’t seem settled enough and doesn’t seem to have accomplished very much

    • Younggil An says:

      You’re right, but since Black’s 3-4 stone is also weak, White’s center group would be fine.

  28. White’s shell starting to crack?

  29. who won?

  30. Jorik Mandemaker says:

    R10 is very interesting. I wonder what white can achieve on the right side.

  31. Anonymous says:

    Wow, they are fighting now! Seems like it is crysis point ….

  32. Maybe he is planning to trade?

  33. Younggil An says:

    Yes, it appears that Lee’s satisfied with the trade.

  34. Spencer says:

    Is anyone else losing noticing the commentators are missing round of moves at the top of the board?

  35. Gil Dogon. says:

    Wow , alphago N2 ! Thats really exciting stuff , What do you think Younggil ?

    • Younggil An says:

      AlphaGo might have felt that she’s losing, so she chose a difficult move to respond.

      However, Black Q5 was a accurate response, and the game was still good for Black.

  36. What is komi for these games? 6.5?

  37. Why can’t he stand there quietly (instead of pretending to know anything at all) and also not block the board?

  38. jaythenorthkorean says:

    who’s wining now?

  39. Anonymous says:

    lee is lost

  40. Andy Z. says:

    Lee Sedol looks nervous. We can’t see AlphaGo, but it is probably not nervous. Where is Alphago, anyway? On site or in a remote location?

  41. Andy Z. says:

    Yes but where is the actual machine? Maybe Lee will be saved by a power surge…….

    • Anonymous says:

      Well as far as i know the HQ of Deepmind is in London, but the server can be in virtually anywhere… or maybe everywhere..

  42. Anonymous says:

    I think there are over dozens of UPS connected to AlphaGo cluster, power surge will not save anyone

  43. And I cannot see the live stream, hope it is recorded

  44. Andy Z. says:

    So it’s not in any particular place? The power surge tesuji fails then.

  45. Younggil An says:

    White seems to be leading the game… and it’s hard for Black to catch up, because it’s nearly end.

    That’s very surprising and scary!

  46. who is white and who is black?

  47. AlphaGo probably using multiple computers connecting together.

  48. β€œThe server is actually located in the Midwest of the United States and we will link the AI system to the hotel with a high-speed connection.” The high-performance version of distributed AlphaGo consists of 1202 CPU and 176 GPU.

  49. Youtube blocked in China so I can’t see anything – someone say something more than just “black is winning” since I don’t know who is black in this game…

  50. what are the prisonners for each?

  51. David Stone says:

    AlphaGo is white.

  52. Thanks all.

  53. Looks like Sedol just resigned; he just tossed a white stone on the board (he plays black).

    • Younggil An says:

      Yes, Lee Sedol resigned, and that’s unbelievable.

      We will update the news with brief comments of the game soon.

  54. Andy Z. says:

    Holy shit.

  55. Even vs a computer…have to review the game πŸ™‚

  56. Looks like a lot of Go teachers are gonna lose their jobs now…

  57. I look forward to the game record and review.

  58. David Stone says:

    Does anyone have an estimate at how many points it was?

    • Younggil An says:

      At the very end, Black was only winning by 2~3 points on the board, so he can’t pay the komi.

  59. Anonymous says:

    how did Lee fail the game? he had big advantage at the beginning

    • Anonymous says:

      What advantage?

    • Younggil An says:

      Maybe Black was slightly behind in the opening though he reversed the game in the bottom left corner.

      However, Black made mistakes in the bottom right corner, and the game was reversed again.

  60. David Tan says:

    Lee probably thought that he could kill the large white group (stretching from top to bottom) if black separated the bottom right white group from it. At the end, he managed to separate them, but both white groups survived. In the process, black lost quite a bit of area in the right.

    On the left, Lee also let white have sente, which consolidated white’s lead.

    Amazing game.

    • Anonymous says:

      No, Lee didn’t think he would kill that at all. It was just about how much profit he would make, not about killing.

    • Younggil An says:

      You’re right, and it’s strange that Lee chose to separate White’s groups while losing his bottom right corner territory…

  61. Apparently Lee underestimated Alpha Go…I can see one day the professional test would be 9 stone against a machine taking white…

  62. Looking forward to the coming matches, will it be 5-0 too?!

  63. It’s really shocking that the technological advancements in artificial intelligence has enabled it to master a complicated chess game. The ability to learn and adapt by the A.I is truly amazing. No wonder Stephen Hawking warned us that artificial intelligence could end mankind.

  64. Oh well. I lost my chance to secure my retirement income by placing a big bet against the 5-0 outcome.

    • Baldomero says:

      Exactly. The only problem with that bet is Lee Sedol might not want to go through with the full humiliation of five defeats and resign earlier. Not that it’s over yet, he could still turn it around, but all the talk about how vastly superior he would be to a machine made no sense. Creativity, ability to play unusual moves… as if Fan Hui 2p, three times European champion, lacked those. He probably tried every, trick play and desperate ploy he could think of, and pros have a vast repertoire. And as if Lee Sedol were many stones stronger. I doubt he can even read much deeper than Fan Hui. He is almost certainly weaker at yose. The Chinese professional exam is no joke. So marginal superiority in a few areas of the game, enough to make him number one, against a machine that grows stones stronger in days.
      I just hope Google leaves it running for longer and shares the software with the world. They might not want to give away their code to Facebook and other rivals to catch up, so we’ll see.

  65. EraZ3712 says:

    For those still wondering, the Korean commentators made an estimate of 15 points (including koma) in favor of AlphaGo near the end of the match.

    • EraZ3712 says:

      The values they listed were something around the mid 50’s for Lee Sedol, and mid 70’s to low 80’s for Alphago.

  66. Did lee resigned early?

    • EraZ3712 says:

      It was pretty far into the end game, so I wouldn’t say “early,” but they didn’t bother to count points. He knew he lost.

  67. is alpha go unbeatible? What do you think?

    • i am an IT guy and i am currently studying machine learning, i have been paying attention to the latest development of AI and machine learning, and looking forward to the ‘ultimate go match between AI and Human’, be honest I don’t any human being can possibly beat the AI now, just like the case of Deep Blue and Garry Kasparov (
      https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer)
      https://en.wikipedia.org/wiki/Human%E2%80%93computer_chess_matches#Komodo_handicap_matches_.282015.29
      )
      and it seems after the match Garry Kasparov joined a computer science team to develop a even better algorithm, i would rather believe that it is another way to play chess for him..

      i think the only thing that can beat AlphaGo is another AI algorithm, which is indeed designed by human, so i guess there is no need for human to lost their faith… right?

      • Rich The Engineer says:

        So far. But it won’t be long before AI will develop ability to develop own algorithms, now that “intuitive” skills, like Go, are becoming mastered. Then who will need humans to write code?

  68. I was 100% for Lee Sedol so the end was painful … Maybe 147 (L11) was too slow compared to sth like D11?
    Also looking forward to the game tomorrow

  69. Anonymous says:

    I am not really surprised by the result, after Fan Hui was defeated I think it was unlikely that Lee could win. I wonder what kind of game would be the most difficult for Alpha GO. Maybe Masaki Takemiya’s natural style would be best ?

    • Anonymous says:

      Fan Hui is orders of magnitude below Lee, so I’m not sure what you base your opinion on?

      • Anonymous says:

        Fan Hui is at most one stone weaker than Lee Sedol, that is not an huge difference and Alpha was already stronger than Fan Hui 6 months ago. 6 months is very little for a human but for a computer…

        • Anonymous says:

          No, I know Fan hui, he was my teacher. But, with all respects, Lee Sedol is much, much stronger than him now. In France, he don’t play many serious game. He’s much weaker than he was when he became pro. I would say Lee Sedol is two stones stronger than him.

        • I agree with you. Fan Hui and Lee are not even close. I don’t understand why the previous post even dare to compare them. The difference is so big, that if they play together, it will look like an professional adult player playing soccer with a small kid. In fact during Fan Hui’s match with AlphaGo, he played so badly that many professional Go players believe he did not even play at a entry professional level. Fan Hui’s level of play during the 5 games vs AlphaGo were like an amateur. That is why people are surprised with huge improvement of AlphaGo over the last few months, because Lee is orders of magnitude stronger than Fan Hui.

  70. If alpha go is still improving isnt it scary. Δ°s there a room for devrlopment? And if so whats the limits?

    • Another IT guy says:

      Yes, there’s a lot of room for improvement for AlphaGo, both in terms of skill and in terms of efficiency. Now it requires quite a lot of resources to run, but I predict that it will be possible to run similarly powerful Go bots on your laptop in the near future.

  71. Pierre Maurier says:

    Is there a kifu of the game somewhere?

  72. I wonder if this will have an effect on 2 day matches in Japan.
    In chess tournaments, before the computer became so strong, some games were aborted and continued the next day, when computers became too strong this

    • Anonymous says:

      This is GO… I am sure that any player who gets to play in a 2 day match, will not use a computer to get help.

  73. pramodit adhikari says:

    does any one have an sgf of the game?

  74. Bill Harris says:

    They coulda left off the loud, stupid-assed intro. This is “Go”, not NASCAR.

  75. Great video but could you please stop the attenuating screen saying there are 0 seconds left before game start, that is really annoying.

  76. Jacques Casanova says:

    The match is five games no matter who wins or just best of five?
    Will the loser of the match receive any money?

  77. steve4427 says:

    haven’t found the kifu yet

Speak your mind