Go Commentary: Lee Sedol vs AlphaGo – Game 1

If there’s any recent game that needs no introduction, it’s this one.

On March 9, 2016, the computer Go program ‘AlphaGo’ defeated Lee Sedol 9p in the first game of the Google DeepMind Challenge Match.

AlphaGo-Lee-Sedol-Aja-Huang

Lee Sedol (right) plays AlphaGo in the first game of the Google DeepMind Challenge Match. Aja Huang (left) places stones for AlphaGo.

The match was widely reported in mainstream media around the world, as well as by Go Game Guru. You can read more about game one of this five game match here.

Let’s have a look at the game!

Commented game record

Lee Sedol vs AlphaGo – Game 1

 

Download SGF File (Go Game Record)

 

Related Articles

About Younggil An

Younggil is an 8 dan professional Go player with the Korean Baduk Association. He qualified as a professional in 1997 and won an award for winning 18 consecutive professional matches the following year. After completing compulsory military service, Younggil left Korea in 2008, to teach and promote the game Go overseas. Younggil now lives in Sydney, Australia, and is one of the founders of Go Game Guru. On Friday evenings, Younggil is usually at the Sydney Go Club, where he gives weekly lessons and plays simultaneous games.

You can follow Go Game Guru on Facebook, Twitter, Google+ and Youtube.

Comments

  1. Thank you for a wonderful review of a very memorable game! ^^

  2. Uberdude says:

    Yay! more commentary on this game. In the variation for black 39 extend at h14, white kosumi j12, black enclose g16 and then white push and cut, I think white 48 doesn’t need to be o17 extend, but that she can extend to j18. If black then plays o17 to capture the 2 stones white sacrifices with l18 and m17 throw in and if she can then exchange g18 for f18 I think she wins the resulting semeai by 1 liberty, but tricky because if black can get n13 for n12 he gains an approach liberty. If that’s the case black would have to play n18 at j18 and white can get n18 in sente, which is a lot of points difference and painful for black. The same argument would explain why Lee didn’t play g16 for move 43, he may well have been happy with the exchange variation shown in which he gets the n18 double sente, but because he saw that white’s descent works (does it?) he realized he could only get the several points unconditionally worse variation with white getting n18, so went for the fighting variation of the game. Is this analysis correct?

  3. I hope you will be commenting some or all of the other games too, as in many ways Game 1 was the least interesting.

  4. Sebastian Heuchler says:

    Thank you very much for this game commentary, which was easy to follow and greatly helpful in understanding. It seems some moves received a slightly different judgement and more sophisticated explanation than seen before, which I found interesting. That being said, …

    sorry for this long comment and changing the topic, but I have a request to you, as an influential writer, with many Go players reading your articles: Please do not gender computer players as female. We ought to use an ‘it’ instead.

    Before I continue, please understand that I am not opposed to AI at all. I am a Computer Science student very fond of AI, doing very well in university, with my Bachelor’s thesis about studying the complex behavior of computer programs, my Master’s thesis likely about self-assembling networks, and various of my personal side projects about applying AI techniques to other games. So I’m definitely not against AI, rather the complete opposite.

    It has been common to make bots female recently, but I firmly believe it is wrong. For one, linguistically speaking, a computer and a computer program have always been ‘it’. While I don’t mean to offend anyone, in my experience, the only people to call e.g. Firefox a female or male (“She won’t let me click on that!”) are people who are rather new to computers or have not much knowledge on how computers work.

    But also, and this is my main point, from a social point of view: Giving AI a female appearance is sending a wrong signal about it being able to reason and understand the way a human would, which is misleading.

    That is not a real problem at the moment, and surely nobody will find ‘female’ Go bots offensive, but it is not as clear in other areas and/or in the future: Imagine your tax forms are handled by a computer program, and the program mistakenly made it so that you won’t get any money back. Would you be happy to be told that “Oh, she decided that.”? It was not a human that made this decision, but a program, and we should know about it, because humans and computers judge and handle situations differently and also make different errors – and offer a different range of responsibility for them. It gets worse if we consider that AI is used in medical and in military applications. You surely would not want to hear “That person was killed by her” when a computer program operated a device leading to the unfortunate event (and neither “by him”!). Sorry for this far-reaching example, but I believe it carries the point across.

    As your articles are very well received and read all over the world, many people would take up this habit of calling AIs female. I am afraid this lowers the bar and AI will be more anthropomorphized in the future even when it should not be. So I humbly request that you refer to computer programs as ‘it’.

    If you prefer, I can provide further references on this topic.

    Yours sincerely,
    Sebastian Heuchler

    • I agree about the “she” issue. He and she should be reserved for living objects of male or female sex. “It” has always been the word for non-living things, and it is the best indicator to make a distinction between a human player and a computer program. And at a certain point in time, not too far away I guess, a lady player will play a computer program: “she played this move and it responded correctly by connecting”.

      Kind regards,
      Paul

    • Hi Sebastian…I may be wrong, but typically white is usually referred to as feminine and black is usually considered masculine.

      Like I said, I may be wrong, but I don’t think they were calling AI a “girl”, just an old tradition coming into play

      • That may be true in Korean, but if so, it’s a case where direct translation doesn’t work: in English, the meaning is only obfuscated by the baggage surrounding the use of “she”.

  5. Thank you for this comment, it was quite insightful. And welcome back, I am glad Go Game Guru is back in business!

    Kind regards,
    Paul

  6. astrolabe says:

    I would like to know your perception of (March 2016) Alphago’s strength. Do you feel it was stronger than any human ever? Did you notice any consistent weaknesses apart from when it was comfortably ahead or behind?

  7. Testveritas says:

    Missed you guys. it’s been a while.

  8. Anonymous says:

    I have a question about move 102. Lee said that there was a computer move in game 1 that “no human Go player would have made”, which I believe is about move 102. I was wondering why this move is so unanticipated. To my amateur eye it looks like a feasible invasion point. Not that I could make it succeed, but I assume that Lee would have considered the invasion. In the live video commentary Chris Garlock at 2:22:25 also stated that he “wondered if this move (white 102) was possible”.

    • Uberdude says:

      Apparently Zhou Ruiyang 9p suggested that invasion during live commentary, so the journalists saying AlphaGo plays all these amazing moves humans would never think of are exaggerating; what is remarkable is that rather than needing a large group of pros to come up with the good ideas it has been compressed into a single player that chooses them wisely with good judgement and timing. Same goes for the 5th line shoulder hit of game 2: Takemiya has been doing high shoulder hits for decades and to suggest AlphaGo was revolutionary in this regard is simply ignorant of the history and variety of human play. The armpit hit of game 3 though was perhaps special, but even so I expect some pro might have found it given enough time.

    • gil dogon says:

      Yes, I also wondered about 102 being called such “genius/unexpected” move. I wonder which other move for 102 would have been the “normal” move. I guess it all rests on your confidence of reading the multitude of variations , as any invasion carries risk, but Whites thickness above practically commands him to invade.
      Indeed 80 is Alphago’s most puzzling move, and Anns analysis shows B h13 does not work well, but maybe there is some other weakness/aji in the position that AG did not like …

  9. Nice to have you back. I was worried if you had abandoned the site :). I was consoling myself that you were busy with games and other work and return in no time.

    After the game with AlphaGo, LSD is in a winning streak. Do you see any substantial change in Lee’s way of playing in recent games. Some of the games, i felt he was playing thin. This is only my guess biased by thought that like how AlphaGo with its almost perfect calculation would not mind playing all over the place, may be LSD was exploring similar strategy. You would have an insights as well as knowledge of his previous games and playing styles. I hope when the time is ripe you would comment if there are any changes.

    Again, its nice to have you back.

  10. Such a fine and comprehensive commentary, thank you!

  11. Great commentary, thank you! I don’t have a problem with W being referred to as ‘she’ in this case – nor in fact with Alpha Go being referred to as a female player. I think it’s great for possibly the strongest player in the world being labelled female! 🙂 I also hope there will be commentaries of the subsequent games.

  12. Thank you for this amazing commentary. I also missed your articles and I’m glad that you are back.

    There has been so much interest around alpha go, I’m not obsessed with alphago myself and rather follow the international matches.

  13. Mate Matolcsi says:

    Thanks you for this nice commentary. This was a very interesting game. I have two questions:
    — if black peeps at O9 at move 93, can white answer the peep in any other way than N9?
    — if black plays R4 at move 121, do you think he is still slightly losing?
    Thanks again,
    Mate

  14. Anonymous says:

    Thank you for the helpful commentary.

Speak your mind