AI vs. Human: The Greatest Go Tournament Ever

AI vs. Human: The Greatest Go Tournament Ever

For fans of artificial intelligence, the past
few weeks have been very exciting. Last October, for the first time ever, an
AI built by Google, known as AlphaGo, beat a professional human player at the incredibly
complex board game Go. And then, in this last week, it went on to
beat one of the best players in the world … four times. Which is a really big deal. And let us explain
to you why. When engineers talk about artificial intelligence,
they don’t mean sentient humanoid robots conversing with us, talking to us, becoming
feeling, then taking over the world destroying us. AI is really just a way of programming computers
to do things that humans normally do. See, computer programs follow a specific list
of instructions to complete tasks, so they struggle with situations that have lots of
options and require decision-making on the fly. Including playing complicated board games,
like Go. The basic objective of Go is to get as many
points as possible, either by capturing the other player’s pieces or by claiming areas
of the board with black or white stones. And a big part of the winning strategy for
any game is the ability to think a few moves ahead, taking into account what the other
player might do each time you make a move. We humans can use our past experiences with
the game to figure out the best moves, and how our opponents might respond to them. But it’s hard to build that kind of pattern
recognition into a computer program. So generally, computers play board games — like
chess — by searching through all the possible combinations of moves to find the one that
means it’s most likely to win. The problem is, Go is played on a huge 19-by-19
grid — and there are hundreds of possible moves that a player might make in every turn. In fact, there are more ways for a game of
Go to play out than there are atoms in the universe. So if it’s that difficult to program an
AI to beat humans at Go, how did engineers teach AlphaGo to do it? Well, instead of having the AI search through
all the possible combinations of moves they tried to help it understand the difference
between a good move and a bad one. To do that, the engineers first fed AlphaGo
30 million combinations of moves, taken from real games with expert human players. Then, AlphaGo played thousands of matches
against itself, to learn new strategies. All that knowledge, plus some clever programming,
helps the AI decide on the best next move. Instead of having to consider every single
possible move, which would take a really long time, it can quickly narrow down the few,
most relevant options. And so far, this strategy has worked really
well against us humans. A few months ago, AlphaGo played a 5-match
tournament against Fan Hui, the European Go champion, and won every single game — the
first time a computer had ever won against a professional Go player at all. Then, Google decided that it was time to test
the AI against Lee Sedol, a South Korean who’s been the top Go player in the world for the
past decade. They livestreamed all five matches, and uploaded
15-minute summaries — which are all linked in the description below. And — spoiler alert — AlphaGo won the first
three, so best out of three of five: it won the tournament. So it was pretty clear that the way the engineers
programmed and trained the AI did work. But AlphaGo and Sedol were still set to play
the last two games in the series. And Sedol actually won the fourth game — which
was a huge deal, because it showed that the AI still isn’t perfect at choosing the best
moves. A big turning point in this game was when
Sedol played a move known as a wedge — one that has lots of possible responses. Basically, he was trying to confuse the AI
by giving it too many options to explore. And that seems to have worked, even after
all that training. After the wedge, the game went downhill for
AlphaGo, until its internally-calculated chances of winning went below 20% — at which point
it’s programmed to resign. Game 5 turned out to be a very close game,
with AlphaGo making a mistake early on, but eventually winning. So AlphaGo might not be the best Go player
in the world right now — but it’ll just keep getting better from here, so odds are,
eventually, it will be. Either way, it’s a huge step forward for
artificial intelligence. Thank you for watching this episode of SciShow
News, and thank you especially to all of our patrons on Patreon who make this all possible.
Thank you so much for being a patron if you are. If you want to become one of those people
you can go to­ there’s a bunch of cool things you can get there as
well. And don’t forget to go to and subscribe!

100 thoughts on “AI vs. Human: The Greatest Go Tournament Ever

  1. I used to play this game all the time when I was a kid! I never realized how complicated the game would be for a computer

  2. That's why we humans will never be beaten by AI. We won't give up, even if our chances are less than 20%. #GoJohnConnor!!!

  3. Why, does my penis love love love vagina??? Seriously tho it never wants to leave, just wants to soak, for eternity! Why????

  4. And here we have the great divide between human intelligen "sentience" and artificial intelligence "programming".
    The go champion might be beatenby the machine in a game, but then he will go home, an extremely complicated calculation and series of action that might very well end in his death in a car crash. Then he will make himself dinner, socialize and interact with other people, work, and be in essense self reliant. While the machine just sits there, waiting for the next time it will get to do the only thing it can, doing nothing in the meantime.

  5. Just remember, although AlphaGo may be exceptionally well at playing Go, that's all it's good for. Other than doing anything else a normal computer can do, and playing Go, that computer can't do much else. Computers are built to do specific tasks, not every possible task.

  6. Why would you say that creating robots who are like humans and who could run the world or even take over isn't what AI is about? If even Elon Musk disagrees with you, you might not want to say that. AI is also the things you talk about in this video, but you are perpetuating, honestly in my view a crime by trivializing it. AI is awesome, but people need to be aware of all its potential risks. Don't minimize them or pretend they aren't real, they are very real. So please, take some personal responsibility over this. Ideally re shoot the video so that what you are saying is actually factual. Thanks.

  7. They programmed it to resign the game after only dropping below 20% chance of winning? Clearly wasn't programmed by people who play sports or go to the gym hahaha

  8. Final comment: why not actually talk about machine learning algorithms in this a little bit? Machine learning is one of the most simple cutting edge technologies to understand. You easily could have explained how AlphaGO works in more depth without adding more than 30 seconds to the vid.

  9. I have been playing since I was 4, but I ended up losing interest after I stopped playing for a while. Then I heard about alpha-go and the Google deepmind challenge. I watched all the matches, and they were amazing; all the moves were so well thought out! It got me playing Go again! 🙂

  10. it's not AI if it's been programmed a specific set of methods written by humans

    most "AI" isn't true AI… whwwwwhyyyyy don't people understand this

  11. It is so funny how humans see the first time, their own behavior. Its like the Google Daydream A.I. -> It is actually for most humans more important, what kind of memories (aka fantasies) they have about other people, then the actual truth. This is why I think, that people love material from dead people. Because as long as they can not ask them, it stays a fantasy inside their head, called memory.

  12. +SciShow my question is, what does this really mean for A.I.? Does it just mean its really good at a certain game or games of that type? Or can this method of not thinking too far ahead be used for other things like, saving people from a burning building, or working with humans on a construction project? I guess i'm asking, what use is this A.I.?

  13. I'm really proud of both of those Go players. Especially the last one. Kudos to him for being so inexplicably smart.

  14. Think about a Go-playing program that could have been made in 1996. It would have sucked at the game. Think of how much better AlphaGo is at playing Go than that program would have been.

    Now, think about a Go-playing program that could be made in 2036 (assuming computing progress doesn't fizzle out by then). Think of how much better this future program will be at playing Go than AlphaGo is. Then consider how pointless it would be for a human to play against it with any real expectation of winning.

  15. imagine if this AI was using quantum computing. It could do every single possible move each turn no problem.

  16. Think AlphaGo was shown THOUSANDS of human-played games, then set to play against itself 30,000,000 times, nor the other way around. In any case, many, many thanks for all the videos. I am a fan.

  17. AI includes strong AI or general intelligence which is exactly what you said it isn't, intelligent computers/robots. Short of studying the field itself, Wikipedia would have told you that.

  18. The fact that it lost game 4 is incredibly important and nullifies the fact it won the first 3 games. Basically, alphaGo plays against itself multiple times plus it beat the EU champion 4 out of 4, before winning first 3 against Lee Sondol. At this point the A.I. has far better "inventory" and it should never lose again, yet it did.
    Also the human brain is doing far more things outside playing Go, like regulating heart beat, maintaining posture, coordinating movement, chemical homeostasis, listening to the environment etc.. than it is doing playing Go. Yet it was able to win even 1 match. It shows just how far advanced human potential is. Even during the Go match the mind only dedicates a fraction of computing to winning so imagine life away from the Go board? Meanwhile AlphaGo is always and only working on improving it's Go skills.

  19. Something small he said needs a correction: AI absolutely does not play chess by assessing all of the possible options. Even though the scale of chess is much smaller to that of Go, it is still unfathomably large. The algorithms that AI use to play chess are just really clever and involve many different metrics for determining the success/failure of a given state in the game.

  20. IN 2017, upgrade version of AlphaGO beat all top class human player in the world by 60:0 in about a week. Only 9 month after this video was made.
    This time, no sign shows human has any chance.

  21. isnt it more impressive the guy managed to outsmart the computer after learning how it played and its weaknesses from the first the games and manipulated it in the fourth and almost won in the fifth with the disadvantage of going second whcih he asked for

  22. Just realized that that guy just won against one of the most advanced A.I ever programed… what a genius

    Sorry for the bad english

  23. Let me tell you the future. Humans almost went extinct from evil robots but someone found a glitch that if you hold your breath, the robot will think you are dead and ignore you. That's why computers will never be anything serious.

  24. the reason the AI will always be the best is that it never gets tired and can run simulations constantly. a human could never keep up with the sheer volume of games that a computer can run. all the computer does is play go, humans actually have to live a life.

  25. The danger with AI is when you give it self learning abilities. If a human can learn to deny rules. self learning AI will do it faster.

  26. Quick update from the future. AlphaGo zero won 100 out of 100 games against the version of AlphaGo that defeated Lee Sedol.

  27. in quantom computing ai would know all combinations at once. once ai and quantum computing marry we are all doomed!

  28. I keep coming back to this channel and remember how you edit all the short pauses between sentences out. I absolutely hate it. This is not how natural speech sounds and it's jarring. I could only listen to the first 17 seconds before I gave up.

  29. i have seen this idea presented over and over again that chess programs(engines) look at all possible moves. simply put – this is not true. an exhaustive search in chess would also be prohibitive as it is in go. because chess is played on an 8×8 board instead of go on a 19×19 grid it does lend itself to a broader look ahead but the idea being presented here is terribly misleading.

  30. I'm all for tech advancement, but not necessarily if it has the potential to send humans back to the stone age or worse… If the AI can think ahead, we have to as well to compete with our own creation. Let's make sure they have some kind of pneumatic or low-fi failsafe included at least.

  31. I hope AIs don't automate everything or get the permissions and powers of SkyNet. I don't want that kind if hell. I could settle for an AI that I can play games with me or teach me how to play them when I am super bored.

  32. I don't know about Go but the description that chess engines check every possible combination of moves from every position is incorrect. There are simply too many possible combinations to check them all in a reasonable (i.e. less than several years). So all chess engines use heuristics to prune their search trees to the most promising moves, in the order of 5 to 8, and check only the combination of these top moves.

  33. They already made, AlphaGo Master and AlphaGo Zero versions. It is now at level Dan 10, which no human can be at consistently if at all. Therefore, AlphaGo can't be defeated. Dan 10 level, didn't exsist, until I made it, because the top human rank level is Dan 10 and obviously it is much above that.

  34. Why did he say Alpha go isn't the best Go player in the world right now. It beat the worlds best player 4 games out of 5. Which by the way go is more complex than the stock market. It's confirmed Google has a money printing machine. Better buy some Alphabet Inc.

Leave a Reply

Your email address will not be published. Required fields are marked *