China Daily | AI beats humans at go - video Dailymotion

The Star

What happens when computers beat humans at a game?  WATCH MORE: https://thestartv.com/c/news SUBSCRIBE: https://cutt.ly/TheStar LIKE: https://fb.com/TheStarOnline

Transcript

00:00 What happens to a game when humans are no longer its best players?

00:05 The game would become interesting again.

00:08 That was the New York Times conclusion when chess became the domain of computers in 1997.

00:13 But we're not here to talk about ancient history.

00:16 We're here to talk about Go.

00:18 Go is a two-player game, much like chess, where players place pieces on a board.

00:24 It was thought to be one of the last games that humans could still beat computers at,

00:28 given the fact that I'm talking about it here, you can probably guess that's no longer the case.

00:33 And there's only one rule for Go.

00:45 After the four consecutive points was occupied, this stone is dead, and it will be taken away from the board.

00:54 These simple rules don't make it easy though.

00:57 There are so many scenarios in Go, like 10 to the power 81.

01:04 That's actually the amount of atoms in the universe,

01:06 and Go has so many combinations that even that number looks small in comparison.

01:10 As I remember, in 2017, AlphaGo was investigated by Google, and it beat the world's best players.

01:23 It's essentially impossible to solve Go like you'd solve tic-tac-toe.

01:27 So how do they do it?

01:29 AlphaGo can have self-learning.

01:31 He can play with himself millions of games per second.

01:35 That's the reason he can find out which move is the best.

01:39 Before AlphaGo, AI was not even close to the top professional players.

01:45 This one is the best move, and you can see the future scenario that he keeps running.

01:52 The AI was saying that white have 69.7% to win the game.

01:59 Let's see if white not playing here.

02:02 Now black has 53% to win the game.

02:06 There's a number of ways. Oh, looks like AlphaGo has resigned.

02:10 Wow.

02:11 We are all happy that human win one game.

02:16 Around 2020, those AI companies can set some challenge,

02:22 but the best player can only win them after they give them two stones.

02:28 If give away one soldier before start the game, so you can win them.

02:32 But if we are start with even, no.

02:36 Like I play game in the Asian game, I will review on the AI software.

02:45 It also can be show on phone.

02:48 So it is quite convenient that after we just finish the game,

02:53 I can just take my phone and just review my game.

02:57 You need nine stones to capture same.

03:01 You can capture nine points.

03:04 Maybe I would try here.

03:06 Oh, that is the best move.

03:08 Yeah, that is the best move.

03:13 It is cheaper if you're reviewing on the AI software.

03:17 Before it's like 800 per hour with those professional player.

03:23 But now it takes only 15 dollars per hour.

03:27 But sometimes they just, AI would just put it something here and say it is the best move.

03:35 For this case, it is real hard to estimate.

03:38 If you put it here, 5% decrease. Here, 10%, something like this.

03:43 But so many place you can place here that you have the same purpose.

03:48 Maybe one day you'll be able to understand why.

03:51 If I can understand, maybe I will become the best player.

03:56 But we cannot understand sometimes the AI move.

03:59 And for the development, if he can explain, that will be amazing.

04:06 The New York Times prediction did mostly come true.

04:09 Pro Chess Today makes use of strategies and tools that 90s grandmasters would never have dreamt of.

04:14 That's also mostly the case with Go.

04:17 I think with the AI, the distance between the professional player and the amateur player will be much closer.

04:25 For example, the professional player, they will play with professional player.

04:29 So they got a much better environment to improve their skill.

04:33 Now we all have AI, so the distance will be shortened.

04:37 But for chess, like with Go, AI tools have meant that cheaters are now in the midst.

04:42 It's a scandal rocking the chess world.

04:45 I feel more comfortable when I play face to face because you never know the opponents online.

04:52 If they use AI, I cannot beat them.

04:54 Cheating did not happen in professional match, but it happened in some amateur game in the early stage.

05:01 And for now, for example, like the Asian game, you need to lock your cell phone into the locker before you enter the place.

05:10 Sometimes if one player was playing too good in the professional game,

05:16 although we know that he cannot cheat because there are no cell phones can be bring,

05:21 but we will say, oh, you just play like AI.

05:24 That's the end of our match together for this week's Tech Asia.

05:29 I'm not much of a Go player myself, but I'm pretty sure the best move would be to subscribe to our channel.

05:34 That will allow you to see next week's episode, which will be about the future of virtual reality in Asia.

05:38 If you just can't wait that long, well, I suppose you could check out last week's episode,

05:42 which was about the use of AI in medical imaging.

05:44 Until then, I suppose, it's the checkmates.

05:47 I know, it's not Go. Take care.

05:49 (upbeat music)

05:51 (upbeat music)

China Daily | AI beats humans at go

Category

Transcript

Recommended