AlphaGo - The Movie | Full award-winning documentary

TL;DR

• DeepMind's award-winning 90-minute documentary (on YouTube since 2020, 38M views) on AlphaGo: from beating European champion Fan Hui 5-0 in secret (Oct 2015) to the Google DeepMind Challenge Match against Lee Sedol in Seoul (March 2016), with 80-100M people watching. AlphaGo won 4-1, about a decade before experts expected Go to fall.
• How it works, in the film's words: first imitate about 100k strong amateur games, then improve through millions of games of self-play reinforcement learning. Three parts: a policy network proposes moves, a value network estimates the probability of winning, and a tree search looks 50-60+ moves ahead. It maximises the probability of winning, not the margin, so "slow" moves signal it thinks it's ahead.
• Move 37 (game 2): a fifth-line shoulder hit AlphaGo itself rated at a 1-in-10,000 chance of a human playing it. Commentators thought it was a mistake, then watched it tie the board together. The film's thesis: learning systems can "go beyond what we as the programmers know."
• Move 78 (game 4): Lee's "God move" (AlphaGo also estimated it at about 1 in 10,000) pushed AlphaGo into one of its known "delusion" blind spots. It played nonsense moves, its value estimate collapsed, and it resigned: humanity's only win, celebrated in the streets.
• The human arc: Fan Hui goes from being crushed ("I don't understand myself anymore") to AlphaGo's advisor and finding its weaknesses. Lee plays "for the human" and is changed by the match. Hassabis frames it as "a human endeavor", and the closing idea is that AI can expand human understanding, as the slack moves are already rewriting Go theory.

Key moments

• [00:18] Cold open — a black screen for 18s, then voices on Go as "putting your hand on the third rail of the universe" and wanting "to understand what understanding is."
• [01:25] Hassabis and games as AI testbeds — the Atari Breakout agent learning from pixels and discovering the tunnel strategy its creators didn't know. "Go is the most complex game pretty much ever devised by man."
• [04:43] Fan Hui is invited — Demis's email; the secret best-of-five in London (frame_0006, subtitle "The second game, I tried to change my style"); AlphaGo wins 5-0. "I lose this program and I don't understand myself anymore."
• [11:50] How AlphaGo learns — imitation of 100k amateur games, then self-play reinforcement learning; deep neural nets "mimic the web of neurons."
• [13:55] Lee Sedol is chosen — "the Roger Federer of Go", 18 world titles; he predicts 5-0 or 4-1 for himself. Seoul, Korean Go culture, his childhood Go school (frame_0009).
• [23:23] The "delusion" weakness — Fan Hui finds lumps of knowledge where AlphaGo misjudges life and death; version 18 is frozen and flown to Seoul with known weaknesses.
• [29:18] Game 1 — AlphaGo team control room (frame_0014), live commentary (frame_0017); Lee resigns: "so shocking."
• [47:13] Inside AlphaGo — Thore explains the policy net, value net and tree search, and that AlphaGo maximises win probability.
• [49:35] Move 37 — a fifth-line shoulder hit rated at 1 in 10,000 for a human; Lee leaves the room, comes back, and thinks for more than 2 minutes; AlphaGo wins game 2.
• [58:17] Game 3 — AlphaGo takes the match 3-0: "we are facing the strongest existence ever in Go history." Lee walks out (frame_0026). Hassabis: "I couldn't celebrate."
• [64:08] Game 4, move 78 — Lee's "wedge" (Fan Hui: "the wolf waits in the forest in the winter"); AlphaGo "goes on tilt", its value estimate collapses, and it resigns; the team laughs in relief (frame_0032); Korea celebrates.
• [76:05] Game 5 and aftermath — AlphaGo's "slack moves" reread as win-probability play; AlphaGo wins 4-1. Move 37 "begat" move 78. Lee (frame_0039): "he showed me something... maybe it's beautiful."

Hook microscope (0-10s)

• Frames: 20 at 2 fps

  Read these hook frames with the Read tool:

  • hook_frames/frame_0001.jpg (t=0.0s)
  • hook_frames/frame_0002.jpg (t=0.5s)
  • hook_frames/frame_0003.jpg (t=1.0s)
  • hook_frames/frame_0004.jpg (t=1.5s)
  • hook_frames/frame_0005.jpg (t=2.0s)
  • hook_frames/frame_0006.jpg (t=2.5s)
  • hook_frames/frame_0007.jpg (t=3.0s)
  • hook_frames/frame_0008.jpg (t=3.5s)
  • hook_frames/frame_0009.jpg (t=4.0s)
  • hook_frames/frame_0010.jpg (t=4.5s)
  • hook_frames/frame_0011.jpg (t=5.0s)
  • hook_frames/frame_0012.jpg (t=5.5s)
  • hook_frames/frame_0013.jpg (t=6.0s)
  • hook_frames/frame_0014.jpg (t=6.5s)
  • hook_frames/frame_0015.jpg (t=7.0s)
  • hook_frames/frame_0016.jpg (t=7.5s)
  • hook_frames/frame_0017.jpg (t=8.0s)
  • hook_frames/frame_0018.jpg (t=8.5s)
  • hook_frames/frame_0019.jpg (t=9.0s)
  • hook_frames/frame_0020.jpg (t=9.5s)

• Word-level transcript (1 words):

  [  6.70s] 😘😘😘😘😘😘😘

0.0-10.0s: pure black (every hook frame is black, frame_0001 included) under ambient score. There's no speech: the single Whisper "word" at 6.7s is a hallucinated row of emoji on silence. The first line lands at 0:18 ("it's intensely contemplative"), then "putting your hand on the third rail of the universe."

Pattern: atmospheric cold open / delayed reveal. Classic feature-documentary grammar. It earns patience with silence and portentous voiceover about what Go means, and holds the machine back until after the human stakes are set. The hook microscope's density analysis doesn't apply to cinema pacing like this.

Editorial profile

• Shots: 40
• Cuts/min: 0.44
• Mean shot length: 135.69s
• Median shot length: 130.44s
• Talking-head ratio: n/a (opencv not installed)

Feature documentary (dir. Greg Kohs): cinéma-vérité in the London office and the Seoul control room, bookshelf-lit interviews with lower-third name titles, archival match broadcast graphics and news clips, subtitled Korean and English, orchestral score. (The 0.44 cuts/min figure is an artifact of "balanced" mode's 40 evenly spaced samples; the film cuts much faster.)

Quotable moments

• [10:32] Fan Hui: "I feel something very strange. I lose this program and I don't understand myself anymore."
• [12:43] "The whole beauty of these types of algorithms is that because they're learning for themselves, they can go beyond what we as the programmers know how to do."
• [51:02] "AlphaGo said there was a 1 in 10,000 probability that move 37 would have been played by a human player... it went beyond its human guide."
• [56:57] "We're really closer to a smart washing machine than Terminator."
• [84:31] "Move 37 begat move 78 begat a new attitude, a new way of seeing the game. He improved through this machine; his humanness was expanded."

Entities mentioned

• People: lee-sedol, fan-hui, Demis Hassabis, David Silver, Aja Huang, Thore Graepel, Julian Schrittwieser, Michael Redmond, Chris Garlock, Janice Kim, Andrew Jackson, Martin Rees, John Daugman, Garry Kasparov
• Companies: DeepMind (Google DeepMind), Google, IBM, Microsoft, Korea Baduk Association, American Go Association
• Tools / products: alphago (policy network, value network, tree search), Deep Blue, Atari Breakout agent (DQN)
• Places: London, Seoul (Four Seasons hotel)

Concepts surfaced

• move-37: a machine-creative move its own policy network rated at 1 in 10,000 for a human, and correct. Learning systems can exceed their human training prior; the archetype for reasoning that emerges from RL (rlhf-is-not-rl).
• superhuman-blind-spots: "tricky lumps of knowledge" where a mostly superhuman system is confidently delusional (misjudging dead groups as alive). Lee's move 78 found one, and it's hard to predict when the system will wander into one.
• win-probability-not-margin: AlphaGo optimises the chance of winning and ignores the score, so its "slack moves" look lazy to humans. Go players had been using margin as a proxy for safety; the machine shows they're different objectives.
• Imitation then self-play: bootstrap from human data (100k amateur games), then improve beyond it through millions of games of RL self-play; the pipeline behind modern reasoning models (rlhf-is-not-rl).
• Anthropomorphizing AI: commentators unconsciously call AlphaGo "he" or "she"; the "Terminator picture" distorts risk discussion when "we're closer to a smart washing machine."
• Human + machine (Kasparov): "a good human plus a machine is the best combination." Both Fan Hui and Lee Sedol come out of the losses with new understanding.

Transcript

Source: captions.

[00:19] it's intensely contemplative
[00:35] it's like putting your hand on the third rail of the universe
[00:39] third rail of the universe if you play go seriously there is a
[00:42] if you play go seriously there is a chance that you will
[00:44] chance that you will get exposed to this
[00:48] get exposed to this experience that is kind of like nothing
[00:52] experience that is kind of like nothing else
[00:52] else on the planet go is putting you in a
[00:55] on the planet go is putting you in a place
[00:56] place where you're always at the very
[00:59] where you're always at the very farthest reaches of your capacity
[01:05] there's a reason that people have been playing go for thousands and thousands
[01:06] playing go for thousands and thousands of years right
[01:09] of years right it's not just that they want to
[01:10] it's not just that they want to understand go
[01:12] understand go they want to understand what
[01:14] they want to understand what understanding is
[01:20] and maybe that is truly what it means to be human
[01:26] be human i was a kid i loved playing games
[01:30] i was a kid i loved playing games i started off with board games like
[01:32] i started off with board games like chess
[01:33] chess and then i bought my first computer when
[01:35] and then i bought my first computer when i was eight with winnings from a chess
[01:37] i was eight with winnings from a chess tournament
[01:39] tournament ever since then i felt that computers
[01:41] ever since then i felt that computers were this sort of magical device
[01:43] were this sort of magical device that could extend the power of your mind
[01:51] virtual environments and games we think they're the perfect platform for
[01:52] they're the perfect platform for developing and testing ai algorithms
[01:55] developing and testing ai algorithms games are very convenient in that a lot
[01:58] games are very convenient in that a lot of them have scores
[01:59] of them have scores so it's very easy to measure incremental
[02:01] so it's very easy to measure incremental progress
[02:02] progress so i'm going to show you a few videos of
[02:04] so i'm going to show you a few videos of the agent system
[02:05] the agent system the ai so let's start off with breakout
[02:08] the ai so let's start off with breakout so here you control the baton ball
[02:10] so here you control the baton ball and you're trying to break through this
[02:11] and you're trying to break through this rainbow colored wall the agent system
[02:14] rainbow colored wall the agent system has to learn everything for itself
[02:16] has to learn everything for itself just from the raw pixels doesn't know
[02:18] just from the raw pixels doesn't know what it's controlling
[02:19] what it's controlling it doesn't even know what the object of
[02:21] it doesn't even know what the object of the game is now the beginning after 100
[02:23] the game is now the beginning after 100 games
[02:24] games um you can see the agent is not very
[02:26] um you can see the agent is not very good it's missing the ball most of the
[02:27] good it's missing the ball most of the time
[02:28] time but it's time to get the hang of the
[02:29] but it's time to get the hang of the idea that the bat should go towards the
[02:31] idea that the bat should go towards the ball
[02:32] ball now after 300 games it's about as good
[02:35] now after 300 games it's about as good as any human can play this
[02:36] as any human can play this and pretty much gets the ball back every
[02:38] and pretty much gets the ball back every time we thought well that's pretty cool
[02:40] time we thought well that's pretty cool but we left the
[02:41] but we left the system playing for another 200 games and
[02:43] system playing for another 200 games and it did this amazing thing it found the
[02:45] it did this amazing thing it found the optimal strategy was to dig a tunnel
[02:46] optimal strategy was to dig a tunnel around the side
[02:48] around the side and put the ball around the back of the
[02:49] and put the ball around the back of the wall the researchers working on this
[02:51] wall the researchers working on this amazing ai developers
[02:53] amazing ai developers but they're not so good at breakouts and
[02:55] but they're not so good at breakouts and they didn't know about that strategy so
[02:56] they didn't know about that strategy so they learned something from their own
[02:58] they learned something from their own system which is
[02:59] system which is uh you know pretty funny and quite
[03:00] uh you know pretty funny and quite instructive i think about
[03:02] instructive i think about the potential for general ai so for us
[03:04] the potential for general ai so for us what's the next step now
[03:06] what's the next step now go is the most complex game pretty much
[03:08] go is the most complex game pretty much ever devised by man
[03:10] ever devised by man beating a professional player at go is a
[03:12] beating a professional player at go is a long-standing
[03:13] long-standing grand challenge of ai research
[03:24] [Music] [Applause]
[03:25] [Applause] [Music]
[04:07] [Music] when i was 18 i won't change my life
[04:11] when i was 18 i won't change my life this is why i go to french i won't
[04:14] this is why i go to french i won't try to forget the goal
[04:17] try to forget the goal but it's impossible because all the
[04:20] but it's impossible because all the things i learned
[04:21] things i learned in my life is this visco
[04:24] in my life is this visco it looks like a mirror i see the goal i
[04:28] it looks like a mirror i see the goal i also see
[04:29] also see myself for me gold is a real
[04:33] myself for me gold is a real life
[04:34] life [Music]
[04:44] dear mr pham my name is demyster harbis i run an artificial intelligence company
[04:46] i run an artificial intelligence company based in london called deepmind
[04:48] based in london called deepmind as the strongest go player in europe we
[04:51] as the strongest go player in europe we would like to invite you to our offices
[04:52] would like to invite you to our offices in london
[04:53] in london both to meet you in person and to share
[04:55] both to meet you in person and to share with you an exciting go project that we
[04:57] with you an exciting go project that we are working on
[04:59] are working on if you would be interested in coming to
[05:00] if you would be interested in coming to visit us please let us know
[05:03] visit us please let us know any thanks kind regards demis
[05:08] any thanks kind regards demis when i see this email i don't know it's
[05:11] when i see this email i don't know it's true or not well i will accept about
[05:14] true or not well i will accept about this
[05:15] this why not for me
[05:18] why not for me iris is an adventure i want to go to
[05:22] iris is an adventure i want to go to visit deepmind to know what is
[05:25] visit deepmind to know what is this go project
[05:29] this go project [Music]
[05:31] [Music] the first visit i think maybe i want me
[05:34] the first visit i think maybe i want me sit in the special room push many many
[05:37] sit in the special room push many many wear
[05:37] wear in my hand also my body let me play
[05:42] in my hand also my body let me play to scan my spring i don't know
[05:46] to scan my spring i don't know to make some research we needed to get
[05:49] to make some research we needed to get fan way to
[05:50] fan way to deep mind to see we were a serious
[05:52] deep mind to see we were a serious operation and we were serious people
[05:54] operation and we were serious people um doing proper research
[05:58] um doing proper research and the search time is getting better
[05:59] and the search time is getting better and better like as we go from for
[06:00] and better like as we go from for example one to eight in any one of these
[06:02] example one to eight in any one of these it always go goes up we think of deep
[06:05] it always go goes up we think of deep mind as
[06:06] mind as kind of like an apollo program effort
[06:08] kind of like an apollo program effort for ai
[06:10] for ai our mission is to fundamentally
[06:12] our mission is to fundamentally understand intelligence
[06:14] understand intelligence and recreate it artificially and then
[06:16] and recreate it artificially and then once we've done that we feel that we can
[06:18] once we've done that we feel that we can use that technology to help society
[06:20] use that technology to help society solve all sorts of other problems if we
[06:23] solve all sorts of other problems if we step through the actual game we can
[06:25] step through the actual game we can see kind of what alphago thinks what's
[06:27] see kind of what alphago thinks what's the most likely variation that he thinks
[06:29] the most likely variation that he thinks will happen
[06:30] will happen we've been working on alphago our
[06:32] we've been working on alphago our program to play go for
[06:34] program to play go for just under two years now all the little
[06:36] just under two years now all the little patterns cascade together
[06:38] patterns cascade together layer after layer after layer after
[06:39] layer after layer after layer after layer
[06:41] layer i started talking about the game of go
[06:42] i started talking about the game of go with demis more than 20 years ago
[06:45] with demis more than 20 years ago and so this has been a really long
[06:47] and so this has been a really long journey the game of go is the holy grail
[06:49] journey the game of go is the holy grail of artificial intelligence
[06:51] of artificial intelligence for many years people have looked at
[06:53] for many years people have looked at this game and they've thought
[06:54] this game and they've thought wow this is just too hard everything
[06:57] wow this is just too hard everything we've ever tried in ai
[06:58] we've ever tried in ai it just falls over when you try the game
[07:00] it just falls over when you try the game of go and so that's why it feels like a
[07:02] of go and so that's why it feels like a real litmus test of progress
[07:04] real litmus test of progress if we can crack go we know we've done
[07:06] if we can crack go we know we've done something special
[07:11] so with fan hurry we started talking around what the real purpose of the
[07:12] around what the real purpose of the visit was
[07:13] visit was it wasn't just a go project we wanted
[07:15] it wasn't just a go project we wanted him to help with that actually we wanted
[07:17] him to help with that actually we wanted to play him
[07:18] to play him and we had a very strong program oh it's
[07:20] and we had a very strong program oh it's okay just the program is so easy
[07:23] okay just the program is so easy so we'll be easy to play a lot of people
[07:26] so we'll be easy to play a lot of people have thought that it was decades away
[07:27] have thought that it was decades away some people thought it would be never
[07:29] some people thought it would be never because
[07:29] because they felt that to succeed at go you
[07:32] they felt that to succeed at go you needed
[07:33] needed human intuition go is the
[07:36] human intuition go is the world's oldest continuously played board
[07:39] world's oldest continuously played board game
[07:40] game and in some sense it is one of the
[07:42] and in some sense it is one of the simplest
[07:43] simplest and also most abstract there's only one
[07:46] and also most abstract there's only one type of piece there's only one type of
[07:48] type of piece there's only one type of move you just place that piece on the
[07:49] move you just place that piece on the board
[07:50] board and then your goal is to create a
[07:53] and then your goal is to create a linked group of your stones that
[07:56] linked group of your stones that surrounds some empty territory
[07:58] surrounds some empty territory and when you surround enemy stones you
[08:01] and when you surround enemy stones you capture them
[08:02] capture them and remove them from the board you earn
[08:05] and remove them from the board you earn points by
[08:06] points by surrounding territory and at the end of
[08:08] surrounding territory and at the end of the game the person
[08:09] the game the person with the most territory wins it seems
[08:13] with the most territory wins it seems really simple
[08:14] really simple but then you sit down to play and you
[08:16] but then you sit down to play and you realize right away it's like well
[08:18] realize right away it's like well i technically know what i'm allowed to
[08:20] i technically know what i'm allowed to do but i have no clue what i should do
[08:22] do but i have no clue what i should do go's incredibly challenging for
[08:24] go's incredibly challenging for computers to tackle because
[08:26] computers to tackle because compared to say chess the number of
[08:27] compared to say chess the number of possible moves in a position
[08:29] possible moves in a position is much larger in chess it's about 20.
[08:32] is much larger in chess it's about 20. in go it's about 200
[08:34] in go it's about 200 and a number of possible configurations
[08:36] and a number of possible configurations of the board is more than the number of
[08:37] of the board is more than the number of atoms in the universe
[08:39] atoms in the universe so even if you took all the computers in
[08:41] so even if you took all the computers in the world and ran them for a million
[08:43] the world and ran them for a million years
[08:43] years that wouldn't be enough compute power to
[08:45] that wouldn't be enough compute power to calculate all the possible variations
[08:48] calculate all the possible variations if you ask a great go player why they
[08:50] if you ask a great go player why they played a particular move
[08:51] played a particular move sometimes they'll just tell you it felt
[08:53] sometimes they'll just tell you it felt right so we have to come up with some
[08:55] right so we have to come up with some kind of
[08:56] kind of clever algorithm to mimic what people do
[08:59] clever algorithm to mimic what people do with their intuition
[09:01] with their intuition nice to meet you hello so with fan
[09:03] nice to meet you hello so with fan hurray we agreed
[09:04] hurray we agreed a best of five match and we agreed it
[09:07] a best of five match and we agreed it would be filmed
[09:08] would be filmed you know we will treat it as a serious
[09:10] you know we will treat it as a serious match
[09:16] i play with alphago aja hwang push the stone
[09:16] stone for alphago and of course i think i will
[09:20] for alphago and of course i think i will win the game with alphago because it's
[09:22] win the game with alphago because it's just a program
[09:23] just a program [Music]
[09:33] the first game i make some mistake and i lose again
[09:41] in that first match i think something clicked for him that
[09:43] clicked for him that this wasn't an ordinary go program we
[09:46] this wasn't an ordinary go program we weren't just doing the
[09:47] weren't just doing the the same as everyone else that something
[09:49] the same as everyone else that something new was happening
[09:50] new was happening the second game i try to change my style
[09:54] the second game i try to change my style but the problem is also i lose the
[09:56] but the problem is also i lose the second game
[09:57] second game and also for third game first game
[10:01] and also for third game first game even the last game i lose
[10:04] even the last game i lose five zero
[10:11] alpha go win all the game after losing losing losing you can feel his
[10:14] losing losing losing you can feel his pressure is getting
[10:15] pressure is getting heavier and heavier and several times
[10:18] heavier and heavier and several times after the game he said
[10:19] after the game he said he wanted to go out for fresh air
[10:23] he wanted to go out for fresh air i said oh i can go with you and have a
[10:25] i said oh i can go with you and have a chat he said no i want to go myself
[10:33] i feel something very strange i lose this program and i don't
[10:36] i lose this program and i don't understand myself anymore
[10:45] i worried whether he'd even come back you know
[10:46] you know he seemed very low and he spent about an
[10:49] he seemed very low and he spent about an hour away from the office he came back
[10:50] hour away from the office he came back completely changed
[10:52] completely changed i thought maddie this is the first time
[10:54] i thought maddie this is the first time in the history
[10:56] in the history the human professional gold player lose
[10:59] the human professional gold player lose with the program she tell me yes we know
[11:02] with the program she tell me yes we know but for you i do i okay
[11:06] but for you i do i okay i told her
[11:12] not and yes i am not happy to lose the game but
[11:15] am not happy to lose the game but i'm very will be happy for play in the
[11:18] i'm very will be happy for play in the history
[11:24] artificial intelligence researchers has solved the game of
[11:25] solved the game of go a decade earlier than expected the
[11:28] go a decade earlier than expected the computer named alphago was able to beat
[11:30] computer named alphago was able to beat the european
[11:31] the european human champion artificial intelligence
[11:33] human champion artificial intelligence researchers have made a significant
[11:35] researchers have made a significant breakthrough
[11:36] breakthrough it really is a big leap forward there's
[11:39] it really is a big leap forward there's a big difference between
[11:40] a big difference between the way the ibm computer beats kasparov
[11:43] the way the ibm computer beats kasparov which was programmed by expert chess
[11:45] which was programmed by expert chess players and the way
[11:47] players and the way the go playing computer more or less
[11:49] the go playing computer more or less learned itself
[11:50] learned itself the way we start off training alphago is
[11:53] the way we start off training alphago is by showing it a hundred thousand
[11:54] by showing it a hundred thousand games that strong amateurs have played
[11:57] games that strong amateurs have played that we've downloaded from the internet
[11:58] that we've downloaded from the internet and we first initially get alphago to
[12:00] and we first initially get alphago to mimic the human player
[12:02] mimic the human player and then through self-playing
[12:03] and then through self-playing reinforcement learning it plays against
[12:05] reinforcement learning it plays against different versions of itself many
[12:06] different versions of itself many millions of times and learns from its
[12:09] millions of times and learns from its errors
[12:09] errors these specific ideas that are driving
[12:12] these specific ideas that are driving alphago
[12:13] alphago are going to drive our our future the
[12:15] are going to drive our our future the technologies at the heart
[12:17] technologies at the heart of alphago they're what are called deep
[12:19] of alphago they're what are called deep neural networks which essentially
[12:21] neural networks which essentially mimic the web of neurons in the human
[12:24] mimic the web of neurons in the human brain it's a very old idea
[12:26] brain it's a very old idea but recently due to increases in
[12:28] but recently due to increases in computing power
[12:29] computing power these neural networks have become
[12:31] these neural networks have become extremely powerful almost overnight big
[12:33] extremely powerful almost overnight big neural networks that
[12:35] neural networks that operate on big data can achieve
[12:38] operate on big data can achieve surprising things
[12:38] surprising things alphago found a way to learn how to play
[12:40] alphago found a way to learn how to play go learning is the key thing here it's
[12:43] go learning is the key thing here it's machine learning
[12:44] machine learning the whole beauty of these types of
[12:46] the whole beauty of these types of algorithms is that because they're
[12:47] algorithms is that because they're learning for themselves they can go
[12:48] learning for themselves they can go beyond what we as the programmers know
[12:51] beyond what we as the programmers know how to do
[12:52] how to do and allow us to make new breakthroughs
[12:54] and allow us to make new breakthroughs in areas of science and medicine
[12:56] in areas of science and medicine so alphago is one significant step
[12:59] so alphago is one significant step towards that
[13:00] towards that ultimate goal
[13:14] my wife called me she told me don't see internet don't connect internet
[13:17] see internet don't connect internet because
[13:18] because people talk terrible thing about you
[13:20] people talk terrible thing about you much with olivago
[13:22] much with olivago fanue is living so long time in europe
[13:25] fanue is living so long time in europe it's not real professional he's just an
[13:27] it's not real professional he's just an amateur player
[13:29] amateur player this is so hard for me so hard
[13:36] the go world were skeptical about how strong really was alphago
[13:38] strong really was alphago and how much further did it need to get
[13:40] and how much further did it need to get to be the top professionals
[13:42] to be the top professionals so our program is improving over time
[13:44] so our program is improving over time and
[13:45] and we want to push the ai algorithm to the
[13:47] we want to push the ai algorithm to the limit and see how far
[13:48] limit and see how far this kind of self-improving process can
[13:50] this kind of self-improving process can go
[13:52] go so we needed to look for an even greater
[13:54] so we needed to look for an even greater challenge
[13:56] challenge a match like no other is about to get
[13:58] a match like no other is about to get underway in south korea
[14:00] underway in south korea lee c dole will be long-reigning
[14:09] alphago and the ultimate human versus machine
[14:09] machine smackdown this is a huge moment for both
[14:12] smackdown this is a huge moment for both the world of artificial intelligence
[14:14] the world of artificial intelligence and i think the world of go so far
[14:17] and i think the world of go so far alphago has beaten every challenge we've
[14:19] alphago has beaten every challenge we've given it
[14:20] given it but we won't know its true strength
[14:22] but we won't know its true strength until we play somebody who is at the top
[14:24] until we play somebody who is at the top of the world
[14:25] of the world likely to don we chose lisa dole because
[14:28] likely to don we chose lisa dole because he wanted a legendary historic player
[14:31] he wanted a legendary historic player somebody who has been acknowledged as
[14:33] somebody who has been acknowledged as the greatest player of the last decade
[14:45] i will not be too arrogant but i don't think that it will be a very close match
[14:47] think that it will be a very close match the level of the player that africa went
[14:51] the level of the player that africa went against in october is not the same level
[14:53] against in october is not the same level as me
[14:53] as me so given that a couple of months was
[14:56] so given that a couple of months was only passed
[14:57] only passed i don't think that it is enough for it
[14:59] i don't think that it is enough for it to be able to catch up with me
[15:01] to be able to catch up with me my hope is that it will be either five
[15:03] my hope is that it will be either five zero for me
[15:04] zero for me or maybe four to one so the critical
[15:07] or maybe four to one so the critical point for me is to make sure i do not
[15:09] point for me is to make sure i do not lose one
[15:11] lose one higher we don't know
[15:14] higher we don't know how well our system will play against
[15:16] how well our system will play against someone as
[15:17] someone as creative as lisa doll also he's very
[15:19] creative as lisa doll also he's very famous for
[15:20] famous for very creative um fighting play yeah so
[15:24] very creative um fighting play yeah so this could be um difficult for us but
[15:26] this could be um difficult for us but we'll see maybe in some ways he's the
[15:28] we'll see maybe in some ways he's the most difficult opponent we can take
[15:30] most difficult opponent we can take there was still so much of a question
[15:33] there was still so much of a question about whether or not
[15:34] about whether or not they could beat someone like lee sedol
[15:37] they could beat someone like lee sedol von way is a good player but he's
[15:39] von way is a good player but he's nothing like
[15:40] nothing like the very top players he said all is a
[15:43] the very top players he said all is a nine down professional
[15:44] nine down professional nine don professional is the highest
[15:46] nine don professional is the highest rank that you can achieve and go
[15:48] rank that you can achieve and go the ranking system fanfare who's the
[15:50] the ranking system fanfare who's the european champion
[15:51] european champion is just a two don professional on the
[15:54] is just a two don professional on the other hand
[15:54] other hand the very top he said all is nine down
[15:57] the very top he said all is nine down professional
[15:58] professional they said all is to go what roger
[16:00] they said all is to go what roger federer is to tennis he's
[16:01] federer is to tennis he's playing in wimbledon to win the grand
[16:03] playing in wimbledon to win the grand slam and it's not just this year he'll
[16:04] slam and it's not just this year he'll be there next year and the year after
[16:06] be there next year and the year after that
[16:06] that and the year after that of course in
[16:09] and the year after that of course in internet
[16:10] internet and all the gold community everyone
[16:14] and all the gold community everyone maybe not 100 maybe 99.99
[16:19] maybe not 100 maybe 99.99 think listener will win very easy
[16:23] think listener will win very easy [Music]
[16:37] [Music] go is taken seriously in korea it's so
[16:40] go is taken seriously in korea it's so much part of the culture like breathing
[16:42] much part of the culture like breathing or you know like
[16:43] or you know like taking swimming lessons or something on
[16:46] taking swimming lessons or something on the surface is a game
[16:47] the surface is a game but inside had a very deep philosophy
[16:51] but inside had a very deep philosophy goalboard reflects the individual who's
[16:54] goalboard reflects the individual who's playing
[16:55] playing the true gonna show itself on the board
[16:57] the true gonna show itself on the board you won't be able to hide it
[17:00] you won't be able to hide it in ancient china japan korea
[17:03] in ancient china japan korea go is one of the four noble things
[17:07] go is one of the four noble things like for accomplishment for literacy
[17:11] like for accomplishment for literacy with music poetry and painting
[17:14] with music poetry and painting so people think the go players
[17:18] so people think the go players are very smart and very noble uh
[17:26] [Music] master guan started this school so that
[17:29] master guan started this school so that he could
[17:30] he could produce great prayer in korea he said
[17:33] produce great prayer in korea he said when he was age of eight he came in here
[17:36] when he was age of eight he came in here attending class 9 a.m till 9 p.m
[17:40] attending class 9 a.m till 9 p.m seven days and then he stayed with the
[17:42] seven days and then he stayed with the master one
[17:45] master one already
[18:06] was a little boy that i remember when i was a student
[18:07] was a student studying go he was very young and
[18:11] studying go he was very young and rural enough to think that pizza grew on
[18:13] rural enough to think that pizza grew on trees
[18:37] [Music] plays things that are interesting where
[18:40] plays things that are interesting where you felt
[18:40] you felt like he's beyond winning and losing he
[18:42] like he's beyond winning and losing he wants to do something that's
[18:44] wants to do something that's innovative or takes things to the next
[18:46] innovative or takes things to the next level so every go player
[18:48] level so every go player started his game for sure roughly 10
[18:50] started his game for sure roughly 10 years he dominates the
[18:52] years he dominates the professional gold world he win 18 world
[18:54] professional gold world he win 18 world championships
[19:11] definitely lizard is going to win the game 5-0
[19:13] game 5-0 yeah all game it's gonna win all game
[19:18] yeah all game it's gonna win all game we had our evaluation match last week we
[19:21] we had our evaluation match last week we won a game
[19:22] won a game and we lost a game and we lost the game
[19:24] and we lost a game and we lost the game in a way that would have made us look
[19:25] in a way that would have made us look extremely foolish if that happened
[19:27] extremely foolish if that happened publicly it means that we still have
[19:30] publicly it means that we still have work to do
[19:30] work to do and we need to take this really
[19:31] and we need to take this really seriously it's just too much risk that
[19:34] seriously it's just too much risk that actually we could we could lose overall
[19:37] actually we could we could lose overall not only that but we could lose in a way
[19:39] not only that but we could lose in a way that makes us look rather silly
[19:41] that makes us look rather silly so i guess um yeah i did you want to say
[19:44] so i guess um yeah i did you want to say anything about what you're trying
[19:46] anything about what you're trying i'm walking home
[19:52] so we're working around the clock at the moment training our algorithms further
[19:54] moment training our algorithms further trying to incrementally keep on
[19:56] trying to incrementally keep on improving right up to the moment where
[19:58] improving right up to the moment where we have the match
[19:59] we have the match we've collected together people with
[20:01] we've collected together people with different skill sets
[20:02] different skill sets onto the same team so we have
[20:04] onto the same team so we have researchers engineers
[20:05] researchers engineers evaluation guys that you mean what
[20:08] evaluation guys that you mean what happens if they create something new
[20:09] happens if they create something new which wasn't used
[20:11] which wasn't used then we don't see it okay the first
[20:14] then we don't see it okay the first thing to know is things can help the
[20:16] thing to know is things can help the perfect solution
[20:22] is the leaf programmer and built the original search engine
[20:24] original search engine so address responsibility is quite a big
[20:26] so address responsibility is quite a big one he will be the one
[20:27] one he will be the one sitting opposite lisa dole and actually
[20:29] sitting opposite lisa dole and actually playing the moves
[20:31] playing the moves alphago makes i'm feeling excited yeah
[20:36] alphago makes i'm feeling excited yeah at least i know he's a great player and
[20:39] at least i know he's a great player and i feel honored to play with him
[20:42] i feel honored to play with him many of my friends they are very excited
[20:44] many of my friends they are very excited about match they
[20:45] about match they keep telling me that the whole world is
[20:47] keep telling me that the whole world is watching
[20:48] watching just prepare alphago hello
[20:52] just prepare alphago hello hello dennis can you hear us yes i can
[20:54] hello dennis can you hear us yes i can hear you guys can you hear me okay
[20:57] hear you guys can you hear me okay yes perfect i was curious to meet you
[21:00] yes perfect i was curious to meet you such an amazing developer who made
[21:03] such an amazing developer who made alphago
[21:04] alphago it's very nice to meet you thank you
[21:06] it's very nice to meet you thank you thank you likewise um
[21:08] thank you likewise um it's a real pleasure to meet you um i
[21:11] it's a real pleasure to meet you um i love
[21:11] love the game of go i'm not very good but i
[21:14] the game of go i'm not very good but i don't know if you know
[21:15] don't know if you know um but we actually have something in
[21:18] um but we actually have something in common
[21:18] common i actually also i'm trained at a game at
[21:21] i actually also i'm trained at a game at a young age
[21:22] a young age i used to be when i was young
[21:24] i used to be when i was young professional chess player
[21:31] i used to play for the england team and then i stopped playing when i was
[21:33] and then i stopped playing when i was about 14. i was um
[21:35] about 14. i was um when i was 13 i was the second highest
[21:37] when i was 13 i was the second highest rated player in the world
[21:39] rated player in the world after the match i would like to propose
[21:42] after the match i would like to propose that i give you one teaching game of go
[21:44] that i give you one teaching game of go and you give me one teaching game of
[21:46] and you give me one teaching game of chess
[21:48] chess that sounds very good i would be very
[21:50] that sounds very good i would be very honored to to do that
[21:52] honored to to do that i'm not sure if it's okay to ask you
[21:55] i'm not sure if it's okay to ask you this question
[21:56] this question but i saw the games against fanfuy and i
[21:59] but i saw the games against fanfuy and i didn't think
[22:00] didn't think it was quite at the level to play with
[22:02] it was quite at the level to play with me
[22:03] me but i heard that it's getting really
[22:05] but i heard that it's getting really stronger
[22:06] stronger can i ask how much it got stronger um
[22:13] can i ask how much it got stronger um yeah i i i can't say too much of course
[22:16] yeah i i i can't say too much of course but it's definitely got uh significantly
[22:19] but it's definitely got uh significantly stronger
[22:20] stronger well with more time i think a logical
[22:24] well with more time i think a logical approach would be to follow up on
[22:25] approach would be to follow up on ash's local search and run that in fast
[22:28] ash's local search and run that in fast alpha
[22:28] alpha and generate a whole new data set and
[22:31] and generate a whole new data set and iterate for more time
[22:32] iterate for more time but we're just out of time we realized
[22:36] but we're just out of time we realized that if we wanted to prepare for
[22:39] that if we wanted to prepare for a mammoth task like taking on lisa dole
[22:41] a mammoth task like taking on lisa dole there would be nothing better than
[22:43] there would be nothing better than talking to a professional and we
[22:45] talking to a professional and we couldn't have picked a better person
[22:47] couldn't have picked a better person than fanway
[22:54] we invite him as an advisor because during the match we found he is a man of
[22:57] during the match we found he is a man of good spirit
[22:58] good spirit welcome back thank you very much he's
[23:00] welcome back thank you very much he's very precious that we have fun
[23:01] very precious that we have fun with us
[23:07] he was crushed by avago but then he became very positive and
[23:09] became very positive and a big help to us when them has tell me
[23:13] a big help to us when them has tell me can you come back to help us make africa
[23:15] can you come back to help us make africa more stronger
[23:17] more stronger and feel this respect of course i
[23:20] and feel this respect of course i accept about this
[23:23] accept about this i play with alphago to understand
[23:27] i play with alphago to understand where is the strong point about alphago
[23:30] where is the strong point about alphago and
[23:30] and where is baby the weakness
[23:34] where is baby the weakness i play this morning afternoon
[23:38] i play this morning afternoon all the time and i found something
[23:43] all the time and i found something i found big weakness about alphago
[23:46] i found big weakness about alphago the big one the superheroes always have
[23:50] the big one the superheroes always have a hidden vulnerability
[23:51] a hidden vulnerability right and the same is true of alphago
[23:54] right and the same is true of alphago it's unbelievably superhuman in general
[23:58] it's unbelievably superhuman in general but it has some particular weaknesses in
[24:01] but it has some particular weaknesses in some situations
[24:02] some situations we can think of there being this space
[24:04] we can think of there being this space of all the things it knows about
[24:06] of all the things it knows about and it knows about most of it extremely
[24:08] and it knows about most of it extremely well but then there'll be these tricky
[24:09] well but then there'll be these tricky lumps of knowledge
[24:10] lumps of knowledge that it just understands very poorly and
[24:14] that it just understands very poorly and it's really hard for us to characterize
[24:15] it's really hard for us to characterize when it's going to enter into one of
[24:17] when it's going to enter into one of these lumps
[24:17] these lumps but if it does it can be completely
[24:19] but if it does it can be completely delusional thinking that it's alive on
[24:21] delusional thinking that it's alive on one part of the board when in fact it's
[24:22] one part of the board when in fact it's dead or vice versa
[24:24] dead or vice versa so there is a real risk that we could
[24:26] so there is a real risk that we could lose the match
[24:28] lose the match everybody in the team try to work more
[24:31] everybody in the team try to work more and more to fix the problem but
[24:34] and more to fix the problem but i think it's difficult to fix this very
[24:37] i think it's difficult to fix this very quickly we've got version 18 in the
[24:38] quickly we've got version 18 in the pipeline haven't we
[24:40] pipeline haven't we well we'll only go for verse 19 if it
[24:42] well we'll only go for verse 19 if it makes it significant if it doesn't
[24:44] makes it significant if it doesn't which were having to rerun all the tests
[24:46] which were having to rerun all the tests because they
[24:47] because they um they went wrong unfortunately so
[24:49] um they went wrong unfortunately so we're sort of basically a day behind in
[24:51] we're sort of basically a day behind in our evaluation because of that
[24:52] our evaluation because of that all the confidence well maybe we just
[24:53] all the confidence well maybe we just need to be realistic that we've tried a
[24:55] need to be realistic that we've tried a bunch of things
[24:56] bunch of things and we've come to the point where we
[24:58] and we've come to the point where we said we would actually start freezing
[24:59] said we would actually start freezing the code and
[25:01] the code and saying look it's
[25:04] saying look it's not and actually not panning out these
[25:07] not and actually not panning out these delusions are still
[25:08] delusions are still a realistic possibility for the match we
[25:10] a realistic possibility for the match we have some weaknesses
[25:12] have some weaknesses that we i don't think we're going to fix
[25:14] that we i don't think we're going to fix fully um
[25:15] fully um before the match so um that's causing us
[25:19] before the match so um that's causing us a little bit of anxiety
[25:21] a little bit of anxiety [Music]
[25:29] fleecy doll is getting ready to rumble on wednesday
[25:29] on wednesday live across the internet this
[25:31] live across the internet this professional south korean go player will
[25:33] professional south korean go player will take on
[25:34] take on artificial intelligence program alphago
[25:43] artificial intelligence program alphago tomorrow julian and george they pack up
[25:45] tomorrow julian and george they pack up version 18
[25:46] version 18 they stick it on a laptop and they fly
[25:48] they stick it on a laptop and they fly out to seoul
[25:56] out to seoul [Music]
[26:14] when i get to the hotel i'm going to catch up with the team what we should do
[26:16] catch up with the team what we should do in the time remaining is
[26:17] in the time remaining is list things that could go wrong in the
[26:19] list things that could go wrong in the solidity of the system
[26:25] solidity of the system [Applause]
[26:32] in a difficult situation he likes to smoke yes there is terrace
[26:34] smoke yes there is terrace and we'll have security so he'll be able
[26:36] and we'll have security so he'll be able to go up then be by himself
[26:51] when i got here i didn't expect the attention on the match it was
[26:52] attention on the match it was literally front page news about eight
[26:55] literally front page news about eight million
[26:56] million koreans play the game of go and even
[26:57] koreans play the game of go and even those who don't you know recognize lee
[26:59] those who don't you know recognize lee sedol
[27:00] sedol he's a national figure and so there's
[27:03] he's a national figure and so there's that
[27:03] that right there's some national pride
[27:04] right there's some national pride involved but it's more than that
[27:07] involved but it's more than that just the very thought of a machine
[27:08] just the very thought of a machine playing a human
[27:10] playing a human at something like this i think is
[27:11] at something like this i think is inherently intriguing to people
[27:50] all the world gave the pressure to listen
[27:51] listen before this he played the tournament for
[27:53] before this he played the tournament for country
[27:54] country for himself but this time he played for
[27:57] for himself but this time he played for the human
[28:09] i just really hope we win this first game if you lose the first game you
[28:10] game if you lose the first game you literally have to win three out of four
[28:12] literally have to win three out of four yeah which is hard work where are you
[28:15] yeah which is hard work where are you going to be
[28:16] going to be i'll start in the match room and then
[28:17] i'll start in the match room and then i'm going to come in here make sure
[28:18] i'm going to come in here make sure everything's here yeah
[28:19] everything's here yeah you should probably be in here
[28:45] hello and welcome to the deepmind challenge
[28:46] challenge game one round one live from the four
[28:49] game one round one live from the four seasons here
[28:50] seasons here in seoul korea i am chris garlick of the
[28:53] in seoul korea i am chris garlick of the american go e-journal i'm here with
[28:55] american go e-journal i'm here with michael
[28:56] michael redmond nine don professional welcome
[28:58] redmond nine don professional welcome michael i want to give a
[28:59] michael i want to give a shout out to all of the folks watching
[29:02] shout out to all of the folks watching around the world
[29:03] around the world well the excitement's pretty palpable
[29:05] well the excitement's pretty palpable here in the hotel i've never seen a
[29:07] here in the hotel i've never seen a crush of
[29:07] crush of interest to reporters from around the
[29:11] interest to reporters from around the world
[29:19] all right folks you're here you're gonna see history made stay with us
[29:20] see history made stay with us five minutes guys five minutes
[29:34] i just thought we should take a moment together
[29:35] together and just think about what's about to
[29:38] and just think about what's about to happen
[29:39] happen extremely excited to be there extremely
[29:41] extremely excited to be there extremely proud of every one of you and what we've
[29:43] proud of every one of you and what we've done
[29:44] done and we're no loss i think it's just
[29:45] and we're no loss i think it's just amazing that we're here
[30:07] [Music] okay
[30:08] okay [Music]
[30:21] [Music] so
[30:49] [Music] boys
[30:51] boys [Music]
[31:04] foreign [Music]
[32:01] we had worked so hard to make sure that this would go
[32:02] this would go technically smoothly we tested it and
[32:05] technically smoothly we tested it and tested it and tested it
[32:07] tested it and tested it and still there comes that moment when
[32:10] and still there comes that moment when you're
[32:11] you're live all the tv cameras are broadcasting
[32:14] live all the tv cameras are broadcasting everything and now it has to do the
[32:17] everything and now it has to do the thing you built it to do
[33:11] i was a bit nervous it's the first time that i
[33:12] that i sit in front of a world-class co-player
[33:16] sit in front of a world-class co-player and i actually can feel the spirit and
[33:18] and i actually can feel the spirit and courtesy of a great co-player
[33:21] courtesy of a great co-player like lisa because i think it is the
[33:23] like lisa because i think it is the first time he
[33:24] first time he faced a strange opponent i think he's
[33:27] faced a strange opponent i think he's not a human
[33:28] not a human has no emotion it's cold but he
[33:32] has no emotion it's cold but he stays very calm and i can feel his
[33:36] stays very calm and i can feel his mental strength
[33:51] [Music] oh
[33:53] oh [Music]
[34:07] here is meow on kim uh nine don pro from the korean baduk association
[34:09] the korean baduk association we are here live at the four seasons
[34:10] we are here live at the four seasons hotel on the 21st floor
[34:12] hotel on the 21st floor what are we looking at how's the game
[34:14] what are we looking at how's the game going
[34:16] going oh it's fighting no from the beginning
[34:18] oh it's fighting no from the beginning from the very beginning
[34:19] from the very beginning yeah yeah he i forgot place very well
[34:22] yeah yeah he i forgot place very well yeah
[34:22] yeah just he's like a top professional just
[34:24] just he's like a top professional just like a top professional
[34:26] like a top professional yeah it's very aggressive
[34:38] now the fight is getting really complicated um this is actually the
[34:41] complicated um this is actually the first time i've seen alpha go
[34:43] first time i've seen alpha go playing a game that has this difficult
[34:47] playing a game that has this difficult fight
[35:11] i just saw him looking at his opponent's face and that's that's just a kind of a
[35:13] face and that's that's just a kind of a habit
[35:14] habit just an instinct as a player to look at
[35:16] just an instinct as a player to look at the person across it yeah
[35:17] the person across it yeah it's it's sort of like um something that
[35:20] it's it's sort of like um something that lisa though would do when he was
[35:21] lisa though would do when he was wondering
[35:22] wondering how his opponent was feeling right it's
[35:24] how his opponent was feeling right it's just a just a habit
[35:25] just a just a habit so it's not as if ajah huang is going to
[35:28] so it's not as if ajah huang is going to do any giveaway because ajahn isn't
[35:30] do any giveaway because ajahn isn't alphago right
[36:30] with human when you play you can have exchange
[36:31] exchange by feeling i look in you i know okay
[36:34] by feeling i look in you i know okay maybe you won't talk with me maybe you
[36:37] maybe you won't talk with me maybe you fear about me
[36:38] fear about me i can feel many many things but alphago
[36:42] i can feel many many things but alphago you can feel nothing so
[36:45] you can feel nothing so when you confirm nothing when you play
[36:47] when you confirm nothing when you play you have a more than more question about
[36:49] you have a more than more question about yourself
[36:51] yourself for beginning you think okay my move is
[36:52] for beginning you think okay my move is good and really
[36:54] good and really good really good oh maybe bad
[36:57] good really good oh maybe bad oh terrible why would you hear more
[37:00] oh terrible why would you hear more and more it's hard to know where to be
[37:04] and more it's hard to know where to be all the different rooms are like
[37:05] all the different rooms are like exciting a different way isn't it
[37:08] exciting a different way isn't it to be here at the heart of the operation
[37:10] to be here at the heart of the operation yeah i feel
[37:11] yeah i feel safe here white's thinking of doing this
[37:15] safe here white's thinking of doing this huge invasion here from its thick wall
[37:18] huge invasion here from its thick wall yes go for it it's done it's
[37:23] yes go for it it's done it's [Music]
[37:29] that is not a confident face he's pretty uh horrified by that
[37:31] uh horrified by that i can't believe what i see right now oh
[37:34] i can't believe what i see right now oh really
[37:35] really she thought you know she's a little bit
[37:37] she thought you know she's a little bit behind
[37:38] behind and she made a very aggressive move here
[37:41] and she made a very aggressive move here okay make it very complicated
[37:42] okay make it very complicated okay i think he said oh if he doesn't
[37:45] okay i think he said oh if he doesn't know respond quickly
[37:47] know respond quickly then he can collapse
[37:50] then he can collapse i mean this one is kind of unthinkable
[37:53] i mean this one is kind of unthinkable as a human group they are
[37:55] as a human group they are what if she knows everything about
[37:57] what if she knows everything about what's going to happen next
[37:59] what's going to happen next where are we here's our search terms
[38:02] where are we here's our search terms 50 or 60 that's the maximum number of
[38:06] 50 or 60 that's the maximum number of moves ahead that alphago is looking
[38:08] moves ahead that alphago is looking from the current game position it's
[38:11] from the current game position it's typically over 50 it's often over 60.
[38:15] typically over 50 it's often over 60. in the games we see often around 150
[38:19] in the games we see often around 150 alpha go goes for the kill we're 100
[38:21] alpha go goes for the kill we're 100 move 115 now so
[38:23] move 115 now so we're getting to that point
[38:31] we are all astonished just in the middle of the game
[38:32] of the game because although go is uh seems to be
[38:35] because although go is uh seems to be doing a much better job than we all
[38:37] doing a much better job than we all thought it would
[38:39] thought it would i thought they said all would be leading
[38:41] i thought they said all would be leading the game
[38:42] the game comfortably but it turned out that he's
[38:45] comfortably but it turned out that he's struggling at the moment
[38:46] struggling at the moment but i think eventually he will prevail i
[38:49] but i think eventually he will prevail i hope
[38:50] hope i was uh you know more five to zero
[38:54] i was uh you know more five to zero but now i'm not sure about that
[39:11] there might be the first mistake kind of clear mistake that
[39:57] why one or not at this time you check this if it's like this white has won by
[39:59] this if it's like this white has won by a lot yeah
[40:16] wow it's so shocking i i expected alphago to win only one
[40:36] i can feel his pain like he was he couldn't believe you know he couldn't
[40:37] he couldn't believe you know he couldn't accept it it takes time
[40:40] accept it it takes time for him to accept the outcome
[40:47] maybe alphago is very strong now but he don't won't believe
[40:48] don't won't believe heavy rules it's just the goal portion
[40:51] heavy rules it's just the goal portion player we
[40:52] player we took we can't believe this because form
[40:55] took we can't believe this because form for us
[40:56] for us is something very very far it's not
[40:59] is something very very far it's not can be coming now it's impossible
[41:02] can be coming now it's impossible but in reality it's not
[41:07] but in reality it's not i think he resigned in a very polite way
[41:11] i think he resigned in a very polite way these are the he's black but he blew
[41:13] these are the he's black but he blew white stone
[41:49] i feel really good i feel like um i i really believe in alphago of course
[41:51] i really believe in alphago of course it's natural that humans want humans to
[41:53] it's natural that humans want humans to win i mean i think that's a natural
[41:54] win i mean i think that's a natural response
[41:55] response but alphago is human created and i think
[41:58] but alphago is human created and i think that's the ultimate sign of human
[42:00] that's the ultimate sign of human ingenuity and cleverness everything that
[42:02] ingenuity and cleverness everything that alphago does
[42:03] alphago does it does because a human has either
[42:06] it does because a human has either created the data that it learns from
[42:07] created the data that it learns from created the learning algorithm that
[42:09] created the learning algorithm that learns from that data creates the search
[42:11] learns from that data creates the search algorithm
[42:12] algorithm all of these things have come from from
[42:14] all of these things have come from from humans
[42:15] humans so really this is a human endeavor
[42:35] in the battle between man versus machine a computer just came out the victory
[42:36] a computer just came out the victory deep
[42:37] deep mind put its computer program to the
[42:39] mind put its computer program to the test against one of the brightest
[42:40] test against one of the brightest minds in the world and one alphago beat
[42:43] minds in the world and one alphago beat a professional player
[42:44] a professional player who has 18 go world championships under
[42:47] who has 18 go world championships under his belt
[42:48] his belt the victor is considered a breakthrough
[42:50] the victor is considered a breakthrough in artificial intelligence
[43:48] in research we normally work to produce an
[43:48] an academic paper it gets published and
[43:50] academic paper it gets published and maybe we get to talk about it in a
[43:51] maybe we get to talk about it in a conference if we're lucky
[43:55] conference if we're lucky this is not normal for research in fact
[43:57] this is not normal for research in fact i've never experienced
[43:58] i've never experienced any media attention remotely close to
[44:01] any media attention remotely close to this
[44:01] this so it's a special moment for us all and
[44:03] so it's a special moment for us all and we're just enjoying it while it lasts
[44:09] we're just enjoying it while it lasts [Music]
[44:11] [Music] in a matchup between man and machine who
[44:13] in a matchup between man and machine who wins well so far
[44:15] wins well so far it's the machine in seoul south korea
[44:17] it's the machine in seoul south korea the artificially intelligent computer
[44:20] the artificially intelligent computer defeated the global champion in the
[44:22] defeated the global champion in the ancient chinese board game
[44:24] ancient chinese board game go lee sidol lost the first matchup
[44:27] go lee sidol lost the first matchup but he's got four more chances
[44:46] it is a bit strange being the front cover and everything as a computer
[44:48] cover and everything as a computer scientist
[44:49] scientist normally you sit there in your corner
[44:50] normally you sit there in your corner you could nobody really knows about it
[44:53] you could nobody really knows about it perhaps heard the joke how can you tell
[44:55] perhaps heard the joke how can you tell that a computer scientist
[44:56] that a computer scientist is an extrovert and not an introvert if
[45:00] is an extrovert and not an introvert if he's an extrovert he looks at your shoes
[45:01] he's an extrovert he looks at your shoes when he's talking to you
[45:03] when he's talking to you instead of at his own
[45:06] instead of at his own like if you look at archer he avoids all
[45:08] like if you look at archer he avoids all the cameras like
[45:10] the cameras like crazy he's like game is finished and
[45:12] crazy he's like game is finished and he's like out the back and
[45:14] he's like out the back and back in his room and i think a lot of
[45:18] back in his room and i think a lot of computer scientists would be like that
[45:20] computer scientists would be like that we were more about doing our work than
[45:22] we were more about doing our work than standing in the spotlight
[45:25] standing in the spotlight hello and welcome to game two round two
[45:29] hello and welcome to game two round two in the google deepmind challenge
[45:32] in the google deepmind challenge throwdown between
[45:33] throwdown between man and machine game one the machine
[45:37] man and machine game one the machine takes down man huge shock headlines
[45:41] takes down man huge shock headlines around the world
[45:42] around the world the reactions on the ground here from
[45:44] the reactions on the ground here from the folks in korea were just stunned
[45:46] the folks in korea were just stunned they estimate 60 million people watch
[45:48] they estimate 60 million people watch the game in china
[45:49] the game in china alone probably bringing up to maybe 80
[45:52] alone probably bringing up to maybe 80 million people watch this game worldwide
[45:54] million people watch this game worldwide it was just incredible and if anything
[45:56] it was just incredible and if anything today is probably
[45:57] today is probably even more of a bad house so he said i'll
[46:00] even more of a bad house so he said i'll knows today he knows that this is just a
[46:02] knows today he knows that this is just a really important game right he's got to
[46:28] i think probably looking for a little uh a little payback
[46:30] a little payback my impression is that maybe he
[46:31] my impression is that maybe he underestimated alphago
[46:33] underestimated alphago and he's going to change his tactics
[46:58] oh well if i say anything about alphago that it's not
[47:00] say anything about alphago that it's not normal it's maybe the way it handles
[47:02] normal it's maybe the way it handles a game when it thinks it is ahead yes
[47:05] a game when it thinks it is ahead yes we're actually going to have a visit
[47:06] we're actually going to have a visit from one of the team
[47:08] from one of the team and we'll talk about exactly that point
[47:09] and we'll talk about exactly that point from the inside thanks so much for
[47:11] from the inside thanks so much for coming by thor i really appreciate it
[47:13] coming by thor i really appreciate it can you sort of share a bit of what is
[47:16] can you sort of share a bit of what is going on
[47:16] going on in alphago so alphago has these three
[47:20] in alphago so alphago has these three main components there's the policy
[47:22] main components there's the policy network
[47:23] network which was trained on high-level games
[47:26] which was trained on high-level games to imitate those players and then we
[47:29] to imitate those players and then we have a second component we call this the
[47:31] have a second component we call this the value net
[47:32] value net and it can evaluate the board position
[47:35] and it can evaluate the board position and say what is the probability of
[47:37] and say what is the probability of winning
[47:37] winning in this particular position and then
[47:40] in this particular position and then third component
[47:41] third component is the tree search where it would look
[47:44] is the tree search where it would look through
[47:44] through different variations of the game and try
[47:47] different variations of the game and try to figure out what will happen in the
[47:52] future so if we now take a position like this
[47:55] so if we now take a position like this first the policy network would
[47:57] first the policy network would scan the position and come up
[48:00] scan the position and come up with what would be the interesting spot
[48:02] with what would be the interesting spot to play
[48:04] to play and it builds up a tree of variations
[48:07] and it builds up a tree of variations and it then employs this value net that
[48:10] and it then employs this value net that tells it
[48:10] tells it how promising is the outcome of this
[48:12] how promising is the outcome of this particular
[48:13] particular variation
[48:20] so alphago tries to maximize its probability of winning
[48:22] probability of winning but it doesn't care at all about the
[48:24] but it doesn't care at all about the margin by which it wins
[48:26] margin by which it wins okay so when you see a slow looking move
[48:28] okay so when you see a slow looking move that's maybe an indication that alphago
[48:30] that's maybe an indication that alphago thinks it has a good chance to win
[48:32] thinks it has a good chance to win yeah that is a little giveaway
[48:35] yeah that is a little giveaway we're looking for a hotel
[48:49] we're looking for a hotel [Music]
[49:08] ooh looks like uh looks like lee is taking a little bit of a break
[49:31] lisa dal go to smoke and alva go just play it don't sing about the opening
[49:34] play it don't sing about the opening will be zero or not
[49:35] will be zero or not so i jab sees alphago plays the move 37
[49:40] so i jab sees alphago plays the move 37 and aja puts the stone in the board
[49:55] that's a very that's a very surprising move i thought
[49:57] move i thought i thought it was i thought it was a
[49:59] i thought it was i thought it was a mistake
[50:01] mistake when i see this move for me it's just
[50:03] when i see this move for me it's just the big shock
[50:04] the big shock what normally humans will never play
[50:07] what normally humans will never play this one
[50:08] this one because it's bad it's just bad we don't
[50:11] because it's bad it's just bad we don't know why it's bad
[50:13] know why it's bad but it's a little bit high yeah it's
[50:16] but it's a little bit high yeah it's fifth line normally you don't make a
[50:18] fifth line normally you don't make a shoulder here on the fifth line
[50:20] shoulder here on the fifth line um so coming on top of a fourth line
[50:22] um so coming on top of a fourth line zone is really unusual
[50:24] zone is really unusual yeah that's an exciting move i i think
[50:26] yeah that's an exciting move i i think we've seen an original move here
[50:29] we've seen an original move here that's the kind of move uh that that you
[50:32] that's the kind of move uh that that you play go for
[50:35] play go for hey
[50:40] interesting stuff this fifth line shoulder head is that
[50:42] shoulder head is that yeah i wasn't expecting that um i don't
[50:45] yeah i wasn't expecting that um i don't really know if it's a good or bad move
[50:46] really know if it's a good or bad move at this point
[50:47] at this point the professional commentators almost
[50:50] the professional commentators almost unanimously said that
[50:51] unanimously said that not a single human player would have
[50:53] not a single human player would have chosen move 37.
[50:55] chosen move 37. so i actually had a poke around in
[50:56] so i actually had a poke around in alphago to see what alphago
[50:58] alphago to see what alphago thought and alphago actually agreed with
[51:01] thought and alphago actually agreed with that assessment
[51:02] that assessment alphago said there was a 1 in 10 000
[51:05] alphago said there was a 1 in 10 000 probability
[51:06] probability that move 37 would have been played by a
[51:09] that move 37 would have been played by a human player so it knew that this was an
[51:10] human player so it knew that this was an extremely unlikely move
[51:12] extremely unlikely move it went beyond its human guide
[51:15] it went beyond its human guide and it came up with something new and
[51:17] and it came up with something new and creative and different
[51:19] creative and different i am very much watching the game
[51:22] i am very much watching the game through these commentators that's the
[51:25] through these commentators that's the way it works so when they're confused
[51:26] way it works so when they're confused i'm certainly confused
[51:28] i'm certainly confused at the same time i'm latching on to the
[51:29] at the same time i'm latching on to the fact that they are confused right that
[51:31] fact that they are confused right that is that is an interesting moment
[51:33] is that is an interesting moment when everyone else is confused who's not
[51:35] when everyone else is confused who's not confused right besides the machine
[51:50] i want to see lisa dal when he sees this he's back lee is back
[52:42] normally have seen about one two minutes no more
[52:43] no more but this time it think more than two
[52:45] but this time it think more than two minutes
[52:57] the more i see this move i feel something changed
[53:00] something changed maybe for human we think it's bad but
[53:03] maybe for human we think it's bad but for alphago
[53:04] for alphago why not go is like geopolitics like
[53:08] why not go is like geopolitics like something small that happens here can
[53:10] something small that happens here can have a ripple effect
[53:11] have a ripple effect hours down down the road in a different
[53:14] hours down down the road in a different part of the board
[53:15] part of the board the game kind of turned on its axis at
[53:17] the game kind of turned on its axis at that moment
[53:19] that moment this move is very special because with
[53:21] this move is very special because with this move
[53:22] this move all the stone played before is work
[53:25] all the stone played before is work together
[53:26] together it's connect it look like a network link
[53:29] it's connect it look like a network link everywhere
[53:30] everywhere it's very special very special
[53:53] this is a tough game for lisa alphago is just not
[53:54] just not letting lisa do what he wants right
[53:57] letting lisa do what he wants right black hat's almost
[53:58] black hat's almost 50 points that's a lot that's not a good
[54:01] 50 points that's a lot that's not a good sign
[54:02] sign oh he said all just slapped himself on
[54:05] oh he said all just slapped himself on the side of the head
[54:07] the side of the head oh wow i think black's ahead at this
[54:10] oh wow i think black's ahead at this point
[54:12] point it's looking good and we're not steady
[54:15] it's looking good and we're not steady steady path now
[54:23] i just see lucidal he lose so much normally we resent people long time ago
[54:25] normally we resent people long time ago but he won't try
[54:26] but he won't try he played he played he played he just
[54:29] he played he played he played he just don't want to reason
[54:39] don't want to reason [Music]
[54:49] because then white oh he resigned it looks like
[54:50] looks like lisa dole has just resigned
[55:23] there was this heavy sadness over that whole floor
[55:24] whole floor you could feel it during the game i felt
[55:26] you could feel it during the game i felt during the game and
[55:27] during the game and i'm leaving the commentary room to go to
[55:30] i'm leaving the commentary room to go to the press conference and i was stopped
[55:32] the press conference and i was stopped by someone
[55:32] by someone another technology reporter now at first
[55:34] another technology reporter now at first all he wanted to talk about was the
[55:36] all he wanted to talk about was the technology and how great this was but
[55:38] technology and how great this was but then even he kind of slipped into this
[55:40] then even he kind of slipped into this moment of melancholy where he was upset
[55:43] moment of melancholy where he was upset as well
[55:54] i am quite speechless i admit that it was a very clear loss
[55:55] was a very clear loss on my part from the very beginning of
[55:58] on my part from the very beginning of the game that was not a moment in time
[56:00] the game that was not a moment in time that i felt that i was leading
[56:01] that i felt that i was leading the game you feel elated and you feel
[56:05] the game you feel elated and you feel a little bit scared there is something i
[56:08] a little bit scared there is something i i think frightening to people about
[56:10] i think frightening to people about a machine that learns on its own for us
[56:13] a machine that learns on its own for us alphago is obviously just
[56:14] alphago is obviously just some computer program but looking at the
[56:17] some computer program but looking at the commentary on the internet
[56:19] commentary on the internet i already saw the commentators call
[56:21] i already saw the commentators call alphago like he
[56:22] alphago like he and she during the games completely
[56:25] and she during the games completely unconsciously
[56:27] unconsciously which alphago is really a very very
[56:29] which alphago is really a very very simple program it's not
[56:30] simple program it's not anywhere close to full ai and we already
[56:33] anywhere close to full ai and we already see that happening so i found that very
[56:34] see that happening so i found that very interesting
[56:42] the tendency to anthropomorphize ai systems is
[56:43] systems is one of the big obstacles in the way of
[56:45] one of the big obstacles in the way of actually trying to understand how
[56:47] actually trying to understand how ai might impact the world in the future
[56:50] ai might impact the world in the future for example the conversation is about
[56:51] for example the conversation is about what what could go wrong like what the
[56:52] what what could go wrong like what the risks are then in invariably you see
[56:54] risks are then in invariably you see this terminator picture
[56:56] this terminator picture every single time there are these red
[56:57] every single time there are these red glowing eyes right we're really closer
[56:59] glowing eyes right we're really closer to a smart washing machine
[57:01] to a smart washing machine than terminator if you look at today's
[57:04] than terminator if you look at today's ai we are really very nascent
[57:08] ai we are really very nascent i'm extremely excited and passionate
[57:12] i'm extremely excited and passionate about ai's potential
[57:14] about ai's potential but ai is still very limited in its
[57:17] but ai is still very limited in its power
[57:18] power i think that people are right to think
[57:20] i think that people are right to think that there is a danger that as
[57:22] that there is a danger that as we continue to improve these systems
[57:24] we continue to improve these systems that we might miss that that threshold
[57:27] that we might miss that that threshold where where we do cross
[57:28] where where we do cross over into danger but the good news is
[57:31] over into danger but the good news is there are already people thinking about
[57:32] there are already people thinking about those dangers
[57:33] those dangers you know there's a lot of talk now we're
[57:34] you know there's a lot of talk now we're leading the discussions on this that
[57:36] leading the discussions on this that maybe there should be a kind of
[57:37] maybe there should be a kind of cross-industry best practices working
[57:39] cross-industry best practices working group or something
[57:41] group or something where the leaders of the research teams
[57:43] where the leaders of the research teams in those organizations the you know the
[57:45] in those organizations the you know the big ones that are working on ai ibm
[57:47] big ones that are working on ai ibm microsoft so on
[57:48] microsoft so on uh come together and make sure that ai's
[57:50] uh come together and make sure that ai's used ethically and responsibly
[57:53] used ethically and responsibly i think what is important is that there
[57:55] i think what is important is that there is this community of people
[57:57] is this community of people who are leading uh the cutting edge of
[58:00] who are leading uh the cutting edge of ai
[58:01] ai who are interact with academics and
[58:03] who are interact with academics and already are thinking about the long term
[58:06] already are thinking about the long term and how we can ensure that innovation is
[58:09] and how we can ensure that innovation is responsible
[58:10] responsible as the power of these machines gets even
[58:12] as the power of these machines gets even greater
[58:18] this is it folks day three game three lease it all go master
[58:22] game three lease it all go master back to the wall he's down 2-0
[58:25] back to the wall he's down 2-0 he's got to win today to keep hope alive
[58:49] to console him and to review africa console him do you think he's
[58:51] console him do you think he's he's upset or i mean i mean is upset
[58:54] he's upset or i mean i mean is upset yeah
[58:55] yeah in the beginning there was a fearful
[58:57] in the beginning there was a fearful fight and then avago
[58:58] fight and then avago played very well so he secured a very
[59:01] played very well so he secured a very early lead
[59:02] early lead from mu 50 the win rate was very high
[59:05] from mu 50 the win rate was very high already
[59:06] already and it was climbing toward 100 percent
[59:08] and it was climbing toward 100 percent oh that's a great move what's your
[59:10] oh that's a great move what's your what's your probability um
[59:36] [Music] he tried to fight directly in the game
[59:38] he tried to fight directly in the game but it's not half stale
[59:40] but it's not half stale when we change our style to play with
[59:43] when we change our style to play with opponent
[59:44] opponent normally it's very very bad
[59:52] so is the game more easy for alphago it's looking good for us locally
[59:56] alphago it's looking good for us locally i think the black group is huge and it's
[59:58] i think the black group is huge and it's got nowhere to go
[59:59] got nowhere to go and it's going to be running around
[01:00:13] i don't know how to describe the situation
[01:00:13] situation if i were black i will resign we should
[01:00:17] if i were black i will resign we should admit
[01:00:17] admit that we are facing the strongest
[01:00:20] that we are facing the strongest existence ever
[01:00:21] existence ever ever in the gold history there's no
[01:00:23] ever in the gold history there's no point in playing out the end game
[01:00:25] point in playing out the end game and you're gonna you know you're gonna
[01:00:26] and you're gonna you know you're gonna lose right so
[01:00:33] even if black can live there oh he resigned
[01:00:33] resigned okay wow wow
[01:00:41] you saw history made here tonight alphago is one again three straight wins
[01:00:43] alphago is one again three straight wins three straight wins has won the match
[01:00:46] three straight wins has won the match when lisa is on the game he he looked
[01:00:49] when lisa is on the game he he looked like unhappy
[01:00:50] like unhappy it's not just he lose the torment it's
[01:00:52] it's not just he lose the torment it's just especially about this game
[01:00:55] just especially about this game because he don't play his game
[01:00:58] because he don't play his game i'm very hard about this very
[01:01:03] i'm very hard about this very but i can do nothing
[01:01:10] you know certainly i feel a bit ambivalent about it given i'm a games
[01:01:12] ambivalent about it given i'm a games player and you know go is the pinnacle
[01:01:14] player and you know go is the pinnacle of board games
[01:01:15] of board games but i really like this statement one of
[01:01:16] but i really like this statement one of the top chinese professionals said of
[01:01:19] the top chinese professionals said of you know if alpha go wins maybe we'll
[01:01:22] you know if alpha go wins maybe we'll really start to get to see what this
[01:01:24] really start to get to see what this game's about
[01:01:26] game's about i couldn't celebrate it was fantastic
[01:01:29] i couldn't celebrate it was fantastic that we'd won
[01:01:31] that we'd won but there was such a big part of me that
[01:01:33] but there was such a big part of me that saw this man
[01:01:35] saw this man trying so hard and being
[01:01:38] trying so hard and being so disappointed
[01:01:59] i see internet many many people are talking about the listener
[01:02:01] talking about the listener maybe i don't play the best you know we
[01:02:04] maybe i don't play the best you know we are go player
[01:02:05] are go player okay sometime in china in korean japan
[01:02:07] okay sometime in china in korean japan we see go
[01:02:08] we see go like ah we are artists you know we play
[01:02:12] like ah we are artists you know we play ours best right for go right so please
[01:02:15] ours best right for go right so please gentle with lisa doll
[01:02:17] gentle with lisa doll he's very very good players great
[01:02:20] he's very very good players great players
[01:02:21] players i am in the room i see this door he
[01:02:22] i am in the room i see this door he won't win
[01:02:24] won't win i try everything we just we can't
[01:03:59] um [Music]
[01:04:08] you could see that he was more relaxed after he had lost three games in a row
[01:04:11] after he had lost three games in a row that said the stakes were still high you
[01:04:13] that said the stakes were still high you know in the end
[01:04:14] know in the end it isn't about pride they don't feel
[01:04:18] it isn't about pride they don't feel like confidence but they feel light
[01:04:22] like confidence but they feel light can lee saddam find alphago's weakness
[01:04:26] can lee saddam find alphago's weakness is there in fact a weakness i was
[01:04:28] is there in fact a weakness i was thinking i just pulled a plug
[01:04:30] thinking i just pulled a plug you just pull the plot out everyone's
[01:04:32] you just pull the plot out everyone's still cheering for this at all
[01:04:34] still cheering for this at all and then yeah it's not going very well
[01:04:38] and then yeah it's not going very well now concerned
[01:04:51] it feels pretty good for for black at this point
[01:04:51] this point it feels pretty good for black guys i
[01:04:54] it feels pretty good for black guys i want
[01:04:55] want eliza play his game because for the
[01:04:58] eliza play his game because for the moment
[01:04:58] moment he tried many many things to play with
[01:05:00] he tried many many things to play with alphago to understand alphago but he
[01:05:03] alphago to understand alphago but he never tried to play
[01:05:04] never tried to play himself i thought man maintained alphago
[01:05:08] himself i thought man maintained alphago look like the real mirror when you play
[01:05:11] look like the real mirror when you play with alphago
[01:05:12] with alphago you feel very strange look like
[01:05:16] you feel very strange look like you and all the turn naked the first
[01:05:19] you and all the turn naked the first time you see this
[01:05:20] time you see this you want to see because oh it's me real
[01:05:23] you want to see because oh it's me real me
[01:05:24] me and more than more you need accept oh
[01:05:28] and more than more you need accept oh it's a real me so how now how i can do
[01:05:36] it's a real me so how now how i can do it's developing into a very very
[01:05:38] it's developing into a very very dangerous fight
[01:05:39] dangerous fight this is really lisado's type of game he
[01:05:42] this is really lisado's type of game he likes this kind of fight
[01:05:44] likes this kind of fight white has to find something inside
[01:05:45] white has to find something inside black's territory i think he is already
[01:05:47] black's territory i think he is already planning on trying something
[01:05:49] planning on trying something so you believe in ethereum's ability to
[01:05:52] so you believe in ethereum's ability to live in the very small
[01:05:54] live in the very small area right yes yes there is a little
[01:05:56] area right yes yes there is a little potential there
[01:05:58] potential there and i think maybe he's going to try to
[01:05:59] and i think maybe he's going to try to do something he said all magic
[01:06:02] do something he said all magic oh what would be the magic move lisa
[01:06:06] oh what would be the magic move lisa he's running short on time but he's
[01:06:07] he's running short on time but he's going to have to use up all this time
[01:06:09] going to have to use up all this time there he just burned like seven or eight
[01:06:11] there he just burned like seven or eight minutes just on this move already
[01:06:14] minutes just on this move already i don't see anything we can't
[01:06:17] i don't see anything we can't find anything
[01:06:24] what is lisa doll up to here yeah he's really concentrating
[01:06:25] really concentrating he really is look at look at that lisa
[01:06:28] he really is look at look at that lisa is very patient
[01:06:30] is very patient he wait he wait away this moment
[01:06:33] he wait he wait away this moment i feel something look like the wolf
[01:06:36] i feel something look like the wolf wheat in the forest in the winter
[01:06:41] wheat in the forest in the winter he cold i feel very very cold but i need
[01:06:44] he cold i feel very very cold but i need patience but the moment coming
[01:06:47] patience but the moment coming they go out to attack
[01:06:51] they go out to attack this is the hinge of the game oh
[01:06:54] this is the hinge of the game oh look at that move that's an exciting
[01:06:56] look at that move that's an exciting move no basically oh he found the wedge
[01:06:59] move no basically oh he found the wedge whoa it's going to change the equation
[01:07:03] whoa it's going to change the equation because now black cannot escape
[01:07:05] because now black cannot escape that would be so cool if that works
[01:07:16] alphago has just played something maybe unusual
[01:07:18] unusual you know i'm not actually sure what
[01:07:19] you know i'm not actually sure what alphago is trying to do here
[01:07:21] alphago is trying to do here what's that about uh you don't really
[01:07:23] what's that about uh you don't really understand it well well
[01:07:25] understand it well well that was a sharp drop in
[01:07:37] this could be that it actually can't find a way through
[01:07:39] find a way through i think this is far enough ahead to see
[01:07:42] i think this is far enough ahead to see that it doesn't work and now maybe it's
[01:07:43] that it doesn't work and now maybe it's on tilt i don't know
[01:07:52] it looks like it's like a viewer yeah it's made a mistake did anything
[01:07:55] yeah it's made a mistake did anything strange happen in there
[01:07:56] strange happen in there no it'll look normal well there we go
[01:07:58] no it'll look normal well there we go well we can definitely say well there's
[01:08:00] well we can definitely say well there's a weakness
[01:08:01] a weakness would we definitely say the mistake
[01:08:04] would we definitely say the mistake i felt this mixture of this sinking
[01:08:08] i felt this mixture of this sinking feeling in my stomach where i was
[01:08:09] feeling in my stomach where i was wondering
[01:08:09] wondering if alphago was becoming delusional in
[01:08:11] if alphago was becoming delusional in this situation where i could see that it
[01:08:13] this situation where i could see that it was starting to play strangely
[01:08:15] was starting to play strangely and at the same time relief of lucidal
[01:08:17] and at the same time relief of lucidal that he was actually
[01:08:18] that he was actually in with a chance now i just like trying
[01:08:22] in with a chance now i just like trying not to look horrified
[01:08:25] not to look horrified i knew after moved 17 after like
[01:08:28] i knew after moved 17 after like 10 or 20 moves i saw avago
[01:08:32] 10 or 20 moves i saw avago strange moves and i know abago somehow
[01:08:35] strange moves and i know abago somehow became crazy but i didn't realize why
[01:08:39] became crazy but i didn't realize why when i was playing we searched for 90.
[01:08:42] when i was playing we searched for 90. at that point okay
[01:08:45] at that point okay i think there's something went wrong
[01:08:47] i think there's something went wrong that's the longest research entire game
[01:08:48] that's the longest research entire game business
[01:08:49] business yeah i think it's like his search so
[01:08:56] i do get the impression that alphago has sort of gone off on a tangent
[01:08:58] sort of gone off on a tangent what are you doing well maybe it has a
[01:09:00] what are you doing well maybe it has a master plan
[01:09:01] master plan no he doesn't even think it has does it
[01:09:04] no he doesn't even think it has does it so it knows it's made
[01:09:05] so it knows it's made because it starts it starts evaluating
[01:09:07] because it starts it starts evaluating it the other way
[01:09:08] it the other way look look at lisa's confused i was like
[01:09:10] look look at lisa's confused i was like what is he doing that's not uh i'm
[01:09:11] what is he doing that's not uh i'm scared confused that's uh what is it
[01:09:13] scared confused that's uh what is it doing
[01:09:13] doing why is it going you know you asked if it
[01:09:16] why is it going you know you asked if it was a bug
[01:09:17] was a bug i i've i've said before that if if yeah
[01:09:20] i i've i've said before that if if yeah right
[01:09:21] right if deepmind has figured out how to write
[01:09:23] if deepmind has figured out how to write code that doesn't have bugs
[01:09:25] code that doesn't have bugs that is a bigger news story than alphago
[01:09:28] that is a bigger news story than alphago oh you're kidding me
[01:09:29] oh you're kidding me hmm there's literally this next movie
[01:09:31] hmm there's literally this next movie we're gonna play
[01:09:34] we're gonna play i think i think they're gonna laugh
[01:09:36] i think i think they're gonna laugh maybe he's gonna laugh
[01:09:54] i don't really know what alphago is trying to do here
[01:09:56] trying to do here that's the understatement of the year is
[01:09:59] that's the understatement of the year is it love
[01:10:00] it love nope that's the move hong makes no
[01:10:02] nope that's the move hong makes no misclips
[01:10:04] misclips is very confused these are not human
[01:10:07] is very confused these are not human moves
[01:10:08] moves this move also um sort of inexplicable
[01:10:11] this move also um sort of inexplicable i mean those you can you can clearly
[01:10:13] i mean those you can you can clearly call those
[01:10:14] call those mistakes yes of course but it's the
[01:10:16] mistakes yes of course but it's the first time in the in the four matches
[01:10:18] first time in the in the four matches that we've seen moves like that right
[01:10:20] that we've seen moves like that right oh the value dropped even more it's
[01:10:23] oh the value dropped even more it's weird
[01:10:25] weird we're like 45
[01:10:26] we're like 45 [Music]
[01:10:28] [Music] lights and white white swinging
[01:10:32] lights and white white swinging come on at least at all this is
[01:10:34] come on at least at all this is unbelievable
[01:10:35] unbelievable all right i was so confident that this
[01:10:38] all right i was so confident that this black moya would just be consolidated by
[01:10:41] black moya would just be consolidated by black and there's nothing there and
[01:10:42] black and there's nothing there and somehow
[01:10:43] somehow he's just erased it all like it's gone
[01:10:46] he's just erased it all like it's gone so he's found his weakness
[01:10:47] so he's found his weakness that that wedge move probably surprised
[01:10:49] that that wedge move probably surprised him yeah right
[01:10:50] him yeah right yeah alphago seemed to have such a good
[01:10:52] yeah alphago seemed to have such a good game and then
[01:10:54] game and then all this whole sequence in the center uh
[01:10:56] all this whole sequence in the center uh sort of changed that
[01:10:57] sort of changed that he pulls off a miracle he manages to
[01:11:00] he pulls off a miracle he manages to make it so complicated that the
[01:11:01] make it so complicated that the artificial intelligence
[01:11:03] artificial intelligence doesn't evaluate it correctly i think it
[01:11:05] doesn't evaluate it correctly i think it looks like aj huang
[01:11:06] looks like aj huang maybe sitting there also knows that the
[01:11:08] maybe sitting there also knows that the game is over
[01:11:10] game is over i can't i can't not stop smiling
[01:11:13] i can't i can't not stop smiling did he smile no i haven't seen him smile
[01:11:15] did he smile no i haven't seen him smile yet
[01:11:16] yet i think he is still checking he's so
[01:11:19] i think he is still checking he's so serious
[01:11:20] serious he wants to be careful yeah is he
[01:11:23] he wants to be careful yeah is he concerned
[01:11:24] concerned a lot she really does his best i think
[01:11:27] a lot she really does his best i think he feels some
[01:11:28] he feels some something like you know uh
[01:11:31] something like you know uh responsibility
[01:11:33] responsibility or burden too much
[01:11:55] looks like alphago is resigned wow the most amazing game um
[01:11:58] the most amazing game um i'm almost going to tear up was it was
[01:12:01] i'm almost going to tear up was it was was game four
[01:12:03] was game four where he comes back and wins right
[01:12:08] where he comes back and wins right [Music]
[01:12:12] [Music] [Applause]
[01:12:55] i heard from so many people saying you know they were
[01:12:56] know they were running out in the street they were so
[01:12:59] running out in the street they were so happy you know they were chanting
[01:13:00] happy you know they were chanting they're
[01:13:01] they're celebrating especially after a 0-3 down
[01:13:05] celebrating especially after a 0-3 down at the time it seems to be hopeless that
[01:13:08] at the time it seems to be hopeless that the end of the world is coming
[01:13:10] the end of the world is coming but we see the light usually
[01:13:13] but we see the light usually i'm happy when i win not like my
[01:13:16] i'm happy when i win not like my colleague wins
[01:13:17] colleague wins but this time it felt like my window
[01:13:33] but this time it felt like my window [Music]
[01:13:45] [Music] [Applause]
[01:13:58] [Applause] [Music]
[01:13:59] [Music] [Applause]
[01:14:36] my question is about number 78 move chinese taco player kuli said it was a
[01:14:38] chinese taco player kuli said it was a god's play
[01:14:39] god's play what were you thinking when you made
[01:14:41] what were you thinking when you made that
[01:14:42] that [Laughter]
[01:15:00] does leave me a little bit in awe of the human brain's power in particular lee's
[01:15:02] human brain's power in particular lee's amazing ability to
[01:15:03] amazing ability to cause alphago problems and find
[01:15:05] cause alphago problems and find something seemingly out of nothing
[01:15:08] something seemingly out of nothing and so we really want to understand what
[01:15:09] and so we really want to understand what happened so it really was lee's move
[01:15:12] happened so it really was lee's move that the key is the center if the center
[01:15:14] that the key is the center if the center go on there's notions
[01:15:16] go on there's notions we were winning before but if alphago
[01:15:19] we were winning before but if alphago thought it's winning and he
[01:15:21] thought it's winning and he couldn't convert that win that means it
[01:15:22] couldn't convert that win that means it wasn't actually winning otherwise it
[01:15:24] wasn't actually winning otherwise it would have had the right responses
[01:15:25] would have had the right responses would we have played it there's a steal
[01:15:27] would we have played it there's a steal in there what probability does it give
[01:15:29] in there what probability does it give it
[01:15:34] [Music] 0.007
[01:15:36] 0.007 is that in the position that he played
[01:15:38] is that in the position that he played that's when he played nicely
[01:15:39] that's when he played nicely so we thought this was the one yeah the
[01:15:41] so we thought this was the one yeah the value is really low
[01:15:48] so the god move was literally a god move because we believe that only one in ten
[01:15:49] because we believe that only one in ten thousands
[01:15:50] thousands uh humans would have found that that's
[01:15:52] uh humans would have found that that's right it doesn't come to mind it's
[01:15:54] right it doesn't come to mind it's the top five in the top five moves that
[01:15:56] the top five in the top five moves that you would consider and i said lisa doll
[01:15:58] you would consider and i said lisa doll he said it was the only thing
[01:15:59] he said it was the only thing and he thought it was the only one yeah
[01:16:06] and he thought it was the only one yeah welcome back to the four
[01:16:08] welcome back to the four seasons world champion lisa doll went
[01:16:10] seasons world champion lisa doll went looking
[01:16:11] looking for alphago's weakness in game four
[01:16:14] for alphago's weakness in game four and he found it today last round
[01:16:18] and he found it today last round he's looking to see if he can repeat
[01:16:19] he's looking to see if he can repeat that we'll see what happens
[01:16:21] that we'll see what happens the place is a mad house downstairs
[01:16:23] the place is a mad house downstairs maybe as many if not more journalists
[01:16:25] maybe as many if not more journalists here today than there were for game one
[01:16:29] here today than there were for game one we all were really excited for the fifth
[01:16:31] we all were really excited for the fifth match to see what would happen was this
[01:16:33] match to see what would happen was this going to be a 3-2
[01:16:34] going to be a 3-2 thing or was it going to be a 4-1 thing
[01:16:36] thing or was it going to be a 4-1 thing and you know there's a different message
[01:16:38] and you know there's a different message with both of those
[01:16:40] with both of those for me i don't think alphago will make
[01:16:42] for me i don't think alphago will make the mistake again
[01:16:45] the mistake again but who knows maybe it will tell me
[01:16:48] but who knows maybe it will tell me and the licital have now have more
[01:16:51] and the licital have now have more confidence about this
[01:16:52] confidence about this look like maybe lizzita finds a key
[01:16:56] look like maybe lizzita finds a key to open the alphago
[01:17:09] you weren't crazy about the timing on this move yeah yeah
[01:17:11] this move yeah yeah and now you don't approve of this yeah
[01:17:14] and now you don't approve of this yeah i'm sort of thinking that maybe alphago
[01:17:17] i'm sort of thinking that maybe alphago hasn't recovered from game four yet yeah
[01:17:19] hasn't recovered from game four yet yeah so
[01:17:20] so white has to do something about this one
[01:17:22] white has to do something about this one here afterwards no hope he doesn't
[01:17:24] here afterwards no hope he doesn't know he needs to play though
[01:17:32] play some move is bad i feel something oh maybe this victim is
[01:17:35] i feel something oh maybe this victim is come back again
[01:17:36] come back again are we seeing another short circuit or
[01:17:39] are we seeing another short circuit or is there something
[01:17:40] is there something what's uh i think it could be a kind of
[01:17:42] what's uh i think it could be a kind of a misreading uh
[01:17:44] a misreading uh and we're you're pretty comfortable with
[01:17:46] and we're you're pretty comfortable with saying that
[01:17:48] saying that you're doing good for at least good for
[01:17:49] you're doing good for at least good for at least no
[01:17:56] there's no reason for white to be playing that move it's a bad move
[01:17:59] playing that move it's a bad move and in some cases it's going to lose a
[01:18:01] and in some cases it's going to lose a point too
[01:18:02] point too it is exactly the same 91
[01:18:05] it is exactly the same 91 certainly someone was telling me that
[01:18:07] certainly someone was telling me that maybe
[01:18:13] the whole game we thought that alphago was wrong about the board position we
[01:18:15] was wrong about the board position we were super worried that
[01:18:16] were super worried that oh i was going to play garbage is going
[01:18:18] oh i was going to play garbage is going to be like loose in a very embarrassing
[01:18:19] to be like loose in a very embarrassing way
[01:18:20] way and this continues for the whole game
[01:18:22] and this continues for the whole game but this is a big weird
[01:18:24] but this is a big weird time when did they play
[01:18:33] yeah as it turned out none of us know go well enough
[01:18:34] well enough to accurately judge what alphago is
[01:18:36] to accurately judge what alphago is doing well why is sweden now
[01:18:38] doing well why is sweden now this one looks like white is winning ah
[01:18:41] this one looks like white is winning ah we all say some of our pago moves are
[01:18:45] we all say some of our pago moves are so weird and strange and baby mistakes
[01:18:48] so weird and strange and baby mistakes but after a game is finished
[01:18:51] but after a game is finished we have to doubt ourselves our judgement
[01:18:56] we have to doubt ourselves our judgement alphago making another kind of
[01:18:57] alphago making another kind of nonsensical throw-in
[01:18:59] nonsensical throw-in we're not really sure what that's about
[01:19:01] we're not really sure what that's about this is what 10 or maybe 11. play
[01:19:03] this is what 10 or maybe 11. play looks like it looks weird and we don't
[01:19:05] looks like it looks weird and we don't quite understand it
[01:19:06] quite understand it i think it is important to study
[01:19:09] i think it is important to study more about hypogo's mistake like moves
[01:19:13] more about hypogo's mistake like moves then maybe we can adjust our knowledge
[01:19:16] then maybe we can adjust our knowledge about go
[01:19:17] about go to me the most amazing thing to come out
[01:19:19] to me the most amazing thing to come out of my understanding of go
[01:19:21] of my understanding of go as a result of watching alphago play are
[01:19:23] as a result of watching alphago play are the the infamous
[01:19:24] the the infamous slack moves well there's something
[01:19:26] slack moves well there's something strange about the way it's playing
[01:19:28] strange about the way it's playing because it's playing some moves that are
[01:19:30] because it's playing some moves that are not
[01:19:30] not really necessary right a slack move is a
[01:19:34] really necessary right a slack move is a move that looks
[01:19:35] move that looks lazy you can see these other better
[01:19:37] lazy you can see these other better moves and alvago is rejecting them but
[01:19:40] moves and alvago is rejecting them but what i think alphago is teaching us is
[01:19:43] what i think alphago is teaching us is that
[01:19:43] that we've been using score as a proxy
[01:19:47] we've been using score as a proxy for chance of winning so the bigger my
[01:19:50] for chance of winning so the bigger my margin
[01:19:51] margin of territory the more confident i am
[01:19:53] of territory the more confident i am that i'm gonna win
[01:19:54] that i'm gonna win and alphago saying no no it shouldn't
[01:19:56] and alphago saying no no it shouldn't matter how much you win by
[01:19:58] matter how much you win by you only need to win by a single point
[01:20:00] you only need to win by a single point why should i be
[01:20:01] why should i be seizing all this extra territory when i
[01:20:02] seizing all this extra territory when i don't need it the lessons that alphago
[01:20:04] don't need it the lessons that alphago is teaching us
[01:20:05] is teaching us are going to influence how go is played
[01:20:09] are going to influence how go is played for the next thousand years am i
[01:20:12] for the next thousand years am i counting why won this game
[01:20:14] counting why won this game how by how much are we talking two
[01:20:17] how by how much are we talking two points
[01:20:18] points [Music]
[01:20:24] go is just a game but we can learn important lessons
[01:20:25] important lessons from our computer be so successful at go
[01:20:28] from our computer be so successful at go machines will have the capability
[01:20:30] machines will have the capability not only to crunch through huge amount
[01:20:32] not only to crunch through huge amount of data
[01:20:33] of data but also to analyze it intelligently
[01:20:36] but also to analyze it intelligently just as
[01:20:37] just as in the case of the go games the machine
[01:20:39] in the case of the go games the machine made moves that
[01:20:40] made moves that surprised even the experts and
[01:20:43] surprised even the experts and eventually the machines will gain our
[01:20:44] eventually the machines will gain our confidence because we will see that very
[01:20:46] confidence because we will see that very very often
[01:20:47] very often they make a better guess than we could
[01:20:49] they make a better guess than we could have made as humans
[01:21:26] i think white might have a slight advantage here
[01:23:25] i mean i think it's just really is uh once in a lifetime thing
[01:23:27] once in a lifetime thing i would say it's the it's the most
[01:23:29] i would say it's the it's the most amazing thing yeah i've experienced
[01:23:31] amazing thing yeah i've experienced you know for us it's the culmination of
[01:23:33] you know for us it's the culmination of 20-year dream
[01:23:35] 20-year dream it started as a pure research
[01:23:38] it started as a pure research endeavor we just wanted to understand
[01:23:39] endeavor we just wanted to understand can neural networks play the game of go
[01:23:42] can neural networks play the game of go and from there it went on to a level
[01:23:46] and from there it went on to a level that i never expected and i'm
[01:23:49] that i never expected and i'm unbelievably proud of the team
[01:23:52] unbelievably proud of the team when i started doing artificial
[01:23:54] when i started doing artificial intelligence it was
[01:23:56] intelligence it was four or five years ago and i was really
[01:23:58] four or five years ago and i was really interested in that but like many people
[01:24:00] interested in that but like many people would discourage me and they would say
[01:24:01] would discourage me and they would say okay there is no future
[01:24:03] okay there is no future so actually seeing that within five
[01:24:06] so actually seeing that within five years
[01:24:07] years we are at this stage right now it's
[01:24:09] we are at this stage right now it's amazing by itself
[01:24:17] there are so many possible application domains where creativity
[01:24:20] domains where creativity in a different dimension to what humans
[01:24:21] in a different dimension to what humans could do could be immensely valuable to
[01:24:23] could do could be immensely valuable to us
[01:24:24] us and i'd just love to have more of those
[01:24:26] and i'd just love to have more of those moments where we look back and say yeah
[01:24:27] moments where we look back and say yeah that was just like move 37
[01:24:29] that was just like move 37 something beautiful occurred there at
[01:24:32] something beautiful occurred there at least in a broad sense move 37 begat
[01:24:34] least in a broad sense move 37 begat move 78 begat a new attitude at least
[01:24:38] move 78 begat a new attitude at least a new way of seeing the game he
[01:24:42] a new way of seeing the game he improved through this machine
[01:24:45] improved through this machine his humanness was expanded after playing
[01:24:48] his humanness was expanded after playing this
[01:24:48] this inanimate creation and the hope is that
[01:24:52] inanimate creation and the hope is that that machine and in particular the
[01:24:54] that machine and in particular the technologies behind it can have the same
[01:24:56] technologies behind it can have the same effect
[01:24:57] effect with all of us i remember hearing a talk
[01:25:01] with all of us i remember hearing a talk by kasparov
[01:25:02] by kasparov who says that a good human plus a
[01:25:05] who says that a good human plus a machine
[01:25:06] machine is the best combination this
[01:25:10] is the best combination this is a unique experience nobody can have
[01:25:13] is a unique experience nobody can have this experience
[01:25:14] this experience play with alphago fight games like this
[01:25:17] play with alphago fight games like this so
[01:25:18] so i hope lisador can find something in
[01:25:21] i hope lisador can find something in this five game
[01:25:22] this five game maybe some change in his game
[01:25:26] maybe some change in his game i see today i fight fight fight
[01:25:30] i see today i fight fight fight it's very good big master real big
[01:25:33] it's very good big master real big master
[01:25:39] even though i
[01:25:39] i [Music]
[01:26:09] me is done but for all the
[01:26:12] is done but for all the story maybe it's just the beginning we
[01:26:14] story maybe it's just the beginning we don't know
[01:26:21] it's just when i play with alphago he showed me something
[01:26:23] showed me something i feel beautiful just it
[01:26:26] i feel beautiful just it i see the world different before
[01:26:29] i see the world different before everything begins what is
[01:26:33] everything begins what is real thing inside the go game with this
[01:26:36] real thing inside the go game with this thing
[01:26:37] thing i will change something with my game
[01:26:45] maybe he just can show humans something we never discovered
[01:26:48] we never discovered maybe it's beautiful

All frames

Total: 40. Hero frames flagged with star.

• frames/frame_0001.jpg (t=00:00) frames/frame_0002.jpg (t=01:10) frames/frame_0003.jpg (t=03:15) frames/frame_0004.jpg (t=05:23) frames/frame_0005.jpg (t=07:36) frames/frame_0006.jpg (t=09:50) frames/frame_0007.jpg (t=12:02) frames/frame_0008.jpg (t=14:06)
• frames/frame_0009.jpg (t=16:22) frames/frame_0010.jpg (t=18:25)
• frames/frame_0011.jpg (t=20:41) frames/frame_0012.jpg (t=25:00) frames/frame_0013.jpg (t=27:11) frames/frame_0014.jpg (t=29:20) frames/frame_0015.jpg (t=31:31) frames/frame_0016.jpg (t=33:40)
• frames/frame_0017.jpg (t=35:48) frames/frame_0018.jpg (t=38:06) frames/frame_0019.jpg (t=40:12) frames/frame_0020.jpg (t=42:22) frames/frame_0021.jpg (t=44:30) frames/frame_0022.jpg (t=46:45) frames/frame_0023.jpg (t=48:49) frames/frame_0024.jpg (t=51:06)
• frames/frame_0025.jpg (t=53:14) frames/frame_0026.jpg (t=55:24) frames/frame_0027.jpg (t=57:39) frames/frame_0028.jpg (t=59:46) frames/frame_0029.jpg (t=01:02:00) frames/frame_0030.jpg (t=01:06:18) frames/frame_0031.jpg (t=01:08:33) frames/frame_0032.jpg (t=01:10:36) frames/frame_0033.jpg (t=01:12:52) frames/frame_0034.jpg (t=01:15:08) frames/frame_0035.jpg (t=01:17:07) frames/frame_0036.jpg (t=01:19:23) frames/frame_0037.jpg (t=01:21:31) frames/frame_0038.jpg (t=01:23:47) frames/frame_0039.jpg (t=01:25:45) frames/frame_0040.jpg (t=01:26:58)