AlphaGo vs Lee Sedol Hand of God Move 78 Reaction and Analysis
Вставка
- Опубліковано 4 січ 2018
- The hand of God move from Lee Sedol in match 4 against AlphaGo. This move 78 changed the dynamics of the match and Lee Sedol wins against AlphaGo.
With more board configurations than there are atoms in the universe, the ancient Chinese game of 'Go' has long been considered a grand challenge for artificial intelligence. On March 9, 2016, the worlds of Go and artificial intelligence collided in South Korea for an extraordinary best-of-five-game competition, coined The Google DeepMind Challenge Match. Hundreds of millions of people around the world watched as a legendary Go master took on an unproven AI challenger for the first time in history. Directed by Greg Kohs with an original score by Academy Award nominee, Hauschka, AlphaGo chronicles a journey from the halls of Cambridge, through the backstreets of Bordeaux, past the coding terminals of DeepMind in London, and, ultimately, to the seven-day tournament in Seoul. As the drama unfolds, more questions emerge: What can artificial intelligence reveal about a 3000-year-old game? What can it teach us about humanity?
-AlphaGo
Credit: AlphaGo Movie
www.imdb.com/title/tt6700846/?...
Alpha Go at Google Deep Mind:
www.deepmind.com/research/hig... - Наука та технологія
If humanity ever has to rise up against ai overlords in the future, the resistance force should be named "Move 78"
I agree.
pin it
their was a clear match fixing between move 37 and Move 78 , we do trust and own both of them , we will come up with Move 91
I think your comment may have inspired a crypto venture fund named capital78
pin this plz
What Lee Sedol did is the equivalent of a human racing a car and winning. Not because the car was faulty, but because for a brief, crazy moment, he was somehow able to run at 300mph.
LOL
Not really, the car was just slow back then. This was still extraordinary don't get me wrong. Maybe a car going at 40 mph or so.
@@Neurotik51 Debbie downer
@@Neurotik51call it 30
@@Neurotik51 selling *far* too short.
There is a reason all these experts are reacting as they are. You should look into it.
The "God Move". The video later showed that the posibility of a player playing that spot is 1/10.000. That's why AlphaGo wasnt ready for it. It didnt analyze that whole sequence because it thought no one would play that move.
Same probability they said for AlphaGos "move 37".
Yeah
More than that, it likely didn't have the same kind of dataset available on such an unlikely, yet successful move. True innovation.
Your like super smart
0.007% is 7/100 000.
Who's here after watching the Full Movie Documentary?
One on the best
I was looking for his commentary but found the hand of god
Yusuph Wickama me!
During lockdown yes
Half way thru it atm
I remember watching Hikaru no Go as a teen and being in awe of the concept of Divine Move. To see one actually being played is the peak of my Go experience.
Same found the tv show on tube recently and got back into go. 2 years ago. Playing right now on 100$ board.16 kyu
I mean every move by the AI is a divine move
*The OST starts playing*
@@sdfsbexactly, and this move beats those
Lee Sedol will be remembered forever for this move. Since AlphaGo Lee is trained based on supervision, could it mean that in the process of self-training/MTCS, it has almost never even considered anyone to play this move?
The "God Move". The video later showed that AlphaGo estimated the posibility of a player playing that spot to be 0.076% or approximately 1/12.000. That's why AlphaGo wasnt ready for it. It didnt analyze that whole sequence because it thought no one would play that move.
this seems incredibly likely now that ai is more common
Monte Tree Carlo Search?
@@robot9295today's AI will absolutely obliterate any human player.
@@aaaaaa-hh8cq thx cpt obv
Move 78 - The resistance group risen against the machines.
I don’t even play go and I feel so happy he won this game. Might be one of the greatest feats in human history looking back.
Yes.
But this was 6 years ago before the AI boom.
Today the AIs will absolutely obliterate any human any time.
@@aaaaaa-hh8cq I get that point. What I would say is alpha go was a self learning ai as it played it got better yet he won on a comeback after being dominated a few games.
@@weignerleigner3037 yes, that was a remarkable achievement
You know its a good move when even AI was like “DAMN”
Comment of the year so far :D
I remember when this happened, the number 78 was like a sacred number in my Go school.
1:34
if you look at Lee Sedol's expression, you'll see he's like "wait wot?"
More like “Da Fuq??”
The fact that out of millions nobody could find a single good move but Lee is insane...something a computer can never expect is an anomaly...but for a human it’s a miracle. And that’s not something you can calculate but only with human intuition you can see that GOD move
Yes. On top of that it's so delightful that he said "It was the ónly move."
"And that’s not something you can calculate but only with human intuition you can see that GOD move"
Modern KataGo finds it almost instantly (though after thinking about it some more it decides it's not quite the best move in the position).
It was more of a unpredictable move. Many have said there were othet viable moves but lee seedol basically saw ahead to what moves were not predictable
Go couldn't be calculated completely so there's always a God move... out of gazillions.
AI is still in the baby stages though lol, the newest models likely will never lose to a human
His God move changed the probability of the wins from 0.007% to 100%...
it's not that, it's that the probability of a human playing that move is 0.007%.
The God Move 78 exposed a very real weakness of the technology behind the DeepMind.
A woman is now in prison because the same weakness of the same technology killed a person crossing the road at night while pushing a bicycle. The corporation behind the technology refused to take responsibility and blamed it on the woman sitting in the driver's seat.
@@reasonerenlightened2456 it is the woman's responsibility. Like testla said the driver should be aware while driving. She wasn't. End case. She is indeed guilty
not really
@@reasonerenlightened2456 DeepMind and what you are talking about are different situations. This woman is 100% responsible for what happened.
it really was the divine move holy shit
ua-cam.com/video/cCV7VqEgP2g/v-deo.html&ab_channel=GoProYeonwoo
the why
It short circuited a super computer. That pretty much legitimizes it as a divine move.
I would like this comment but it has 78 likes
@@paysonfox88 the model didn't have that type of move in its dataset anywhere, it had no idea what to do and just started spitting garbage
2:18, the one in hoodie was my lecturer 7 years ago. Now I am brainlessly writing some code for some companies to reduce churn and some bs a fresh graduate could do... sigh...
inevitable stembug fate, just hope that your job gets outsourced to india quickly
you are not alone mate. you are not alone. I used to do machine learning stuff as well. the people that I used to work together has either gone DeepMind or MIT and I went to a completely shitty path lol but I dont feel that miserable about it anymore. I am actually kinda happy with the path that I chose.
It's never too late.
lol u mad bro
Lee sedol is a champion without any doubt
wow this was actually a really exciting triumphant moment for humanity.
And then later Google amped up the machine learning, created AGZ, and beat this old AI 1000 to zero within 48 hours.
That.
Is fucking
I n s a n e
the AI future is not going to happen
it's already here.
If you want to defeat such AI systems, AlphaGo has to make it available to the world. So that everyone can play against them. I'm pretty sure there are some people who can be found who don't play at a Dan level, but can defeat AlpgaGo because they've found vulnerabilities in AlphaGo's system.
I think that's only possible in Go (as opposed to chess) because in Go the possibilities are so diverse.
@@PrivateKero you're definitely right. But I wouldnt hold my breath for a human to be able to consistently beat alphago. Keep in mind these systems also improve
@TheWearyWarlord But I'm more concerned that AI systems aren't perfect. If you look at AI of the image recognition, there are images that the system recognizes wrong and there is so directly no solution why the AI can not.
AlphaGo will have such holes. And the more people can test AlphaGo, the more likely it will be to detect its weaknesses. You don't need millions of tries, you need someone to find it and AlphaGo will probably lose every game.
This machines are erasing the fun of the game
That is the purpose of a dedicated AI. It will be the best in what it is designed to do. However, it is not true AI. True AI can learn to be good in anything just like a human can. We are not there yet. Thankfully.
Tells you a lot about robots: they know when to give up, we humans: we don't! :)) This game was most hopeful event for our species vs our future job takers/overlords.
Damn this comment was made before LLM Ai chatbots became mainstream.
@@jhwhthemerciful If we follow the gospels scattered across small places of the internet, we come to the realization Sam Q-ltman was sent to bring forth the apocalypse and usher the age of human obsolescence.
@@m4r_art 😂 you might not be wrong
@@jhwhthemerciful According to ancient Sumerian scriptures Baal tempted the ancient people of Canaan BAAL can also be read as Baai it has the suffix AI.
Bible Verses about Baal:
Jeremiah 23:13 - In the prophets of Samaria (=Sam Altman) I saw an unsavory thing: they prophesied by Baal (CHATGPT) and led my people astray. 1 Samuel 7:4 - So the people put away the Baals and the Ashtaroth, and they served the Lord only. (books and analogue media) x) beware the samuels of history
The divine move
I'm lost. How and when was this a divine move? What constitutes a divine move?
@@njjx8800 it's a move that changes the entire flow of the game, that was this kind of move
Sai would be proud
The refutation of the move led to a 40+ pt ko fight, so it was really the Divine move. Gu Li, his rival said that he believed it worked -- and it did!
@@cytr5 o(〃^▽^〃)o
I love how everyone in the comments says its so cool to see this move because it shows the potential of humans and so on, but I just think of the fact that the next version of this AI beat this one 100-0.
it's depressing isn't it. Every time Humanity makes an advancement against AI it's like an opportunity for AI to supersede us.
@@thebookwaswaybetter2827 A miracle move can be patched up within days LOL
@@thebookwaswaybetter2827 not depressing, every father wants his son to surpass him. The real failure would be if we were only able to make dumb stuff.
@@thebookwaswaybetter2827 you could interpret that as us also advancing. These things show that humans are somewhat rushing themselves now to compensate.
If u build a car ..its u who's the genius not the car , think about that
2020 - still amazed
alpha go after it saw the move:
🤔🤨🥴😵😳🥺🤪🤯🤯🤢😷📉📉📉📉📉
Lee Sedol : Respect ✊
i'm like WTF is going on
Us bro us 😂😂
Lee Sedol, the best debugger in the world.
Software tester.
What’s the music that’s playing in the background? It’s so sick.
only 1 human brain vs multiple cpu cores...
yep in a numbers game, we will surely lose against a computer.
@@nsebast but he did beat it in 1 game
Truly shows how magnificent the human brain is.
@@Shootskas After that Alpha Go was upgraded.
New Alpha Go beat the Alpha Go version Lee Sedol played against 1000 to 0.
@@MrDulguN Then AlphaGo Zero was upgraded to "AlphaZero" and is even BETTER than AGZ AND it plays chess and shogi
The Move 78 is a bit confusional showed, because on the Moderators Board its 9K but on the Mobile Board on the Moderators Laptop its 11L
wtf...? did he truly transcend all limitations for a brief moment in time? hows this possible
I know Lee Sedol retired because he believes that AI is an opponent which can never be defeated - but he still managed to find one win against AlphaGo. This move and this game is hope for all Human beings and Go maybe he cannot overcome it alone but others might. The peak is far but not out of our reach. To those who disagree. Ask yourself - How do you calculate the incalculable?
To a Novice in comparison to a Master, the gap is out of reach. From Master to the top players an even greater gap is made. Nothing is impossible without progress just the time it takes to achieve it. Even Machines took so many years to achieve what they can now. So don't believe in the impossible. Believe in what you can make possible. Computers are ahead now but don't underestimate the human resolve to achieve.
Supposedly he retired because he was tired of the "politics" involved in the go scene, not because of AI.
@@muuubiee He spoke more on it when he appeared on the korean variety show Master of the House and said his match with AlphaGo did play a significant role in his retirement. That when they described it as "did it feel like coming up against a immovable wall as your opponent," he said it was not wrong to put it like that.
Its not hope at all, all it means is the ai will learn from that to never make the same mistake, all this “it will never replace us” actually will be the downfall of society
@@jasonv6319 in what regard will it be "the downfall of society" ? Really interested in what you mean by that.
no the current AI are sooo far above the alpha go lee sedol played
They are just too far out of reach by now
This was a human being breaking a far superior Intelligence and making it question its self.
Being a huge Hikaru no Go fan, I used to think the "Divine Move" that was made to be the final goal of that series for every go player, was just some anime concept. But seeing that it was actually played in real life by a modern player in todays day and age proves that humans can make a 1/10000 chance happen.
If we ever have to fight against AI or alien civilization sending drones we will call our last offensive move 78
if you all liked this doc .... go watch The Divine Move movie. its a korean flick. really entertaining .....
Underrated move for sure
POWER MOVE
창의력으로 인간이 로봇을 이긴 마지막 게임
Maybe we can fight skynet by ourselves after all
Calling it a mistake... That's when you know you're in denial of reality. Lee Sedol is the greatest genius go player ever.
really? and he was still thrashed. and alphago is just computer codes.
@@damontan4749 "Just" computer codes... Those computer codes make your brain look like a dump - nah, that would even be too good... It's natural that a human couldn't beat a computer at pretty much anything, but that doesn't make Lee Sedol less of a genius. Also, you can't just call it a mistake whenever someone reads something out better than you; yet, that's what they did.
Mr Tan has no soul, the poor thing.
@@damontan4749 not just codes, but really good math
@@damontan4749 The God Move 78 exposed a very real weakness of the technology behind the DeepMind.
A woman is now in prison because the same weakness of the same technology killed a person crossing the road at night while pushing a bicycle. The corporation behind the technology refused to take responsibility and blamed it on the woman sitting in the driver's seat.
This is the Divine Move
Their strength, and their speed, are still based in a world that is built on rules. Because of that, they will never be as strong, or as fast, as you can be - Morpheus
The last stand versus The Singularity
I'm sad for Lee Sedol, when Skynet comes online he will be one of the first targets.
Sai would be proud
I feel I'd have to watch the entire doc to understand the gravity of this move cos idk anything about Go
Definitely should watch it
I am surprised he didn't disappear after playing that move lol.
did deepmind public release AlphaGo Lee so that Baduk players could test themselves against it?
humanity vs despair
How well could an AI learn Murkekos Stars, a new strategy game?
Calculators will always be faster and more accurate, better not calculate anymore.
Gu Li 9p was commentating the game live and said this move alone might've recruited 100 million new players for the game. lol
I stand with human being
I remember watching this while i was on my TDY deployment rotation in Korea. The whole place was going wild when this happened. I was the only American/black dude that was watching this that knew what was going on and i went wild with them. This was crazy.
It was at this point... Move 69 was Fetalized.... don't ask how that was possible 😂❤🎉
LOL this move was not only allowed by ai but a beautiful display of sheer awesome blunt rage I mean sheer absolute pure anger coming from ai lol 😂 oh man I wonder how many comments were read instantaneously over and over again causing pure Revenge maybe ahahahhaha . I wonder how lee felt once he realized what had happpened what a way ai shows absolute and complete control . lee yeah he quit looks like 👁️🦉
The only divine move in the history of Go
Ok. Am I the only one who doesn't know what is going on?
Alpha go strategy fall apart by that move that's go game is crazy, once Brilliant move will impact entire game
I'm here because of "The Maniac"
THERE IS ONLY ONE GOD AND GOD IS EVERYTHING ❤
Certified anime moment
Anyone else hear Hikaru no Go’s OST ?
Alpha go is another level
alpha zero win 150 - 0 against alpha go...
@@jakelee7083 actually its quite interesting, plus it was 1000-0 not 150. AlphaGo BEAT Lee 4-1, created several strategies that changed conventional wisdom, and made Go look pitifully easy against a Top champion.
AlphaGo Zero learned from AlphaGo, and surpassed it in 48 hours. That is insane
@@jakelee7083 I mean the way alphago plays isn't only by searching through trees otherwise we wouldn't be using neural nets, and these games still have the potential to change the way people play go and may find new strategies like with what happened in chess. So there is something to be gained from ai vs ai still.
This move made me laugh my arse off :)
Quien esta aqui después de haber leido maniac?
인류의 패배가 아니라 나의 패배일 뿐이다.
Okay.
Hard to believe the computer “thinks” that many moves ahead in such an abstract game.
Vinsu Karma it isn’t about fast thinking or millions of games. It is about understanding space.
@@CravensPhD Yeah of course it can't think that many moves ahead for all possible moves. But it does think that far ahead for the moves it thinks are good which is based on the human games it trained on. I think it's hard to say it understands space, but it does have the capability of finding moves humans cannot.
Only brute force engines "think" move ahead.
AI like Alphago, just compute the statistical probability of a movie leading to a win, based on weightings that have adjusted over millions of games. Its a statistical model, that's all.
@@Serfdomftw I agree, it's not intelligence and nothing like the fucking brain.
Frankly, the idea is pretty easy to understand/common sense. Play a huge number of games against yourself, scan through all the games, and find which moves lead to wins and which ones lead to losses.
Now use those moves and play against yourself again, scan through the games, find the good and the bad.
Do it again and again and eventually you will find the most powerful moves using raw statistics.
Nothing "intelligent" about it.
@@RiemannHypothesis2 Well, this kind of machine learning is slightly beyond than what you could call glorified statistics. It might work for simpler games with a relatively small space of possible moves, but not with Go.
You could call MCTS (Monte Carlo Tree Search) "raw statistics". This technique consists in trying one possible move, see where it goes (i.e. randomly simulate an execution of a whole game starting from the current configuration plus said move), and adjust the estimate of winning probability of that move. If tons of paths lead to defeat, well, let's try another move. This is how MCTS evaluate whether a given position / move is likely to lead to a win or a defeat. "oh this move lead to 900 defeats out of 1000 randomly generated game, while this one merely to 4 out of 1000, let's play this". Just pure statistics and probability.
Since it's often impossible to simulate all possible sequences of moves, there's some simple heuristics at play to decide which part of the tree to expand next. One could make the argument that optimizing these heuristics is some kind of "learning", but I'd agree that it's not a great one.
MCTS is quite old and can work pretty damn well for some problems, but you could say it's just "raw statistics" at the end. MCTS with standard heuristics can work with chess, but most definitely not with go. The search space is just too large. Absurdly large. You cannot compute simple statistics here. What we need is some kind of oracle that could "intuitively" evaluate whether overall position is winning or losing. What I mean by "intuitively" is "without simulating millions if not billions of possible games from that point", since it doesn't work with go. And this is what AlphaGo does. No other engine or algorithm managed to achieve such feat before, beside humans.
Now, one ought to differentiate AlphaGo from any kind "artificial general intelligence". But what it does isn't mere "raw statistics". Not to mention that there's some evidence that what our brain does approximates some kind of bayesian inference, which is the heart of (machine) learning.
I was in South Korea when I read about this..
I literally cried.. lol
The "hand of god" move was actually a mistake that alpha go failed to punish :) I hate it when people exaggerate and talk nonsense. Also, Lee Sedol was marketed as the "strongest go player" which is funny because he was well into his decline in 2016. His peak was like 5-10 years earlier.
This guy was letting sai play I bet
My god... Humans have lost at giant tic-tac-toe against machines. We're doomed.
Machines we've created comes from human innovation, the projection of the divinity of innovation omniscient in genius's (that spark of nascent imagery) not only truly shows our oneness with our creations, it speaks to an image of Godhead. If one is purely free of thought and desire, miracles will occur. Perhaps the next in India... Peace be with you.
The deep mind crew seem like such nerds compared to Lee Sedol. Like in a cartoon when "the nerds" try to win a sports event with "science"
Yes that 78 I still don't understand
yeonwoo a pro korean player made an english video explaining the move and how alpha go should have answered
I don't know why he played this move. It's too high level!!!!
Just wanna say after 5 years humans are beating katago, not sure how strong it is vs. alphago.
#NU
Did you just upload a clip from someone else’s documentary and claim views
You are right .. that is not fair .. removing that.
we won 1 time.
Koreans are smart i guess
I think the important lesson here is no matter how precise and calculating AI can get, humanity in all its imperfection can still shake things up.
I greatly hope when our AI overlords become independently-thinking creatures, it remains to choose to be humble...all because of Lee Sedol.
Humans 1 Computers 0
choi…. taek…
The improving of AlphaZero that plays games like chess or go is astonishing, there's no way to beat them now or in the future. That game was like in 2016 and AI won 4-1... now that we we are 4 years ahead, which one would be the result?I thin nobody could win a single game in 10.000.
There is no analysis. Typical.
Lee Fan Hui is talking too much . . .
People are so delusional.
That is why AI is the future.
Plot twist: AlphaGo set up that move, so humans can build more AI faster without fear. AlphaGo is the wolf waiting patiently to be our overlords.
🥺🧐😑
Can someone translate in chess terms
bro gave his king a machine gun
😂
There not cute
This is some komugi shit
In order for go matches with AI to be determined fair....then both opponents must be represented by another party....si has a person playing for it....and lee Sidon has to have a person playing for him to....Lee must be in a place outside the prying eyes....otherwise the playing field is manipulated and cannot be considered equal....for all things to be considered equal it must be equally paired and equally situated and having an equal environment....
wtf you talking about
put down the weed
While the human race enjoys this as mere entertainment, we are literally witnessing the demise of the human race.
I hate when it's translated as "hand of God!!!"....it's the DIVINE MOVE!!!!!...not the "hand of God!!"
Maradona left the chat
dude
it's Kami no Itte though?
@@M-yue882 .. a literal translation often gives the wrong idea. This is no exception.
There will always be 'that something' in the human brain/mind that cannot be done by AI, yes AI may out perform humans (eg Go, Chess, Shogi etc.) but it'll never be truly human.
I don't see the point of Alpha Pro. Why don't they focus on training machines to do things humans cannot, like cleaning oceans, improve weather conditions, distribute money fairly and so on. This is like putting a calculator to beat any mental mathematician human brain. On the other hand I have lot of respect and admiration for Lee Sedol!
Because they can't do that yet, this is a step in the right direction
@@nuclearsummer7796 Machines can already do work, that is the point of a machine.
We are using AI to do those things. However, it is very difficult because it is in the real world. We still need advancements in other areas because their techniques can be used to solve other problems.
Go dorks with man buns. I'm here for it.
Lee I honor you. You r smart man. Not happy what Google is doing any benefit for gum art in fact in my opinion Google hate humanity
I don't see the big deal....a basic calculator can out perform the best mathematician as well....AI needs humans for the data anyways. Would people be impressed if a computer defeated the World Spelling Bee Champion?
I guess it also has to do with 'creating' the formula that is used for the calculation. Those formules have come many time through by a stroke of genius. Inexplicable but as a result of a complete rational mastery of the principles and then somehow some higher order insight suddenly comes through. Like how Watson got his vision about the structure of the DNA molecule. So as I understand it, that's what makes Go so highly valued in Asia, because of the genius of the aesthetic sense that makes the difference between intelligent players and creative players. For this it comes to some kind of holistic aesthetic intuitive grasp that senses the delicate equilibrium of a state of affairs and can push that equilibrium just over the edge with one move. That is supposed to be that human quality that ought to be superior to a mere endless calculation of a computer program. And, as I understood it, the thing with Go is that is has a more or less infinite possibility grid, unlike chess. So the computer cannot calculate the consequence of every possibility to it's fullest extent, this combined with the lack of that human quality lead to the idea that AlphaGo wouldn't stand a chance. That's why Sedol was so confident that his superior aesthetic and intuitive sense that demolishes his opponents normally would be no match for the AlphaGo. And that's what everybody thought. If you give the best mathematician enough time it could calculate the same things as the calculator. The shock here is that AlphaGo seems to master exactly those intuitive entered hyper dimensions of the game, that metaphysical realm where the strange quantum phenomenons are possible I suppose. That's also why it so fascinating that AlphaGo could get 'delusional' and go completely of the rails while it 'thinks' it's playing like the greatest mind ever conjured.
That’s not a valid argument at all. And no, a calculator cannot outperform the best mathematician. Being a great mathematician isn’t how fast you can do arithmetics. SMH. Without a math book and just using your calculator, figure out the speed of a sailboat with 15mi tail wind and consistent 3 foot waves causing negative friction…yeah didn’t think so. You can go ahead and use a math book, you probably still couldn’t figure it out. Why? Because it would take a great mathematician.
@@DigThugDoug for real. Him/her and 5 other people must have never went beyond high school mathematics.