GO champ Ke Jie talks about his match with Google's AI

The future belongs to artificial intelligence (AI), said the Chinese Go prodigy Ke Jie after he was defeated by the Google-owned AlphaGo computer program.
The world's number one Go player gave an exclusive interview to CGTN on Sunday, a day after his defeat in the town of Wuzhen, east China's Zhejiang Province.

[@raymondge1520]

At 2:01 Ke Jie says "他实在太狠了“ which the channel translates to "AI is really Awesome", but in actuality, it means "The AI is way too brutal." Just something to think about.

[@haoqi9596]

"狠", under that situation, means aggreesive with no mistake. Your statment showed the importance of interpreter, direct translation cannot solve all the laguage problems.

[@Ohmriginal722]

Lee Sedol was Garry Kasparov battling Deep Blue and Ke Jie is Magnus Carlson messing around in Stockfish knowing there's no hope.

[@JonathanXLindqviust]

A very bad analogy because the computing, thinking and "feeling" of a Go-AI is entirely different than the mechanics of Deep Blue and Stockfish. The chess AI's already "know" every possible match, they literally calculate every move from a spreadsheet of all possible outcomes and will always pick the one leading towards victory. The Go-AI is more like Magnus Carlson writing gibberish on google-search and the search-engine still managing to go "Is this what you were looking for?" and finding the perfect answer regardless of what was written.

[@Ohmriginal722]

@@JonathanXLindqviust Chess is not a perfect game, there’s a reason that the latest version of stock fish combines the traditional symbolic chess approach with neural networks like the kind used by alpha zero and Leela zero. Chess may not feel the same as go, but does still very much rely on an element of intuition.

[@huleyn135]

@@JonathanXLindqviust No offense but you know very little about what you're talking about as you compare the two games and the two engines, the initial analogy is very much on point. Modern chess engines are very similar to AlphaGo. They don't "know literally every move". They don't search millions of moves deep like older engines, they don't rely on brute force. They're just good players.

[@JonathanXLindqviust]

@@huleyn135 Read your first line, then nothing more. I am happy in my knowledge that that'll haunt you. Petty punishment for petty minds.

[@dipjyotimitra186]

@@JonathanXLindqviust huleyn135 is right, bro. The main difference between AlphaGo or the new AlphaZero from any previous software is that IT DOESN'T SEARCH EVERYTHING. It is this "Human-like" intuition that is making it so powerful. AlphaZero searches 60,000 moves per second compared to 60 million moves per second for stockfish So, it's quite opposite than what you said! There's still NO machine that knows all moves; then all the future match would become draw! (as such machines would play with each other)

[@ThatJourney]

"Ill leave 'him' to the task of exploring the universe...." My god it's happening already.

[@infinitysalinity7981]

Exploring is good fun though

[@puct9]

It's possible that is a translation inaccuracy. In Chinese, there are pronouns for him, her, it (他，她，它), however, they are all pronounced the same. Your point is still valid. There are numerous occasions where commentators or other people have referred to AlphaGo with a term typically reserved for humans or living beings, but just a little insight from me.

[@kathyxu1513]

There’s no phonetic differences between he/she/it in Mandarin, i think the translator added a bit of their own interpretation there.

[@kimshakhov1516]

These are the words that make you think

[@malharjajoo7393]

terminator.skynet.

[@gabrielgonzales5907]

Actually, Lee Sedol beat the AI just once. To be fair, though, the program may not have been as polished as it was when it played against Ke Jie.

[@MrSpiritmonger]

There is a new AI called AlphaGo Zero that beat original AlphaGo software 150-0, so even AIs can beat AIs.

[@beasthundred4709]

True. But Lee Sedol was not in his prime. Ke Jie was.

[@huleyn135]

@@beasthundred4709 Except that doesn't matter. Humans against humans are not so different. But the variant that fought against Ke Jie was an order of magnitude stronger than the variant that fought against Lee. Becomes even worse considering this is alphago Master, but alphago zero is MUCH stronger than that also.

[@asianamericancasestudies6434]

Ke Jie reviewed the play by Lee Sedol, that's why he made the big claims. But the new AlphaGo is 1000x more complex.

[@aeacussheng2339]

AlphaGo Master is way more powerful than AlphaGo Lee, the ELO rating is about 1000 higher.

[@rangerroe3739]

Beautiful analogy about the universe and the pond

[@giusgra7411]

it's so weird to consider how he looks like the chinese version of Magnus Carlsen..

[@joseluna6262]

So true!!

[@KnifeYoMom]

You may not like it, but this is what peak performance looks like.

[@MrWizardjr9]

@Anonymous Panacea magnus sometimes wears glasses also

[@szymonharbuz9052]

I just wanted to write that. The resemblance is uncanny.

[@jaycho6747]

That's what happens when you play too much board games...

[@claudiaaaa2476]

Fun fact: Ke Jie has a dog named Alpha, and the word "dog" in Chinese is pernounced "Gou", so it's a "Alpha Gou"

[@STJukes]

Is that true?

[@claudiaaaa2476]

@@STJukes yep :) he posted videos about his dog on his Bilibili channel too, calling it Alpha Gou in it

[@chengcongpan4108]

I can't find it in bilibili. where?

[@MundaneMuser]

Clever and cute.

[@Digdigwuhu]

I cannot beat you,but i can pet you instead

[@Yewbzee]

They have now created a new AI called AlphaGo Zero which beat the previous AlphaGo by something like 150 games to zero. Not only that it actually taught itself to play from first principles without any human teaching or intervention.

[@hey8174]

"Tabula rasa" style neural networks are the key to generalizing AI! AlphaGo Zero is not a go game engine, but generalized modular AI engine that can be applied to a number of different applications. The only thing it needs is a fitness function defining good or bad results, and some tools to affect those results. If you allow an engine to manipulate investments, manufacturing, news outlets, social media, education, legislation, military, the possibilities are infinite.

[@Yewbzee]

Tuc yes the power of this type of AI could be unbelievable. Nobody can guess when general AI will arrive but when I look at this type of software it makes me think it’s probably not too far away. Maybe less than 20 years?

[@hey8174]

I don't think we are many lifetimes away.. AI used to be a mirror of human behavior. Then we took two mirrors pointed inward on themselves with some fitness criteria to create AlphaGo Zero. If someone were to program a bot that prioritizes self sustainability rather than winning board games, we will begin to see programs that replicate artificial sentient behaviors to the point we can't distinguish computer from human. A nuclear bomb is a hammer, generalized artificial intelligence is on a different level.

[@Herman47]

The previous version was *AlphaGo Master (not AlphaGo Lee).* According to Wikipedia, AlphaGo Zero defeated AlphaGo Master 89 wins to 11.

[@infinitysalinity7981]

They already had Zero when Ke Jie played Master. So in a way, Deep Mind were holding back and they still managed to beat him.

[@u2blr]

He is one of the most talent Go player in the world. He is young, He could dominate Go for years. I feel bad for him.

[@Zarozian]

He needs to keep fighting it everyday. That AI is not a God but a measuring stick to help us evolve.

[@asianamericancasestudies6434]

He's the number 1 of the world.

[@asianamericancasestudies6434]

@@Zarozian Natural selection works to adapt to environment. Go is not.

[@asianamericancasestudies6434]

He's actually rank 1 in go after winning many championships. But he quit to go to the number 1 university in China.

[@Jason-jb1vs]

@@Zarozian dude the AI is learning faster than you if you fight it everyday, the AI three weeks later can beat the AI three weeks before in 150 - 0 fashion, you simply do not evolve faster than AI with a human brain. But what are the top ranking Go players do now is learning how AI plays and makes them better at playing other human.

[@Aramis7]

(I think) I can understand the sentiment of Ke Jie and Lee Sedol after having lost to AI. In particular Lee Sedol was going in very sure of himself and Alpha Go beat him easily (minus one game), and he was visibly shaken by the defeat. Not only it was a personal loss, but he was effectively representing humanity on the table. After the loss, it would be very natural to think of yourself (Lee) , or of all humanity for that matter, as a complete failure, being beaten by computer program barely a decade old. But this is so far from the truth. Computers will from now be the undisputed kings of Go (or any intellectual game for that matter) , but that doesnt diminish one bit the worth and intellect of human. Will we have human-machines matches in the future? Unlikely . Where's the fun of watching a snail competing against an F1 car or a pigeon against a F18? None. However machines can teach us new and innovative ways of playing the game that humans would have never thought of, and that's something that we'd all benefit from. Machines will enhance our game, not so that we can beat them, but that we play a more enjoyable game against ourselves. Go is a game, and as such its meant to challenge you and provide enjoyment (or despair...depending on which side you are on). Its not a life or death matter and humans will continue to play it even if a machine can easily beat us up. Let machines be machines, let humans be humans.

[@sutapagoswami116]

this is such a beautiful take. never thought about machines and humans co-exisiting in harmony before your comment.

[@Bea_remembrance]

Most ominous and prophetic interview on KZbin

[@michaelpace7847]

Damn, that AI ruined this guy's self-image

[@NO1xANIMExFAN]

yea no kidding

[@asianamericancasestudies6434]

It's like saying Bolt lost to my race car. Big deal...

[@asianamericancasestudies6434]

If the smartest guy says this, we're probably too dumb to even understand our stupidity.

[@davidyang6074]

He's not just any guy, he's the highest ranking player in the world

[@zwang3909]

He had a super long winning streak against other top human players after beaten by the AI. Seems like he learned a lot.

[@maqboolfida786]

This is the fate of the world champion in Go...then imagine the fate of mere mortals like you and me!

[@retryhikaru184]

"I really regret not losing brilliantly"

[@bruceli9094]

I hear there's going to be rematch. Alphago vs Ke Jie 2020?

[@852derek852]

Something about the reactions of the top go players to having been beaten makes me, as a human, feel deeply uneasy. I also don't like the fact that all of these powerful AIs are mostly in the hands of enormous corporations run by billionaires. What happens when AIs can create art, music, literature, science, philosophy, programming and mathematics better than any human?

[@kevzter15]

Komugi is better

[@dannygjk]

The version that beat Ke Jie is weak compared to the AlphaGo Zero version. Let that sink in.

[@koba_Lyle]

Looks a bit like our current Chess champion don't he? Lol

[@richardlee1721]

This translation. Sigh

[@jennyxie5382]

What about this translation?

[@manyuerainyo0o0]

ke jie already is a god, that alphago is cheating god

[@undeniableluck3260]

1:16 exquise me

[@meghakaushal4780]

Go cannot be solved by brute force like chess or checkers. Go involves a lot of creativity and intuition. Chess Super Grandmasters already know that they cannot stand anywhere near the AI but up until just a few years ago there was no engine or AI even close to defeating elite Go players. Hence it was not expected for the AI to humiliate Lee Sedol and Ke Jie so easily. But now we know that AlphaGo is capable of doing that.

[@bruceli9094]

Beaten by an iPhone lol

[@user-ev1jg3dd7z]

Being arrogant to complement the Al so exaggerating

[@Justme-to6yu]

Social credit go: weeeeeeeeeee

[@sam6007]

all i hear is “sha sha da sha a ga duh sha sha guhh shu duh” thank god for subtitles

[@aaronlin3823]

Well you should definitely go check your ears with a doctor.

[@sam6007]

@@aaronlin3823 dooly noted

[@aaronlin3823]

@@sam6007 duh

[@MundaneMuser]

LOL

[@eqh1593]

"I knew from the very beginning i could not beat the computer" Literally said "I think i will beat it 5-0 before the first match"

[@NO1xANIMExFAN]

no, that was lee sedol who said that when he versed alphago. lee played 5 matches with alphago this is ke jie, who played against alphago master, the next generation of alphago. they only played 3 matches. you not only confused the two players but also confused the two versions of alphgo.

[@oliverludwig6148]

@@NO1xANIMExFAN however i also think, i've read somewhere, that at some point he claimed to be confident of beating the A.I., and before refused to play, out of fear, that it would copy his style.

[@asianamericancasestudies6434]

@@oliverludwig6148 He made the statement after watching Lee Sedol played AlphaGo, and he explained the play why he would win whereas Lee Sedol made many mistakes. But then the AI also advanced.

[@dannygjk]

@@asianamericancasestudies6434 and even that AlphaGo was weak compared to the later version.

[@irrelevantideology9640]

What if....they paid him to lose?

[@Aziz0938]

Just NO

[@thePronto]

That would be self defeating. The system learns from its mistakes, which is how it got to this level of exprtise. Ditto the developers. They probably had a bonus in the contract if he won a game...just like bug or security exploit bounties.

[@asianamericancasestudies6434]

And what's the point of that?

[@SebastianBaltar]

now everyone can download an ai way stronger than alphago called katago and play them everyday on their pc, no one has beat it unless they make it play with 0 play outs or smth

[@fisophia1734]

Lol, thats not God, machine is machine, God of God.thats stupid try to fight math with calculator machine thats going scary or claim her is God.

[@rgaha5676]

Why can some people be so stupid enough to make someone compete against a machine in term of memory. Isn't AlphaGo just a program which memories all the previous playstyle and strategies of Go history. Its just using whats best suited going through it's memory. We human can't compare to machine in this field. Shame Google if they feel proud about doing this.

[@ericqiu6174]

not exactly

[@user-ni7hm3ry8o]

@@ericqiu6174 Human vs Calculator

[@gabrielgonzales5907]

That's how it was for the previous version, but with Alphazero there was no memorizing of previous strategies. They simply told the AI the rules and objective of Go and had it play itself thousands and thousands of times. In the case of chess, they told the same AI the objective and rules of chess (no memorizing any people's styles or strategies) and after playing itself for 4 hours it pretty much mastered the game more than any human.

[@gdx3phos434]

Gabriel Gonzales Would you bring a car in a race against Usain Bolt? Calculator in a math test? Submarine in a diving competition? A search engine to Jeopardy? You’re bringing something that doesn’t need to care about any other parameters such as stress, peer pressure, pride into a field that is meant for people only.

[@lucifer2133]

Fair point, but sooner or later there will be no field in which it will be "fair" for humans to compete against machines, and this by definition is the moment when we've become redundant.

[@marisasalbego1321]

GO takes way too long

[@thePronto]

Unless you play the 9x9 game.

[@finaldestinationr101]

The old saying goes" A student is only as good as his Master" AI is not and will not be smarter then humans period. It's only what humans put into it that comes out.

[@useewhynot]

Here's the brilliant thing though: we never put anything into the AI. AI doesn't learn by us telling it what to do and it doing it, AI learns by being given the rules of the game, then being allowed to explore the game for its own equivalent of lifetimes. Its like being born as a newborn and playing the game of Go nonstop until you're thousands of years old. Its incredible how by simply being given a specific set of rules, these systems are capable of inventing strategies beyond anything a human has come up with, and that's what makes me hopeful for the future; maybe we as humans can't solve the giant problems ahead of us, but AI can.

[@real_dddf]

青出于蓝而胜于蓝。The student is taught by the master, but inevitably surpasses. This is how technology and the world has advanced to what we see today. My teacher teaches me addition. I go on to discover multiplication. My student then figures out calculus. AI is that exceptionally bright student that can do wonders if guided in the right direction.

[@metsnanatdemeke9965]

@@useewhynot much respect to you sir!

[@dannygjk]

Faulty reasoning.

[@Tortuosit]

Haha, those Asians. Basically what he's trying is face-saving. No need to do that. But they do that even more than when westerners started losing chess. Just a cultural thing. After all, computers do always win Go from now on.

[@aaronlin3823]

Haha, those westerners... eh?

[@malharjajoo7393]

omg, so much hype around AI ... kinda BS

[@kittyscratch580]

How is it B.S.? To be honest A.I. beat chess before the year 2000. Go has more combinations so it took to the year 2020. We currently see a Chatbot cannot speak proficiently towards a human. Why? It has more combinations... It's only a matter of time until A.I. will supersede us in that as well.

[@dannygjk]

@@kittyscratch580 Yes and now we have ChatGPT-4 which is superior to most human experts in multiple fields.