Former Go champion beaten by DeepMind retires after declaring AI invincible

Lee Se-dol is seen in 2016 during his matches with the AI program AlphaGo.
Photo: Google / Getty Images

The South Korean Go champion Lee Se-dol has retired from professional play, telling Yonhap news agency that his decision was motivated by the ascendancy of AI.

“With the debut of AI in Go games, I’ve realized that I’m not at the top even if I become the number one through frantic efforts,” Lee told Yonhap. “Even if I become the number one, there is an entity that cannot be defeated.”

For years, Go was considered beyond the reach of even the most sophisticated computer programs. The ancient board game is famously complex, with more possible configurations for pieces than atoms in the observable universe.

This reputation took a knock in 2016 when the Google-owned artificial intelligence company DeepMind shocked the world by defeating Se-dol four matches to one with its AlphaGo AI system. The games had a global impact, alerting the world to a new breed of machine learning programs that promised to be smarter and more creative than AI of old.

Lee, who was the world’s number one ranked Go player in the late 2000s, initially predicted that he would beat AlphaGo in a “landslide” and was shocked by his losses, going so far as to apologize to the South Korean public. “I failed,” he said after the tournament. “I feel sorry that the match is over and it ended like this. I wanted it to end well.”

Despite the outcome, Go experts agreed that the tournament produced outstanding play. AlphaGo surprised the world with its so-called “move 37,” which human experts initially thought was a mistake, but which proved decisive in game two. Lee made his own impact with his “hand of God” play (move 78), which flummoxed the AI program and allowed Lee to win a single game. He remains the only human to ever defeat AlphaGo in tournament settings. (During training AlphaGo lost two time-capped games to Go player Fan Hui.)

Since the tournament, though, DeepMind has only improved its AI Go systems. In 2017, it created AlphaGo Zero, a version of the program which surpassed even AlphaGo.

While the original AI learned to play Go by studying a dataset of more than 100,000 human games, AlphaGo Zero developed its skills by simply playing itself, over and over. After three days of self-play using hugely powerful computer systems that let it play games at superhuman speeds, AlphaGo Zero was able to defeat its predecessor 100 games to nil. DeepMind said at the time that AlphaGo Zero was likely the strongest Go player in history.

In a statement given to The Verge, DeepMind’s CEO Demis Hassabis said Lee had demonstrated “true warrior spirit” in his games with AlphaGo. Said Hassabis: “On behalf of the whole AlphaGo team at DeepMind, I’d like to congratulate Lee Se-dol for his legendary decade at the top of the game, and wish him the very best for the future ... I know Lee will be remembered as one of the greatest Go players of his generation”

According to Yonhap, Lee isn’t completely giving up on playing AI, though. He plans to commemorate his retirement in December by playing a match against a South Korean AI program called HanDol, which has already beaten the country’s top five players. Lee will be given a two-stone advantage.

“Even with a two-stone advantage, I feel like I will lose the first game to HanDol,” Lee told Yonhap. “These days, I don’t follow Go news. I wanted to play comfortably against HanDol as I have already retired, though I will do my best.”

Update: Comment from DeepMind CEO Demis Hassabis has been added to the story as well as clarification about players who have beaten AlphaGo in different settings.


Sorry to see him Go

We are all GOing to lose anyways.


Take your rec and get out.

This….this comment is what I am thankful for this Thanksgiving! Frankly!

I understand why you would do this. At the same time, people didn’t give up weightlifting because machines are able to move more than we can. It’s just new that mental challenges can be defeated by technology as well as physical challenges. After the initial shock, I don’t expect many will follow suit by leaving the sport.

For sure. Lots of people still play chess right?

Stockfish defeats them all, but still loses against DeepMind’s AI.

I think it’s different for things that are perceived as tests of intelligence. Animals have always been stronger than the strongest humans, and even simple mechanisms like levers and pulleys that have existed for millennia demonstrate that machines have advantages over humans for physical tasks. People can still compete with for the trophy of strongest human, sure.

But humans have always maintained (or perceived, at least) superiority over all other things in existence in abstract intelligence, and we’ve used games like Go and chess to represent that. The fact that the world’s best human Go player is the best entity in existence at that game probably means something to the players who strive to achieve at its highest level.

I imagine that the idea that no matter how much you try or improve, the limitations of your physical neurology mean that you can never again credibly compete at the highest levels of this test of skill is probably a bit of an existential crisis for someone who has dedicated so much of their life’s work, and gained so much of their social prestige from it.

For him – definitely. But a new generation of players is going to grow up acknowledging AI’s superiority in the game and being motivated to play nonetheless, same as what happened for chess. This doesn’t mean much for the game overall other than maybe the development of some new techniques and strategies.

paxys, this has tremendous ramifications for Go. Just as it had tremendous ramifications for chess. Besides throwing out a lot theory based on human games, players will study alphago’s games instead of human games. Players will train by playing computers instead of playing human players. Today most analysis of chess games is done by the computer. And in the future so will Go games.

Kind of different, strength has never been compared to other forms because man was never the strongest in the known universe.. let alone our own planet – so it was always a competition between ourselves.

As far as the mind, humans have been considered the most intelligent of the known universe – so we welcomed all challengers and now that day has come to an end – - which is shocking.

AI will continue on its exponential trendline, and is expected to reach human-level general intelligence in just 9 short years. Super-intelligence follows shortly afterwards. The Singularity is Nearer, but most people are oblivious.

9 years? Does that mean Alexa will actually listen to me then?

That is Going to be a challenge even Amazon may not be able to succeed at.

The short time frame from AI scare-ers always blows my mind. I lose cell service over a hill, self-driving cars are slipping ever further away, and we still don’t have good battery tech. If an AI can play Super Mario though, we are doomed.

generally these people don’t actually program and have no idea how these AI systems actually operate. our AI isn’t even smarter than an infant

BenderIsGr34t, Not smarter than an infant, really? I can ask google on my smartphone any question under the sun. Google will give me an answer. Can an infant do that? People don’t even realize they are using AI all the time with Google. LouisPiper, And by the way Waymo does have self driving cars. They ordered 80,000 cars. They just need to get up to scale.

Exactly. This link mentions it but I don’t like that site or its lack of sourcing, but I don’t have time to find a proper source right now.
This Wired link is a nicer read.

Basically, these AI assistants have the intelligence of a 4-6 year old when it comes to general intelligence like common sense.

Wait and see what happens when that 4-6 year old AI turns 7.
It’ll be like you cranked it up to 11.

Not smarter than an infant, really? I can ask google on my smartphone any question under the sun. Google will give me an answer

You need to understand what the measure actually is before you can say with any confidence which entity is better at it.

If the measure is things like general intelligence, reasoning, or problem solving, Google is literally nowhere. However if its matching search results in a database…

My calculator can add way larger numbers than an infant (or any other human) is capable of. That doesn’t make it smart or intelligent.

nos, Every time a computer does a thing that requires intelligence, you say it’s not intelligence, it’s just a computer following an algorithm. But before the Go champion defeat, people did say beating a Go champion requires lots of intelligence. This is the AI effect. Which means AI are things the computer has not done yet. Once a computer does it, it’s not AI. What is your definition of intelligence?

scoob101, what you are really saying is only the human brain is intelligent. A computer can’t be intelligent because it’s not a human brain.

paxys, Then your calculator is not very intelligent. Because it is restricted to a narrow domain. It has intelligence of an elementary school student in the area of arithmetic. But not a lot of intelligence.

There are 2 reasons for the AI effect. A computer can’t be thinking because a human doesn’t think that way. Or because a computer can’t be intelligent because then a computer is more intelligent than a human. That’s not possible.

Something new from DeepMind, the creators of AlphaGo. MuZero learned to play chess, Go and Atari games by watching. Without being given the rules. One general learning algorithm learned to play many different games. Then play better than the strongest human player. More signs of high levels of artificial intelligence.

View All Comments
Back to top ↑