[Kim's e-Column] presents IT issues that become topics of discussion online, unraveled through Kim's curiosity. We will explore IT issues with keen and sparkling eyes!

First Match Begins on March 9!
A Battle Between Human and Computer for Approximately 1.2 Billion Won in Prize Money



Last year, Google's 'AlphaGo' played against European Champion Fan Hui, 2-dan, and secured five consecutive victories. This caused a global sensation, though compared to IBM's 'Deep Blue' defeating a chess champion in 1997, it came at a slower pace. However, Go, with its 10 to the power of 170 possible moves, was expected to be difficult to win easily.

Google has already issued a challenge to South Korea's Lee Sedol, 9-dan, and plans to hold a total of five matches, identical to those with Fan Hui. The century's matchup, beginning next month, will take place on March 9, 10, 11, 13, and 15, and will be broadcast live on YouTube. The prize money is set at 1 million dollars (approximately 1.2 billion won), and Google has announced that if AlphaGo wins, it will donate the entire prize amount to charitable organizations.

As befits this century's matchup, predictions about the outcome are also heating up online. As opinions on the result of the human versus artificial intelligence battle vary, we examined diverse public discourse.

First, most netizens on domestic portal sites predict and hope for Lee Sedol's victory. A user ID 'goldwing' said, "This is funny. A computer beats a professional Go player? For Lee Sedol, it will be five losses," while 'euni****' left a comment saying, "I hope Master Lee wins."

Netizens Say "Lee Sedol Will Obviously Win, but It's Surprising and Frightening"
to "It May Be Difficult Now, but It Will Soon Surpass Human Ability"—There Are Also Minority Opinions

On the other hand, there were many comments expressing fear about the advancement of artificial intelligence. User ID 'dtb****' wrote, "This is a serious problem. We are entering an era where computers will dominate humans," while 'pens****' expressed surprise, saying, "I can't believe it has already reached this level. In 10 years, computers will be competing against each other for the world Go championship."

Though the recommendation counts were not high, considerable numbers of opinions from those who sensed AlphaGo's victory also appeared. User ID 'nani****' said, "AlphaGo is not merely a machine that calculates possible moves through algorithms," and explained, "It is an artificial intelligence that mimics the human cerebral cortex, using parallel processing devices to simulate the electrical action of neuronal synapses, acquiring in a day or two experiences that a human could never obtain in a lifetime." Continuing, "There is absolutely no possibility of Lee Sedol winning," the user stated. Additionally, 'gold****' warned Lee Sedol not to underestimate AlphaGo at all, saying, "Right now a professional Go player will win, but like chess, it will achieve a complete victory within five years."

With Go as a Powerful Nation—Korea, China, and Japan—Interest in Go Is Hotter Than Ever
Lee Sedol Victory Predicted, but Some Comments Say "We Don't Know the Result Yet"

Interest in this matchup appears to be significant not only in Korea but also in China and Japan. A user ID 'hxu2[Shanghai]' who left comments on the Chinese site Sina.com asserted, "Google's computer lacks the ability to think; no matter what, it will never be able to beat humans," while a contrasting opinion came from user ID '光慧[Chengdu, Sichuan],' who revealed they work in the artificial intelligence field, saying, "It's interesting how many people seem to think AlphaGo only played against Fan Hui. Those who make artificial intelligence conduct extremely extensive testing before launching an AI. This program has probably faced many skilled players on internet Go sites in Korea, China, and Japan," speculating that "the AI has probably undergone extensive training on the internet."



<Demis Hassabis, co-founder and chief executive officer (CEO) of Google DeepMind, a subsidiary of Google that developed 'AlphaGo,' held a press briefing on the 28th at Google Korea in Yeoksam-dong, Seoul, via video conference system from London, England. Provided by Google Korea>

In fact, DeepMind CEO and Google Vice President of Engineering Demis Hassabis stated during a video press conference on January 28, "'AlphaGo' has played approximately 1 million Go games so far. If we assume that a typical Go player plays about 1,000 games per year, AlphaGo has received training equivalent to nearly 1,000 years in a single session."

Other opinions included user ID 'WifitCSA' saying, "I hope the Go powerhouses of Korea, China, and Japan would conduct joint research for the advancement of artificial intelligence."

Japanese netizens also left their own comments, expressing anticipation for the match. User ID 'EHa1Oewk' on a Japanese 2-channel Go thread wrote, "Sedol will win decisively and people will say how great Sedol is, and that will be the end of it, but I'm looking forward to it," while user ID 'ZN3lpHC3' stated they saw the game record provided by Nature magazine, saying, "Even with Lee Sedol at 9-dan, it seems risky. I personally love Sedol's Go and hope he wins. But it does seem risky for Sedol."

Experts: 'This Time It Won't Defeat Lee Sedol'
However, 'It Is Much Superior to Existing Go Programs'

Experts are also making their own match predictions. On January 28, Professor Kim Dae-shik of KAIST stated to Hankyung.com, "Lee Sedol, 9-dan, will obviously win," adding, "Google probably doesn't think AlphaGo will win either." Additionally, he explained, "Rather, for Google, it would be advantageous for AlphaGo to lose and obtain data from the highest-level player."

Game Records of AlphaGo versus Fan Hui <Provided by Nature & Google DeepMind>

Yang Jae-ho, Secretary General of the Korean Baduk Association, also predicted, "Lee Sedol, 9-dan, has never played against artificial intelligence computers, so he may win one or two games, but won't be able to win all matches," while Lee Ha-jin, Secretary General of the International Go Federation, stated, "After looking at the game records, I didn't get the feeling that a computer played," adding, "Existing Go programs place stones through calculation, so if a miscalculated value occurs, they place stones in strange locations, but AlphaGo doesn't seem to do that."

AlphaGo connects 1,202 CPUs to reduce the number of possible outcomes, inputs 30 million games worth of big data, and then AlphaGo itself conducts games to find the best moves and accumulate data. Additionally, it can recognize images, so in situations without data, it eliminates all possible scenarios related to that situation.

After 'Deep Blue' won in chess in 1997, it was speculated that Go would be more difficult to win because it had far more possible moves. Now, this may seem like a foolish thought. Indeed, only computers could handle calculations with such vast numbers of possibilities, could they not?

Now, only the readers' choice remains. Will it be a human victory or a machine victory? Or should the question be changed to 'Whose victory are you hoping for?'