What would you do if you learned true magic?
Fang Yu, a sophomore in college, never expected that he would actually acquire a mage tower!
This is an urban magic "face-slapping" ...
Chapter 218: Man vs. Machine (I)
A conference room on the sixth floor of the Four Seasons Hotel in Jongno District was set up as a temporary media reception room. Demis Hassabis and David Silva, representatives of DeepMind, were standing at the tea break area at the door, looking at the cameras of the reporters on the scene, their faces a little stiff.
The reporters on the scene were mainly Korean reporters, but there were also a considerable number of Dazhou media and Youmai media.
Before the press conference officially started, reporters were whispering, discussing in low but intense voices whether the artificial intelligence called AlphaGo could possibly defeat Lee Se-sil.
Soon the staff came to urge them, saying that the time had come and Lee Se-sil had arrived at the scene and would be present with representatives from Deepmind and the Korea Go Institute.
The Korean reporters' questions were mostly focused on Lee Se-sil, and the fanatical national sentiment almost scared Hassabis to the ground.
When Hassabis was asked about the chances of AlphaGo defeating the world's number one player, Lee Se-ji, he had to keep a low profile for fear of being beaten.
"It's 50-50." Hassabis replied briefly.
Isn't Lee Se-sil ranked fourth in the world right now? How did she become number one?
Obviously, this answer did not satisfy the reporters on the scene. Whether Lee Se-sil would win or lose was secondary to them. What they wanted to hear more was the flattery of the Koreans by these tall-nosed and deep-eyed foreigners.
What about Goryeo being number one in the world at Go? I really like Goryeo, kimchi is the most delicious food in the world, Koreans are very polite and well-mannered, and Goryeo deserves to be in the World Cup semi-finals...
These are what the Koreans are willing to hear.
After the press conference, both parties entered the preparation stage. Fortunately, Google is a Malaysian company, which naturally has a superior position in Korea, thus avoiding a lot of trouble.
Finally, after Lee Se-sil had lunch and smoked half a pack of cigarettes, at one o'clock in the afternoon local time, the two sides finally arrived at the game room on the seventh floor of the Four Seasons Hotel.
Hassabis, David Silva and Huang Shijie have arrived for a long time and are doing the final check on alphaGo.
"Lee Se-sil has five matches in March, so it's normal to ask for them to be played in February. It just leaves us with much less time to practice."
"If we hadn't adopted the new training structure this time, we might not have been able to complete the last training session. In that case, defeating Lee Se-sil would have been somewhat risky."
Hassabis picked up the bottle of water and took a sip.
"The applicability and efficiency of the youzi0.25b framework are amazing. If I hadn't used youzi0.25b in my last training session, I might not have been able to complete V13 training until now. If I hadn't completed it, this competition would definitely have been postponed."
David Silva nodded in agreement. As the project leader of AlphaGo, no one was more nervous than him about whether this project would be successful.
"It's a shame their baseline architecture isn't fully open source, so they didn't choose a license," Hassabis said, shaking his head regretfully. "But it's also understandable. It seems like this company is just getting started, so it makes sense to build some technological moats. Just like Hinton, who didn't open source AlexNet back then, otherwise he wouldn't have gotten the $40 million from Google."
The Hinton Hassabis was referring to is Geoffrey Hinton, who is now in his seventies and is a recognized leader in deep learning. He pioneered the concept of deep learning through the backpropagation algorithm thirty years ago.
Four years ago, Hinton released the groundbreaking AlexNet eight-layer neural network model. This model uses the CNN deep neural network and GPU-accelerated deep learning to significantly improve the computer's visual image recognition capabilities.
It was thanks to AlexNet that artificial intelligence first had real commercial value.
Hinton also became famous because of this, officially broke out of the circle, and was sought after by major companies.
At that time, in order to maximize profits, Hinton did something out of the ordinary. He and two students set up a shell company. This company had nothing but the three of them.
Then Hinton organized an acquisition tender and invited major companies to sign up for auction-style bidding. Whoever bid the highest would be acquired and they would work for that company.
At that time, Qianxun was played a trick. After several rounds of bidding, she always offered the highest bid, but the other party refused to recognize her and still chose Google.
I'm just using you as a backup, what can you do?
Deepmind also participated in the bidding at that time. Deepmind offered equity instead of cash, so Deepmind was eliminated in the first round of bidding. Therefore, Hassabis has always been dissatisfied with Hinton.
As a result, the two of them now work in the same company. Isn’t that a coincidence?
"I heard that Google is raising the price of EOS Technology. Currently, their EOS framework project has over 70 branches and over 300 contributors."
"Their official website updates examples almost every day. The framework has had five major updates and 11 minor updates in three months. The pace is very fast, but they are a big company, so Google has some concerns."
Hassabis chuckled.
"This company is indeed quite mysterious. No one knows who Professor Jorz is, and no one knows if it's a stunt inspired by Satoshi Nakamoto. However, if they hadn't kept it secret, Jorz might have been poached by a big company long ago." Huang Shijie shook his head.
Hassabis' eyes flashed slightly. He had just sent an email to Professor Jorz a few days ago, promising that if Professor Jorz was willing to join Deepmind, in addition to giving corresponding shares, he would also be willing to let Professor Jorz be the chief technology officer of Deepmind.
Of course, this kind of thing cannot be mentioned in front of David.
At this time, two staff members came forward to remind them that the game time was about to end.
Hassabis smiled and put his arms around Huang Shijie and David Silva's shoulders: "Let's go, let's make history!"
——
Although it is a man-machine game, a chessboard is still placed on the scene, and next to the chessboard is a computer. Huang Shijie will place the chess pieces on the corresponding points according to the instructions of alphaGo and play against Li Shishi on the chessboard!
"It's an honor to play chess with you." Li Shishi politely shook hands with Huang Shijie and sat on one side of the chessboard.
After the referee introduced the rules, Lee Se-sil guessed that black would go first, picked up a black piece and placed it lightly in the upper right corner of the board.
He looks very relaxed.
At the same time, Huang Shijie was also entering Li Shishi's first move into the computer.
AlphaGo's moves were not very fast, and it took about a minute to consider each move, while Lee Se-sil made his moves very quickly without thinking.
"The combination of moves 1, 3, 5, and 7 doesn't happen in real life. Lee Se-sil probably wanted to test the AI's logic, but his advantage as the first mover was essentially lost."
Da Zhou was also broadcasting this game live, and the commentator was none other than the true number one player of the time, Gu Jie.
"The AI was not affected at all. It separated the two upper black pieces naturally and responded normally. Li Shishi had a bad start. However, for Li Shishi, it should not be a problem to make up for this small loss at the beginning from the AI in the middle game." Gu Jie looked a little gloating and seemed very relaxed.
But as the game progressed, both Lee Se-sil on the scene and Gu Jie, the commentator, were dumbfounded by alphaGo's performance.
"The combination of 24 and 26 moves is very powerful. AlphaGo seems to be much stronger than we thought." Men and women working together make the work more enjoyable, and the live broadcast platform also assigned Gu Jie a female commentator.
In fact, they were also afraid that Gu Jie would talk nonsense on the spot, so they gave him insurance.
"That's because Li Shishi was too hesitant in his response!" As expected, Gu Jie immediately expressed his dissatisfaction with Li Shishi's play. "If he keeps playing like that, he's bound to lose. It's not because AlphaGo is so strong, it's because Li Shishi's play is so bad!"
The progress of the whole game seemed to be just as Gu Jie said. After being shocked by AlphaGo's chess skills at the beginning, Lee Se-sil did seem a little nervous and frequently fell into long thoughts.
By the 67th move, he even started waving the chess piece around on the board.
Even the KBS commentator exclaimed: "Li Jiudan seems to be doubting himself and is a little scared."
"The winning rate has risen to 73%." David Silva was keeping track of alphaGo's backstage with a pad in hand, and quietly reported the winning rate calculated by alphaGo based on each move.
At this moment, alphaGo seemed to have made a wrong judgment in the lower left corner. In such a good situation, it actually chose 88 Tiger!
In this way, the black chess piece can successfully occupy the lower left corner!
The situation is reversed! ?
The studios of more than a dozen TV stations and five or six live broadcast platforms were all filled with jubilation.
AI made a mistake! Turn defeat into victory!
Is this really the case?
"The winning rate is still 70%." David Silva remained calm. He understands Go, but he understands alphaGo even better.
Li Shishi was also very excited. In his opinion, White's move 88 was undoubtedly a stupid move. It was obvious that White could have taken advantage of the victory to kill him, so why didn't he do it?
Li Shishi snapped his hands twice and accepted the real estate in the lower left corner.
Immediately afterwards, the AI seemed to have made another wrong judgment, giving two pieces of territory to Li Shishi.
Are we really going to turn defeat into victory?
After Lee Se-sil confidently broke the black chess piece in move 101, AlphaGo made a brilliant move almost without thinking!
White, play 102.
What! ?
Not only Lee Se-sil, almost all the commentators in the studio exclaimed in surprise.
"Orange!?" Gu Jie blurted out.
Chen Ying, who was commentating nearby, was shocked: "Gu Jie, you said AlphaGo is Orange?"
Gu Jie pondered for a moment and said, "No, I just said that this move reminded me of Orange. AlphaGo is not yet as strong as Orange, and of course, it is not as strong as me either."
Chen Ying: “…”
"Orange?" Seeing this move, Li Shishi also raised his head and stared at Huang Shijie closely.
Huang Shijie was puzzled: "What oranges?"
They have not fed human chess records to AlphaGo since November. Instead, like the Orange model, they have adopted the method of self-playing to learn. They have basically not paid attention to the news in the Go industry in the past two months. Even if they did pay attention, they only paid attention to the world championships. An online chess player who emerged in the Great Zhou Dynasty was not enough to catch the attention of DeepMind.
This is what it means to be a different person.
Orange is already well-known in the Go world, but the artificial intelligence that specializes in Go doesn't even know about him.
"No, nothing. It's not Orange. AlphaGo's chess skills are not as strong as Orange." Li Shishi shook his head, but this scene had been faithfully recorded and broadcast by the cameras on the scene.
As soon as 102 is played, the outcome of the game is decided.
"Lee Se-sil is going to lose." Gu Jie was the first commentator to confirm this.
"Impossible? Black should have the upper hand now." Chen Ying was a little unsure.
Just as we reached the middle of the game, judging from the situation, Li Shishi did not seem to be at a disadvantage. Moreover, Li Shishi had such a big advantage just now, how could he possibly lose?
"The move just now created the only opening for the black piece on the right." Gu Jie arranged the subsequent pieces on the large chessboard. "Black can only choose to take a long move, and then..."
After Gu Jie made this move, the audience who knew a little about chess understood it.
Black is indeed going to lose.
The black chess piece in the lower corner, which originally seemed impregnable, was easily captured by the white chess piece with three pieces left, leaving no room for resistance.
The 88 tiger, which everyone previously thought was a stupid move, happened to save a small piece of white chess real estate.
But just this little piece is enough!
No matter how you calculate it, by the final stage, Li Shishi will lose at least three points.
Three points may not seem like a lot, but for a game of chess, a difference of half a point means the difference between winning and losing!
"AlphaGo's control of the situation really reminds me of Orange. They both have one thing in common: they judge their moves based not on how good the situation could be, but on how controllable it is."
"Of course, I still think 88 Tiger wasn't a good move, but it's clear that AlphaGo believed the situation was under control and securing victory was paramount, so they gave up the lower left position. Orange also made this choice in our first game, but I won that time."
The comments below the live broadcast room were so loud that you couldn’t even see the screen. They were all telling me, Jiebao, you lost the last two games.
Gu Jie didn't notice the comments and continued, "Judging from this game, AlphaGo's skill is roughly the same as when Orange first played against me, but not as good as mine. Compared to the last two games between Orange and me, the gap is even greater."
"So I'm wondering..." Gu Jie touched his chin, "Is the orange that defeated me in the last two games also an AI?"
"Besides, Yicheng is the website of the Great Zhou Dynasty. Orange appears on Yicheng and is named Zhou Wenming. Maybe it's an artificial intelligence developed by a company in our country?" Gu Jie said excitedly. "If that's the case, I'm still number one in the human race, hahahaha."
"But, if Orange is an AI, then even you, the current world number one, lost to him, and you even admitted that you couldn't beat Orange. Doesn't that mean that humans have already lost to AI? Doesn't that mean that Go is dead?" Chen Ying blinked her big eyes.
"Ah!" The smile on Gu Jie's face froze suddenly.
The barrage of comments below became more frequent, and people who knew chess and those who didn’t were all posting barrages and comments.
"It's possible. I've seen Orange's chess records, and they're not something a human can play!"
"Jiebao is right. Our Great Zhou's artificial intelligence is no worse than Mai's!"
"Zhou Wei is awesome! It must be Zhou Wei!"
"The person above doesn't understand Go or AI. Zhou Wei doesn't work on AI. The only AI in China that can beat him is Qianxun."
"If Orange is AI, it must be from Miami. China doesn't have this technology, and even ballpoint pen refills have to be imported."
"That's all nonsense. Orange is a real human being. Li Shishi lost, which only means that he played poorly. Do ordinary people understand that masters are better than ordinary people?"
——
"Snap!" Huang Shijie placed the 186th move on the chessboard according to the feedback from alphaGo on the screen.
Li Shishi picked up two black pieces from the chess box and tapped the box gently.
"Snap"
"Snap"
It's as if time has stood still.
Then, Li Shishi gently placed the two chess pieces on the chessboard.
"I lost."
As the Korean referee announced that AlphaGo had won the first game, Huang Shijie, Hassabis and other members of the DeepMind team cheered together.
AI defeated the top human chess player for the first time!
"I'm very sorry, I would like to ask, have you ever let alphaGo play Go on the Go website?" Li Shishi was not as depressed as imagined. He stood up and shook hands with Huang Shijie and Hassabis.
Hassabis shook his head. "Until now, apart from our team members, AlphaGo has only played against two humans, you and Fan Hui."
"So it's not you." Li Shishi showed a hint of disappointment on his face.
???
What's the meaning?
"AlphaGo's strategy for several moves is very similar to that of a chess player I met online. Oh, by the way, his name is Orange. No one knows who he is. Gu Jie and I both lost to him."
Li Shishi said calmly.
"Orange?" Hassabis frowned, but didn't care.
In this world, no matter how strong a human chess player is, he can never be a match for alphaGo!
(End of this chapter)