"Dalula?" Hasabis was shocked to see Dalula, perhaps no less than Columbus' first discovery of the New World.

Dalula smiled slightly and said gracefully: "Hello, Demis, no see for a long time!"

"Why did you come?" Hasabis still looked puzzled.

Dalula said: "I will formally introduce myself. Dalula Riley, the head of the Silicon Valley R Incubator, is here to negotiate the acquisition with your company this time."

Hasabis was stunned: "So, Elon and Thiel know about this?"

"Elon and I have been divorced for a long time. R Incubator is a venture capital agency I run independently." Dalula said lightly, "Before coming here, Elon and I did not communicate, but I believe he will be soon

You will know."

"Well, I probably shouldn't ask too much." Hasabis shrugged, "No matter what, you're welcome. The last time we met, it seemed that we were at Thiel's house..."

"Yes, that's been three years ago." Dalula said, "I'm glad to see you getting closer and closer to your dream!"

With the dual relationship between Hinton and Dalula, the atmosphere was very harmonious after the meeting between the two parties.

After visiting the company, everyone returned to the conference room.

DEEPMIND is not big, with only more than 30 employees in total, but the entire company is in a very vigorous state. Although it is located in the old-fashioned London, it presents a Silicon Valley-like atmosphere everywhere.

When he arrived at the conference room, Hasabis prepared coffee for everyone himself.

Hinton could not sit down, so he could only stand in the corner of the wall, pressing his back tightly against the wall. He pointed to the chess board hanging on the opposite wall and said, "Demis, are you still playing chess?"

Hasabis smiled and said, "I just occasionally go to the online battle platform to find a game room next game when I encounter difficult problems, and hope to relax in this way."

"That's a pity. If you insist on playing chess, maybe we will see you in the World Chess King Competition." Hinton said.

Dalula also said: "I still remember that the last time we met at Til's villa on Baker Street, you and Til didn't talk about artificial intelligence at all, they were talking about chess from beginning to end, your point of view made me

Refreshingly, you said that the reason why chess is full of competition is because of the tension between horses and elephants about occupying the black and white grid. At that time, I was thinking, Oh my God, you are a truly smart genius!”

Hasabis smiled and said, "Yes, Thiel later admitted that the reason he was willing to invest in DEEPMIND was not because he was optimistic about artificial intelligence. In fact, he didn't ask what the artificial intelligence we developed could do. He just thought

I'm very interesting and willing to invest no matter what I do."

Dalula smiled and said, "Later, Til and Elon recommended you this way. He said that Demis is a very smart person and can succeed no matter what industry he works in. You will definitely not be wrong to invest in you. But I

I still feel it's a pity that you could have become a world-famous chess player."

Hasabis shook his head and said, "Artificial intelligence has conquered chess. Even if I can get the chess king, I am still no match for artificial intelligence. So my goal has changed a long time ago. I hope to give humans through artificial intelligence.

Provide more powerful logical computing capabilities and conquer all the games in the world! We have made great progress in this regard!”

Hinton said in surprise: "You have conquered "Bracks"?"

Hasabis smiled and said, "That's right!"

"Display it to me!" Hinton was quite excited.

Since the concept of artificial intelligence first formed in the 1950s, games have always been an important test method for artificial intelligence.

As early as 1990, artificial intelligence defeated the human world champion in the checkers event.

In 1997, IBM's supercomputer "Deep Blue" defeated world chess champion Kasparov.

In 2011, another IBM computer, Watson, defeated all human players in the game "Edge of Danger".

DEEPMIND also tried to tap the potential of artificial intelligence from game projects. They developed an artificial intelligence that can play games, and their main goal was to Atari games.

Atari is a popular console in the 1980s. It was once the overlord of the game market, but it was slumped due to the launch of the most trasty game in history, "ET", and was eventually replaced by Nintendo's FC.ωωw..net

But in the hearts of many European and Magnesium people, Atari games are still classic. Among them, the most famous games, "Space Invader", "Table Pong" and "Bricks" were included in the AI ​​conquering list by DEEPMIND.

When I came to the demonstration computer, there was a game "Bracks" on the screen. No one controlled it, and the baffle automatically wandered under the screen, bounced the ball towards the brick wall upward again and again.

Hassabis said: "Vlad, come and explain it to you."

Vlad Mney is a senior engineer at DEEPMIND. This artificial intelligence system that is being made by him leading the team. He nodded and said, "Everyone, we use a brand new deep neural network.

The construction method was created. Through repeated trial and error, we learned the nuances of "Bricks" game method and each hitting angle. We have played 14,283 games, and the system has been able to analyze which one.

The action is effective hit, which one is invalid hit? We call this learning method reinforcement learning."

"In the initial system, it took more than 300 hours to master the game and pass the level. After we continuously adjust the algorithm and structure, it is now able to learn the basic rules of the game in 30 minutes, 1 hour

Achieving the proficiency of ordinary players, they can accurately receive a small ball hit a brick every time. After 2 hours, it has learned advanced skills, can find the right angle and hit the ball into the narrow space behind the brick, and use the small ball

The rebound collision before the bricks scored quickly.”

"At present, the score of this system in the game has surpassed everyone in DEEPMIND." Mney's last sentence is thought-provoking.

Hinton kept leaning against the wall and listening quietly until he asked: "So your system really learns the game skills of brick-making from scratch every time?"

Mney said: "That's right. We just set a goal, such as getting the highest score, which can automatically trial and error and learn, and develop some unique game strategies. The ultimate goal of these strategies is to get the highest score. Then,

It can succeed every time, but the time it takes is getting shorter and shorter."

"It's shocking..." Hinton murmured.
Chapter completed!