Before Meng Fanqi was reborn, under pressure from OpenAI's ChatGPT and Microsoft's Bing, the two major artificial intelligence organizations, the [Google Brain] team and the [DeepMind] team, merged.
After the merger, Demis served as the CEO of [Google DeepMind], and Jeff instead took a virtual position as the team's chief scientist.
As the founder of a small company that was acquired, Demis was able to surpass Google veteran Jeff in the post-merger position in artificial intelligence, a field that Google is very concerned about. This shows that Demis has some means and talents.
But that was ten years later. Now Demis has just got a new owner and wants to share the resources of the Google Brain team. He is still quite uneasy when facing Jeff.
Fortunately, Jeff is still a technical manager after all, and he also appreciates the achievements of the DeepMind team.
The communication between the two parties was smooth and pleasant.
But when talking about Meng Fanqi, a member of the Google Brain team, Jeff said that he could not make a decision.
"He is now a famous and privileged person. If you need his help and participation, you should contact him directly yourself."
This answer surprised Demis.
It is really unimaginable that Meng Fanqi has such a high status without even officially joining Google.
After DeepMind was acquired, in addition to the computing resources it just needed, the first thing it paid attention to was Meng Fanqi of Google Brain.
This is an idea advocated by Dr. Huang Shijie.
From his perspective, Meng Fanqi is probably the resource DeepMind needs most right now, an outstanding and creative brain.
And as both Chinese, Meng Fanqi is naturally better able to understand the cultural heritage of Go than European and American scholars.
Meng Fanqi was confused when he received the call from Dr. Huang, but fortunately, Dr. Huang acted as the spokesperson of AlphaGo in his previous life and replaced AlphaGo in playing chess with Li Shishi and Ke Jie.
As soon as Meng Fanqi connected to the video, although he looked a little different, he recognized him immediately.
After some introductions and pleasantries, Meng Fanqi went straight to the topic.
"Dr. Huang suddenly contacted me. Is there anything important?"
"It's not that important. The main thing is that we have a large-scale intelligent chess project. I wonder if you are interested in participating. The goal is the most difficult board chess game in human history, Go. We hope to make the artificial intelligence program in In Go, the most difficult chess game in the world, he truly defeated professional Go players."
Of course, Huang Shijie didn't want Meng Fanqi to join full-time and work for free. He knew that Meng Fanqi was at the peak of his technology and could release new methods as fast as laying an egg.
There is no reason or qualification to ask such a rising star to join the team and work for him.
All he asked for was for Meng Fanqi to participate in the discussion intermittently and provide them with some opinions and opinions.
Many of Meng Fanqi's previous thoughts have played an important role in the construction of AlphaGo's basic framework.
"I used your residual idea some time ago, which helped our performance a lot. Your generative confrontation idea also inspired us." Dr. Huang is not flattering. The residual idea basically helps everyone. Algorithms that require deep networks of a certain scale are of great help.
In addition, Meng Fanqi's work on optimizers, normalization and other aspects is also very refreshing to Dr. Huang.
It can be said that every time I read Meng Fanqi's articles, I have some new ideas and inspirations. Since both of them are Chinese, Dr. Huang really hopes that Meng Fanqi can participate in this project.
Regarding the matter of Go, after all, the Chinese can better understand the feelings in his heart.
But the only thing I'm worried about is that Go AI is not a direction with great commercial value or promising industrial implementation.
To put it bluntly, this thing doesn’t make much money.
The project is like a Pixiu, you can only see the money going in, but you can’t see the things coming out.
Meng Fanqi, as a rapidly rising star in the field of artificial intelligence, has made many breakthroughs, especially in detection-related algorithms.
The conditions for the implementation of many detection-related technologies are mature, and research in this area will definitely be fruitful.
In contrast, he may not be able to like this Go AI project, and even if he does, he may not have the time to participate.
However, Dr. Huang still made efforts and attempts to invite.
"It happened that DeepMind had just been acquired by Google, so I asked Jeff about you and wanted to see if there were any opportunities for cooperation. Jeff and Hinton said that although you are now a researcher at Google Brain, you enjoy a lot of With high autonomy, they won’t impose any restrictions on you, and they don’t know your schedule very well, so just let me contact you directly.”
When Dr. Huang said this, he was actually a little curious.
A position like a researcher does have a certain degree of autonomy, and you can make some choices and choices in research directions based on your own strengths and preferences.
But like Meng Fanqi, there are very few researchers whose two top leaders clearly expressed the attitude of "leave him alone" and "I can't control this guy."
"I mainly brought in a little bit of revenue for Google, just a few hundred million points." Meng Fanqi laughed, having discussed this matter with Hinton last time.
Ever since Meng Fanqi gave Google one-third of its advertising revenue, the privileges of the entire Google Brain department have been completely established.
Money, money, is the lifeblood of all directors.
Whoever can generate income so powerfully will enjoy unlimited privileges.
As soon as Meng Fanqi made money, the entire department followed him.
As a result, Google has completely established its corporate purpose of putting artificial intelligence first, and the status of Google Brain is naturally extremely transcendent.
And of course, the biggest contributor to all this, Meng Fanqi, can no longer be controlled.
People just didn't obey the orders, and that's why they made so much revenue.
What, you want to control him? How much revenue did you generate for Google?
"To tell you the truth, I'm actually quite interested in Go AI. If it's convenient, would you like to talk about the current progress today?"
Unlike Dr. Huang, Meng Fanqi is very good at chess, almost at the level of amateur 6-dan.
His level is basically that of someone who has just understood the basic rules of Go and has never played a good game with anyone.
I only watched a few basic tutorial videos on War Eagle and Ice Cream, but I still didn’t pay too much attention to them.
But this does not affect his interest in Go AI at all. He pays more attention to AI than Go.
Over the past year, Meng Fanqi has focused on accumulating original capital and monetizing the visual technology he is most familiar with, almost forgetting some basic pursuits in life.
Vision algorithms, especially this series of operations in detection, YOLO-multi-target-face, were all Meng Fanqi's comfort zone in his previous life.
The various techniques and source codes are so familiar that one can form an army by one person, and the entire set of functions will be developed without any lag.
These are the most mature technologies that have been realized in the past, and the processes and principles are not that troublesome.
A complex system like AlphaGo cannot be reproduced by Meng Fanqi alone.
There is a massive amount of chess record data, input methods, scoring standards, search and confrontation principles and screening strategies.
Meng Fanqi was basically not familiar with this part of the content. There were too few interviews for technologies that could not be monetized.
