Meng Fanqi was in a good mood when he walked out of the private club with Robin Li.
Robin Li is very optimistic and persistent about cutting-edge AI technology, and he is very good at talking about many conditions in this area.
Meng Fanqi felt quite good after coming into contact with him.
If Baidu does well overall in this tentative cooperation, Meng Fanqi will not be too lazy to continue to look for other partners in the country.
After all, at this point in time, the popularity of AI technology is far from high enough.
"Then let's make an appointment to meet again in a week or two." Li Yanhong shook hands with Meng Fanqi. "At that time, you have prepared the algorithm and model, and we have prepared the contract and other relevant regulations. After the matter is finalized, I hope there will be A real-time intelligent detection App, let’s see how the market responds to this technology.”
For the detection of content on pictures, a huge key point for qualitative change is real-time or real-time.
Not all images that need to be detected are prepared and preprocessed in advance.
The input of many scenes will be real-time shooting by cameras. In this case, the speed of the algorithm is a very big bottleneck.
Gamers should be very familiar with the concept of FPS. To put it bluntly, it is the number of frames per second and how many pictures are displayed per second.
The larger the number, the smoother and silkier it will naturally be. But at the same time, the number of images that the algorithm needs to process will also increase exponentially.
The current high-precision traditional algorithm can calculate a picture in one or two seconds, which is already very fast.
Generally speaking, the FPS must reach at least ten or twenty to truly be considered a smooth real-time detection.
If it can reach 60-70FPS, it will obviously have that extremely smooth and completely real-time processing feeling.
"In this period, to deal with the problem of slow detection algorithm speed, sampling is usually the method." Meng Fanqi recalled.
The principle of sampling is very simple. I can only process one picture per second, so I will ignore other frames and only extract one frame per second for detection.
However, this approach will cause the frame to chase people. By the time the detection frame is predicted, the objects in the frame have already moved outside the frame, giving people a very poor user experience.
If most of the objects in the video content are moving quickly, there will be empty frames flying all over the sky, which is pure visual pollution.
"No problem, I will keep my promise and will not submit a paper or announce technical details until at least 6 months later." This is also an important part of the conditions just mentioned. Baidu does not exclusively possess the technology.
But Meng Fanqi must wait until June 2014 before he is allowed to submit a manuscript or make the technology public. Meng Fanqi is completely acceptable at this time.
After Meng Fanqi shook hands with Mr. Li and said goodbye, he declined Robin Li's offer to send him a driver and left Baidu's campus alone.
Li Yanhong stayed where he was, looking at the thin pieces of paper in his hand, thoughtfully.
After returning to the headquarters, Robin Li lay on his boss's chair and continued to think for a long time.
Immediately, he made two important decisions.
First, Baidu wants to open an AI research laboratory in Silicon Valley overseas.
Meng Fanqi's reasons for rejecting him made him realize to a certain extent that the overall academic atmosphere and environment are also very important.
Baidu lost the sky-high auction with Google last year. Canada is too close to the United States, which is a very important factor.
If you have an excellent Silicon Valley AI research laboratory, then you will have much better competitiveness when you recruited Hinton's team last year and just hired Meng Fanqi.
At this point, my reaction was slow. Li Yanhong reflected.
The second point is that Baidu needs to hire industry leaders with a truly international vision and preferably experience in cooperating with giant companies.
We need such an outstanding figure to keep whiteness in check.
Robin Li's current favorite role is Andrew Ng, a Chinese-American, director of the Stanford AI Laboratory, and one of the founders of Google Brain.
If you don't move faster, then more and more Baidu will fall behind.
The third point was recommended to him by Meng Fanqi in the previous conversation.
The AI deep learning platform must be easy to use and open source.
Baidu is actually secretly developing this platform, and Meng Fanqi’s suggestion surprised him. This shows that the technical talent in front of him also has the vision of a leader or entrepreneur like himself.
By building such an open source community, we can attract a steady stream of AI algorithm talents to conduct research based on these tools.
This direction is still a desert, waiting to be explored.
Meng Fanqi's suggestion made Li Yanhong feel like the sun was clearing from the clouds.
Even if you cannot recruit enough talents, as long as you maintain this platform, Baidu will never lack related conveniences.
The threshold for ordinary people to access AI algorithms must be lowered and lowered again.
It is necessary to allow everyone to use these models and technologies with as little operation, contact, training, and modification as possible.
The Baidu AI platform must be transformed from a product that only a few scientists and large enterprises can afford to become the infrastructure of the entire society.
This is indeed something that Li Yanhong has not yet begun to consider at this stage. Li Yanhong, who has just woken up from a dream, has improved his evaluation of Meng Fanqi by a few points.
While Li Yanhong contacted Yu Kai and other capable subordinates to discuss various details, Meng Fanqi returned to the school library as always to make final adjustments and tests on the code that was about to be handed over to Baidu.
He still prefers the feeling of a library. It is too comfortable at home or in his own rental. There are countless temptations that may interrupt his work and thoughts.
Once the rhythm of study and work is interrupted, the efficiency is likely to be drastically different.
"I'll just check my phone." I thought to myself, and the whole afternoon passed.
Meng Fanqi, who was immersed in modifying the code, had no idea that Robin Li would be so resolute. Today's meeting directly advanced many of Baidu's decisions by 5-7 months.
In fact, in the final analysis, workers underestimate the determination of major technology companies to invest in this area.
Baidu's net profit last year exceeded 10 billion yuan for the first time, but Robin Li himself had a strong sense of crisis.
On the one hand, Baidu's core business is still contributing to steadily growing profits; but on the other hand, the mobile Internet is rapidly penetrating, and huge changes in the business world are about to happen, which makes Baidu fall into some confusion.
Meng Fanqi's AI algorithm performance, which is two and a half years ahead of its predecessor, is like a shot in the arm, giving Robin Li, who has already locked onto the AI track, a shot in the arm.
A year later, Robin Li dug out the internal letter he sent to all Baidu employees at this time last year, "Change starts with you and me."
"When the business is still growing rapidly, we should not quickly pursue net profits, but should invest more money in more new businesses and innovations."
