With these features added by Meng Fanqi today, Baidu can basically handle the first wave of government demands.
The functions commissioned by Li Yanhong are also completed.
Indeed, as Wang Kai worries, if Baidu continues to focus on the project without sufficient technical accumulation, he will still encounter the same problems in subsequent needs.
It's just that this incident is a bad thing for Baidu, but not for Meng Fanqi.
Of course, Baidu is not completely devoid of technical strength. For example, based on the detection of vehicles, Baidu can completely perform functions such as recognizing license plates.
At most, it is far behind Meng Fanqi in terms of speed and accuracy. However, at this stage, other people do not have Meng Fanqi's technical level, so it is actually completely acceptable.
"If this giant Skynet project goes smoothly, the next step should be face detection, right?" Meng Fanqi began to predict the government's needs.
AI has gained momentum since then, but its applications in government projects are not that many.
Among them, the largest one, the most widely used one, and the one with the most core demand is probably face recognition.
This technology has attracted the attention of the government for a long time. It has been used to prevent and combat major criminal crimes since 2001, and has been put into large-scale use in 2008.
However, these methods are all products of the old era based on traditional methods.
After Alex unveiled the era of deep learning and Meng Fanqi led the true deep network, there will be a qualitative leap in the speed and accuracy of face recognition.
According to the original timeline, in June 2014, SenseTime would independently develop the DeepID series of face recognition algorithms, with an accuracy rate of 98.52.
While defeating Facebook's DeepFace in the same period, DeepID's face recognition capabilities surpassed the level of human naked eye recognition for the first time, breaking through the warning line of large-scale industrialization.
In August of the same year, SenseTime received tens of millions of dollars in financing, and in October SenseTime Technology was officially established.
By 2023, SenseTime's market value will be close to 100 billion. It initially relied on facial recognition technology that surpasses human levels.
In that timeline, the residual network has not yet been proposed. Meng Fanqi thought for a while and decided to take action at this time. Even if he does not use too many later technologies, it should not be difficult to achieve a recognition ability of 99.5+ with the residual network + ArcFace alone.
Is it necessary to start a short-term business? Meng Fanqi hesitated on this matter. If he took action at this time, his technical strength would definitely be enough to win the facial intelligence part of the government system.
It’s just not clear whether rashly establishing a small company with a small number of employees is qualified to contact the person in charge or the responsible department of the government.
If you are not qualified to join the game, it will be a bit useless for the company to run, and it may end up being acquired.
emmmm, think about it carefully, it seems acceptable? Meng Fanqi looked at Wang Kai thoughtfully.
A very key thing is that many AI technologies cannot be monetized in time after they are developed, such as the basic technology of ChatGPT, attention mechanism->Transformer->GPT->GPT2->GPT3->GPT3.5( ChatGPT)->GPT4.
The technical upgrades along the way, the first five versions are just a waste of money, and you may not be able to afford it.
But it is extremely valuable to develop such results in a big company like Google.
In order to lay out the technology, he had to develop many technologies first. Under this situation, it would be a better decision to wait for the time to mature while sharing the shares with large manufacturers.
However, the huge Skynet project and the funds received much earlier than expected made Meng Fanqi change his plan.
Meng Fanqi originally thought that he would be noticed at the end of 2013, and that he would not be able to accumulate tens of millions of wealth until the middle to late 2014. However, he never thought that 2013 was not over yet, and he had already accumulated a lot of funds on hand, and he was qualified to participate.
Meng Fanqi focused too much on the shocking technologies after 2020, and somewhat ignored the early implementation capabilities of technologies such as face recognition.
If he can take over the face algorithm part of Skynet, his products will be available all over China, and the company will become an officially endorsed technology company.
After explaining the code part for Baiduo vehicle detection and tracking, Meng Fanqi called Wang Kai to the corner of the office and leaned against the window.
Wang Kai, a muscular man who is about 1.8 meters tall, looks like he is being blocked by Meng Fanqi, who is about 175. The picture is a little weird.
"I thought about it carefully and realized that Skynet's demand for face technology is a rare opportunity. It's also a share of the money with big manufacturers. Now that there's no shortage of money, it's better to take some things yourself."
The former has a smaller commission, but it is more stable and does not need to worry about so much. It is the best choice for accumulating assets in the early stage.
Now that there is no need to worry about funds, you should eat the big cake by yourself.
"Brother! You finally figured it out! If you have this technology, you must be able to work alone. There are not a few brothers like me who are willing to work hard for you even with a little bit of equity!" Wang Kai was overjoyed. The opportunity for financial freedom fell from the sky. .
"However, if you start a business, you must be fast and win the project within two or three months." Meng Fanqi has published too many technologies and articles in the past two years. Paying too much attention to the company's projects is overkill, and he also lacks Sufficient senior management experience.
First, win a few large projects with absolute technical advantages, and then entrust professional company managers to take charge of management.
"I can provide funds to register a company, recruit some non-technical personnel, and first make the product to test the waters. You can pay more attention to this matter. If the subsequent key goals can be achieved, the initial members will all have a certain degree of success." proportion of the initial shares.”
"What is our first major technical product?" Wang Kai was overjoyed at first, but he quickly calmed down.
Although Baidu's work is a bit tiring now, it has been on the rise. It was definitely right to defect to Meng Fanqi, but you had to ask clearly about the situation and not be too rash.
"The first core technology is a face recognition algorithm that surpasses human level." When Meng Fanqi said this, Wang Kai understood the implication of Meng Fanqi's words without any unnecessary explanation.
After all, during this period, he knew a lot about the content of government projects and knew that a high-precision facial recognition system was a very important government need.
"The face detection algorithms on the market today are still old methods. I believe that with your level, the products you make will be able to get rid of those things by a long way." Wang Kai suppressed his inner excitement and thought as calmly as possible. .
"The demand for facial recognition algorithms is no less than that of road traffic. It has a wider range of applications in many more fine-grained scenarios. Whether it is government projects such as customs and safety supervision, or specific company check-ins, or even combined with this year's popular Mobile payments.”
Wang Kai became more and more excited as he talked, while Meng Fanqi's eyes darkened when he heard the company clocking in, and he recalled the unbearable years.
If the self-developed technology is applied to face check-in, merit will definitely be deducted.
"But this matter is not that urgent. It still needs a few weeks of preparation. You should first do the things at hand in Baidu and this part of the Skynet project. If you really want to start a business with me, you should complete the handover as soon as possible. Okay, don't affect Baidu." After all, Wang Kai is Robin Li's employee. To be fair, Robin Li has taken good care of himself during this period.
Therefore, Meng Fanqi gave a few more words: "If you want to find someone through your other connections, such as classmates and friends who studied AI before, if you have an idea to start a business with me, it's best not to bring Baidu away." Someone from the AI Research Institute."
Although the first batch of people from the Baidu AI Research Institute basically left to work on autonomous driving two years later, Meng Fanqi is still reluctant to take the initiative to poach these AI talents.
On the one hand, he is not good-looking, and on the other hand, he plans to get a large amount of data from Baidu in the future.
"I can give you this technology in about a year, and you don't have to rush to tell me whether to participate."
"Why are you not in a hurry? If you are in a hurry, I will definitely follow you!"
