It was Wang Kai's own idea to start a business with Meng Fanqi, but Meng Fanqi still felt a little embarrassed. After all, Li Yanhong had specially sent him to learn this part of the code, and he felt a little guilty for taking him away.

So the next day, Baidu came again and optimized the current method for recognizing license plates in images to make up for it.

Embarrassingly, during this time in China, he had been working behind closed doors, eager to break through the technology and accumulate his technical prestige in various directions, without considering entrepreneurship or making early preparations.

If he wanted to start a business temporarily, it seemed a bit hasty. It was estimated that he would spend a lot of effort on recruiting interviews early next year.

First of all, there were no technical personnel around him that he could use, and he hadn't looked for them in advance. The familiar classmates were still in undergraduate school and were not suitable for major roles.

Although this was not a big problem in the early stages, Meng Fanqi could produce results on his own, but to start a company, experienced programmers were still necessary.

After all, Meng Fanqi only had a basic understanding of the front-end and back-end of the app. Most of his former colleagues from his previous life were still in school, and it would be absurd to approach them without warning. He knew them, but they didn't know him at all.

There were a few former superiors from his previous life who had some skills and vision, but at this time, they had just entered the workplace as employees and could be recruited.

"You have to think about it. If you start a business with me at this stage, I can pay you more, but I can't guarantee that it will work in the end," Meng Fanqi said frankly. "And after I go to the United States in a few months, besides being responsible for the technical aspects, the rest will depend on you."

"How can I let you worry so much!" Wang Kai immediately expressed his stance. Other than that, nothing else mattered. Why didn't he have the motivation to work when he was employed? It was because the company wasn't his own, so he just took the fixed salary and did the corresponding amount of work.

If the pay was low and there was constant nagging, it would be difficult for him to help you.

But now that there was hope of getting shares, it was different. If the early-stage company was initially formed and only focused on the pure software part, it would only take about ten people to support Meng Fanqi's AI algorithm and turn it into a product to be sold.

As long as the product was developed, getting one or two shares would be the starting point of financial freedom.

The small salary he earned at work, after tax, was only around three hundred thousand. After deducting daily expenses, he could save up to two hundred and fifty thousand at most in a year.

From twenty-five thousand to sixty-five thousand, he would barely be able to save up to ten million.

The biggest fear of starting a business was to die before achieving success, and the stock options would become worthless.

This biggest risk was not a problem at all in Wang Kai's eyes and was not worth doubting.

With the level of technology that Meng Fanqi had demonstrated, it was definitely possible for the company to go public.

Just by relying on the advanced face recognition algorithm itself, it would not be difficult for the company to sell it for tens of millions.

If there were a few complete projects, the valuation would easily reach double digits.

If they really got involved in government projects and provided this technology for customs, government agencies, and transportation hubs, it would be astronomical.

Face recognition was what Wang Kai had been thinking about, but it was not important in Meng Fanqi's eyes and was almost forgotten.

"There are too many things to do. After Google's recommendation and advertising algorithm update, I need to carefully plan and record it," Meng Fanqi said to himself.

At this moment, Meng Fanqi was updating his Google Scholar profile. It had not been long since he publicly released a large number of papers in Sydney, but it had been some time since he published the paper on generative adversarial algorithms.

He wanted to see if his actions had caused any major changes.

The most convenient way to do this was to see which papers had cited his work and if there were any significant research results.

After updating his Google Scholar profile, Meng Fanqi was surprised to find that in just a few days, he had more than twenty citations.

Upon closer examination, these twenty or so citations actually came from only four or five papers.

Because Meng Fanqi's latest publication had revolutionized the entire paradigm and the code was open-source, any research related to deep learning would likely have to cite several of his papers from the beginning.

Residuals, optimizers, training methods, data augmentation - almost no one could avoid these four pillars.

When you're just trying to make great content at pawread.com .

With each new paper in the field of deep learning, the number of times Meng Fanqi's work was cited would almost quadruple, and this multiplier would continue to expand in the future.

By 2023, the most cited scholar in human history would have been cited millions of times.

The number of articles in the field of AI, which was over 20,000 per year in 2012, had quickly increased to about 135,000 per year by 2021.

At this rate, in just four or five years, Meng Fanqi would become the most cited scholar in history at the age of 25 or 26, and his citations would continue to grow exponentially in the years to come."Up to the time before my rebirth, the original author of the residual network, kaiming, has been cited more than four hundred thousand times," Meng Fanqi recalled slightly, the AI technology he is now publishing and planning to publish is several times more than kaiming.

It's not impossible to break three million by the year 23.

The academic papers are catching up with online novels, and the number of citations is equivalent to subscriptions, and the actual number of times the articles are read is tens of times more than this.

To be able to write academic papers with this level of popularity, it seems that there are no predecessors.

Among these published articles, Meng Fanqi's Google mailbox has long been flooded with emails from all over.

When Meng Fanqi clicked in, the computer froze half to death and took more than half a minute to recover.

In the inbox, there were questions, magazine invitations, requests for unpublished code, and greetings from colleagues.

Most of these emails were in English, but there were also some written in Chinese, which should be clear about his nationality.

After scanning through, one email from the Shanghai Public Health Center caught his attention.

Meng Fanqi searched and found that the center is a century-old city-level tertiary hospital with many types of advanced medical equipment, especially skilled in the diagnosis and treatment of liver diseases.

Meng Fanqi checked the detailed information of the hospital and inferred that they should have many imaging results of different instruments for various diseases.

After reading the email, Meng Fanqi understood the other party's intention. The residual U-Net segmentation method he updated before has greatly improved the technical level and segmentation effect of image segmentation.

Especially in the segmentation of fine-grained objects, there has been a significant leap.

And the segmentation type of task is very important in the application of medical imaging.

Because in medical images, they are usually taken after preliminary diagnosis by various departments.

The significance of classification and detection is not great, and further content analysis is the main requirement.

For example, separating the lesion area in detail, or assisting in diagnosing the degree of lesion, which will make it easier for medical staff to diagnose and save a lot of time.

"The actions of the tertiary hospital in Shanghai are really fast, they have already noticed me," Meng Fanqi knew the significant impact of the U network on medical care, but he thought this would be a task for the next year.

But he didn't expect someone to come knocking on his door the day before yesterday.

Meng Fanqi straightened his attitude and felt that the government and the hospital, these two behemoths, seemed to be within reach.

This start-up seems to require a more serious approach.

Not to mention the large-scale government project demand for facial recognition, just the medical AI alone, if done well, is enough to go public.

Meng Fanqi carefully pondered for a moment and felt that these two directions would be best handled by two separate companies.