"How about coming to study with me? You can start directly from the last year of your undergraduate degree and then go straight to a Ph.D., with a full scholarship."

Li Feifei sat in Han Ci's seat and started talking to Meng Fanqi.

"Hahaha, the full scholarship is not the key. My main concern is that I may not have much time at school in the following years," Meng Fanqi laughed. Stanford's tuition is not cheap, about 50-60 thousand US dollars per year, including living and accommodation expenses, it would be around 70-80 thousand US dollars per year.

For ordinary students, this is a huge amount of money. Many American students need to rely on student loans, and they still have a large amount of debt even after working for several years.

But for Meng Fanqi, several hundred thousand US dollars is already a negligible amount.

Just the profits brought by the recent rise in Baidu's stock in the past few days have already exceeded 500 thousand US dollars.

Speaking of these stocks, Meng Fanqi plans to gradually sell these shares in the US stock market in a few days and invest in some promising companies.

He has already contacted Secretary Liu and transferred the second installment of 18 million yuan that was advanced by Li Yanhong to his domestic account.

Because he suddenly remembered that at this current time, it seems that there is still some money to be easily obtained in China, and this is the last window of opportunity.

That is Mi Huyou, a company that later became worth billions, but currently seems to have not seen the light of day.

"Let's go to Shanghai in the middle of this month." Meng Fanqi didn't remember much about the non-AI companies, he rarely paid attention to this before his rebirth.

But the experience of Mi Huyou is quite extraordinary, so it's worth it for him to make a special trip to Shanghai before leaving.

"It doesn't matter if you're not at school. Wherever you do research, it's still research," Li Feifei said. Besides, Li Feifei herself has a close relationship with Google. If a student can produce such high-quality academic achievements on their own, the supervisor wouldn't care that much.

"Is that so?" Meng Fanqi actually just wants to symbolically obtain a degree, he doesn't really care about this matter.

Since he has to publish so many papers and achieve results anyway, why not get a degree along the way?

As long as the supervisor doesn't interfere too much, he doesn't expect the supervisor to provide much guidance and resources.

Apart from his unique advantages, he already has Hinton's guidance and Google's resources at this point, so his conditions in this regard are already at the top.

Li Feifei obviously also understands this. The main selling point is a relaxed and favorable environment. "You can come directly to study the last year of your undergraduate degree, get the credits first. After graduation, you can come to me and continue studying. We can directly collaborate with Google on the papers."

Not only does Li Feifei herself have a close relationship with Google, but in 2017, she also officially served as the Vice President of Google, mainly responsible for AI research and the integration with cloud products.

Even Google's two main founders, Sergey Brin and Larry Page, are both computer science Ph.D. students from Stanford.

So for both Stanford and Google, it doesn't matter if Meng Fanqi works at Google and studies for these degrees at Stanford. They are all one big family.

This is also the main reason why Meng Fanqi specifically mentioned Stanford.

"Is that settled then?" From Meng Fanqi's perspective, Li Feifei is a very good choice.

First of all, she is an immigrant of Chinese descent and takes care of Chinese people. For example, Jia Yangqing, who later became the VP of Ali, although he was not Li Feifei's student at Stanford, she still urged him to study multiple times.

Although his current status is not that high, he later became an academician of the American Academy of Sciences and is considered a top-tier mentor in the field of AI at Stanford.

This novel is available on "p𝑎wread.com".

Secondly, her relationship with top journals such as "Nature" and "Proceedings of the National Academy of Sciences" and top conferences in computer vision is very good. Her IMAGENET competition is often held at these top conferences.

By collaborating with her, it would be much easier to publish his own achievements. With the quality of his results, he can basically secure awards at several conferences.

Finally, her influence at Google would to some extent make it easier to promote his achievements and even facilitate specific collaborations and other internal matters at Google.

Meng Fanqi had already planned to apply to Stanford directly at the conference. Therefore, he had prepared a printed resume, research proposal, and even a recommendation letter written by Hinton.

He was well-prepared, and Hinton could only sigh and lament why his own school is so far away in Canada. It's true that proximity to water brings early success.

At this time, after some confusion on stage, Han Ci began her presentation.

"So far, the development of AI-related disciplines has completely changed people's previous understanding of AI. Meng's residual thinking has achieved amazing results in many image tasks.

For example, it can identify images more accurately than humans, or even generate completely non-existent images out of thin air. And these remarkable achievements are mainly achieved through solving.

For any image problem, what we are interested in is the mapping function from the image to its specific meaning, such as the category of the content.

The usual training practice is to give an efficient approximation of the objective function based on a limited amount of data. Or to approximate the underlying unknown probability using a limited sample without labels."

"The basic components of a neural network are: linear transformation and one-dimensional non-linear transformation. A deep neural network is generally a repeated composition of the above structure.

For a constructed network, we design an optimization problem, fit the data based on empirical errors, sometimes adding some regularization terms, and solve this optimization problem."

Han Ci's slides began to show numerous formulas. "From this, we can decompose the error into three parts: the approximation error is completely determined by the selection of the hypothesis space, the estimation error is the additional error brought by the size and quality of the dataset, and the optimization error is the additional error brought by optimization or training."Although he had already roughly understood Han Ci's thoughts in this area, Meng Fanqi still held great admiration for her as he watched her nervously yet confidently present these highly professional topics in such a formal setting.

Many of the conclusions she presented were well-known to later generations, but most of them were empirical conclusions drawn from a large number of experimental observations.

Even though Meng Fanqi would pioneer many AI application fields in this life, the inherent mathematical proof would always be an insurmountable gap for him in both of his lives.

People inevitably look at those who can accomplish these things differently.

At this moment, in the conference hall, it wasn't just the awestruck Meng Fanqi who was impressed by Han Ci.

The academicians led by Hinton, such as the old professors from the Oxford team, were full of approving gazes.

This was a treatment that even Meng Fanqi had not just received.

"Perhaps this is the old scholars' preference for theory," Meng Fanqi said with a smile, shaking his head. As someone who works on AI application technology, he might never be able to win such favor from these eccentric old men.