In fact, Meng Fanqi has also fantasized about such a thing as rebirth countless times, in his dreams, the progress of career and love is always fast and smooth sailing.

However, when there was a real opportunity to walk back on the college campus ten years ago, Meng Fanqi found that it was easier to buy BTC.

He did not have the memory of writing "Dream of Red Mansions" silently in the other world, but memorized 300 Tang and Song poems.

And in the age of AI, the update of code and tools is so fast.

As of this time, not to mention the two AI tools that have not been released in the future, even the earlier Caffe has not started development.

The entire academic community is still in the aftershocks of the AlexNet earthquake, and many people in the academic community still have not fully understood what the deep neural network looks like, and where is the ready-made framework for him to use?

Meng Fanqi walked on the way back to the dormitory and began to think about how to deal with the first problem of being reborn, as long as the server survives these few months, a part of the BTC is enough to buy.

And after winning the first place in the competition, Google will not be stingy with a few machines to lend you. Meng Fanqi is very aware of this past, as long as the algorithm is successfully implemented, Google will give him an astronomical amount.

The trickier part is the code framework used to build this model. When he first got involved, the hottest tool was TensorFlow (TF) in 15, and two years later, Facebook proposed PyTorch in 17, which was easier to use, and gradually caught up. Meng Fanqi also switched to PyTorch for the most part.

He'd heard about some of the previous frameworks, but honestly never touched them.

The later it gets, the better the packaging becomes, and even elementary school students can directly construct a high-performance cutting-edge model with two lines of code.

But that was ten years later, and now there are still more than two years before the release of the original version of TF, and Meng Fanqi is a little irritable.

His current plan is to take a look at the source code of the original version of AlexNet last year, and then look at Alix's own homepage, which should yield some gains.

I really can't do it, so I can only take advantage of the fact that Alix is proud of the spring breeze now, and he has not signed a contract with Google, and he asks him for advice.

Since AlexNet is a milestone event in the era of deep learning, and the name of its model is named after him, Meng Fanqi has a little understanding of his paper, the network, and even himself. Clicking on the web and searching for their name will easily find their simple profile on the University of Toronto website.

Not only that, but this personal homepage also details some of the code he developed and accelerated himself during his time at school. Mainly C++ and CUDA code based on NVIDIA graphics cards, basically in the 10-12 years time period.

"A good man's life is safe!" Meng Fanqi clicked on the link and found that he not only recorded the code and results in it, but also remembered clearly what was corrected every time he changed it, and even left a sentence next to it, "If any code can't run, please feel free to contact me." ”

Two lines of tears flowed from Meng Fanqi's eyes, it is precisely because of the selfless dedication and sharing of generations of algorithm pioneers that the AI industry has developed so rapidly in the past decade.

"Good code, I'll copy it!" The data and code are being downloaded, and if the cuda-convnet framework can be used, then even if the problem of the code is half solved, Meng Fanqi can initially implement some key algorithms based on this prototype framework.

Thinking of this, Meng Fanqi had to sincerely thank the algorithm interviews that were becoming more and more involuted in later generations.

He is very aware of his abilities, and even if he goes back to 10 years ago, his coding skills are not outstanding. His strength lies in his clear understanding of many technologies and the company's development path.

The reason why I know these well-known algorithmic techniques well and can reproduce a considerable part of them. Thanks a lot to the programmer's unique interview mechanism. In larger companies, it is common for you to meet you for three or five rounds of technology in one interview, test you to write code on the spot, and ask you about classic technology for you to implement.

In addition, the programmers are very fond of sharing, and they love to write "faces", that is, interview experiences and experiences. There are often well-wishers who provide very detailed and perfect answers to these questions.

In order to screen candidates, large enterprises have become more and more difficult and tricky. As a result, Meng Fanqi is familiar with the basic core technologies and classic papers of each route.

It's just that some people in later generations have written the tools well, and understanding these classic principles is not very useful in their work, but they didn't expect to be able to make the most of them here.

In addition to the code, another important task is the details and reasoning of several important papers.

Meng Fanqi basically remembers the main context and logic, but the subtleties are not thorough, and he still needs to work hard to polish it.

If you want to share the biggest cake, there are still quite a lot of things to do, Meng Fanqi carefully calculated, many of the cornerstones of the future direction of AI, in fact, most of them are proposed in the period of 13-15 years.

In addition, if you frequently propose algorithms and publish papers one or two years in advance, it is likely to cause a butterfly effect and make progress in the field more rapid. As a result, you have less time than you would otherwise be.

Therefore, Meng Fanqi's preliminary plan is to try to implement or put forward theories in the first one to two years, that is, to about 15 years.

In the following 15-18 years, the resources that have been exchanged are mainly used, whether it is economic resources or traffic popularity. Shift your focus to focus on a few projects that break through the human level, such as AI Go, AI games, AI painting, and AI language. With AI based on its own algorithms, it has frequently set records that beat the top level of humans. The paper only needs to post a few important pieces of content.

Once you have accumulated enough assets in the process, you can take a large number of shares in some technology companies in advance. In 18-19 years, you can gradually transform, gradually leave academia, and focus on investment or AI entrepreneurship to avoid the technological advantages you gradually lose.

Other than that, if anything, it can only be a separate residence. Yenching Electric Power University, as a university with electricity written in its name, cuts off power to students every three or five nights. Sometimes model or algorithm tests run for days on end, and Meng Fanqi doesn't want to be affected by lights out and power outages at night.

In addition, if the computer is doing algorithm work, it will run under high load, and the movement will be extremely high. Even if you can do this in the dormitory, I'm afraid it will make your roommates sleep.

Rent a small room near the school, the money I have saved in the past two years has just been invested in BTC, and I really have no choice but to ask my brother to borrow some.

Ask your brother to borrow money, and his seniority will inevitably plummet. I haven't taken a few steps on the road to the godfather of AI, and "calling my father" has become true. If you don't call your dad twice, how can those roommates who lost the goods lend money to themselves?

"Cherish the last few months of picking." Meng Fanqi patted his empty wallet with some sadness, "I may no longer know what it feels like to be short of money in the future." ”