Many schools will have an exhortatory motto similar to "Today you are proud of your school, tomorrow your school will be proud of you".

For most people, it's just chicken soup. But there are a small number of people who can really do their own to make the school popular.

The same is true for papers, some of which are proud to be published in top journals and conferences, while others are what make journals and conferences top-notch.

People who have a little understanding of scientific research will basically think that if a good paper is issued, it will be issued SCI, and a high-score, SCI journal in the first district. There is really nothing wrong with this view most of the time.

However, for computer science, especially AI, conferences are actually much more important than journals.

The conference is a form of receiving manuscripts at regular intervals every year, reviewing them regularly, and then bringing together all the authors of the manuscripts to exchange and present. He is characterized by fast review and a fixed time. In addition, participants can communicate and discuss directly with the authors face-to-face.

Journals, on the other hand, are traditional forms of submitting manuscripts at any time throughout the year, with an irregular review time and no organization for authors to meet.

For many disciplines, journals are more authoritative, more formal, and more rigorously vetted.

But journals are too slow for AI. The speed of AI is so fast that many people can't wait.

For example, last year, AlexNet won the championship, and you did a lot of research on your own at the beginning of 13, studied it for four or five months, and then submitted the manuscript. If you are voting for the conference, the review results will already be announced in August and September.

In November and February, you can go to the venue comfortably to talk and laugh with others, and then graduate.

But if you are applying to a journal, you may still be reviewing the manuscript after half a year, and when the new results of 2013 come out, the reviewer will take a look and find that they seem to be several points higher than you.

It's also very normal for articles that could have been accepted, so it's also very normal to be rejected.

At this time, it can really be said that the sky should not be called the earth, and if you are lucky, you may be able to go smoothly if you are lucky enough to transfer to a worse conference and journal.

If you are unlucky, the article will be rejected 2-3 times, and it is very likely that it will be completely outdated from then on, and half a year's hard work will be in vain, and for many students, it may be the end of a year of delay.

Due to this unique and fast discipline rhythm, the AI direction focuses on conferences over journals, and it is often when the conference is successful that the content of the article will continue to be expanded and submitted to the journal for slow review.

If scholars in the field want to know about the latest developments, they are basically not going to journals to search, but will pay attention to the top academic conferences, and participate in the conference to communicate directly with the authors.

Periodical? When it was published, it was already a year and a half ago. Times have changed, my lord.

But at this moment, even the meeting was still too slow for Meng Fanqi. The results of IMAGENET have been preliminarily announced, and he believes that those who are interested in this matter have their eyes fixed on DreamNet and the team Dream.

Now that the New York Times has taken care of this, it seems that it is time to send out a generative adversarial network to add another fire to DreamNet.

Some articles rely on the reputation of high-level conferences and journals to bolster themselves, while others do not.

The deadline for receiving CVPR is November 1, 2013, and Meng Fanqi can choose to anonymously publish his submission after submission.

It is not against the rules to do so.

It's just that sometimes, it's just a matter of covering your ears and stealing the bell, taking off your pants and farting.

As is the case now, Meng Fanqi is preparing to publish his preprint, DreamNet-based Generative Adversarial Networks.

Even if he chooses to remain anonymous, who doesn't know who the author of this article will be?

No information has been released about DreamNet's details right now, so who else could have published this DreamNet-based paper at this time?

In the same way, AI technology in later generations has gradually developed towards large models, often with hundreds or thousands of GPUs. In many cases, just look at the cards used, use whose unique big data, and look at the content to guess which research group of which company.

Other people don't have the resources or the ability to do what they want to do.

Since it was an extra move, Meng Fanqi naturally would not do such a boring thing.

"At this moment, what I'm most afraid of is that others won't recognize me, so why be anonymous?"

With this in mind, Meng Fanqi published the paper on the ARXIV website. As for whether the people at the CVPR meeting will feel that such an approach violates the principle of double-blind review, it is not within his care.

ARXIV is an open repository for scholarly preprints and a modern way to share research results.

At first, it was because it took too long to review manuscripts for some basic subjects, especially mathematics and physics, and some papers may not be read by anyone for several months, and no one can read them at all.

In this case, many people will consider putting a relatively drafted version, or simply the final version, directly on the public platform of arxiv.

This promotes communication and the speed of development of the discipline, and is also a kind of evidence to prove when the results of one's research were obtained.

Later, computer, statistics, biology, economics and other disciplines gradually joined, and arxiv became more and more all-encompassing.

Unlike serious conferences and journals, ARXIV is just an open platform. It's the same as pixiv, an illustration sharing site. Both arxiv and pixiv generally don't do a very strict review of the uploaded content, so the level of content on it can be high and low.

ARXIV is also not a formal conference or journal, and its content is naturally not actually published. The content posted on it has not been peer-reviewed and runs the risk of not being recognized.

Minke can also publish his perpetual motion machine masterpiece on it, pretending to be unfathomable.

If nothing else, it really works wonders to coax the uninformed layman. The paper introduction, various field labels, and citations on the page are all in English. It is easy for non-researchers to be fooled at first glance.

However, Meng Fanqi was not worried about this matter at all. The logic of modern AI is as simple and crude as that.

The principle and code are here, with the same random seed, anyone can reproduce my results.

Whether others admit it or not, question it or not, Meng Fanqi doesn't care at all.

Even, he vaguely expects that this kind of conflict will bring considerable attention.

A short time later, Meng Fanqi submitted the LaTeX source file for the generative adversarial network and compiled a PDF file on the website. Only two days later, the website will be updated with these new articles.

At the same time, Ian Goodfellow, the original author of the paper, has recently come up with the idea and is still in the process of thinking about it, and there is still a long way to go before a truly formed-confrontational framework can be formed.

Ian is on his way to the office of his Ph.D. supervisor, Bengio, one of the AI Triumvirate.

"It's a brilliant idea and I need some help." Ian thought, "I'm going to finish this amazing job in my final year of PhD. ”

He would never have imagined that a more complete article, with more detailed experiments, and more comprehensive discussions and reflections on the field, would have been published.