Many schools have a motto of encouragement, similar to "Today you take pride in the school, tomorrow the school takes pride in you."

For most people, this is just a motivational phrase. But for a few, they can truly contribute to the reputation of the school.

The same goes for research papers. Some papers take pride in being published in top journals and conferences, while others are the reason why those journals and conferences are considered top-tier.

For those who have some understanding of scientific research, they generally believe that good papers should be published in SCI, with high scores, and in first-tier SCI journals. This viewpoint is generally valid.

However, for computer science, especially in the field of AI, conferences are actually more important than journals.

Conferences are held annually, with fixed deadlines for submission and review, and all the authors of the accepted papers gather together to exchange and showcase their work. The review process is fast, with guaranteed timeframes. Attending the conference allows for direct face-to-face communication and discussion with the authors.

Journals, on the other hand, accept submissions throughout the year, with no fixed review timeline, and they do not organize author meetings.

For many disciplines, journals are considered more authoritative, formal, and have stricter review processes.

Theft is never good, try looking at [ pawread dot com ].

But for AI, journals are too slow. The pace of AI is so fast that many people cannot wait.

For example, last year when AlexNet won, if you had done a lot of research in early 2013, studied for four or five months, and then submitted your paper. If you submitted it to a conference, the review results would have been announced around August or September.

In November or December, you could comfortably go to the conference and have a good time, and then graduate smoothly.

But if you submitted it to a journal, it might still be under review after six months. By that time, when the new results of 2013 came out, the reviewers might find that someone else had achieved higher results than you.

It is normal for an article that could have been accepted to be rejected in this case.

At this point, it can be said that luck plays a big role. If you are lucky, you can transfer your article to a slightly lower-ranked conference or journal and still succeed.

If luck is not on your side, your article may be rejected 2-3 times, and it may become outdated. All the effort put into it for half a year would be wasted, and for many students, it could mean delaying graduation for a year.

Due to this unique and fast-paced nature of the field, AI values conferences more than journals. Success at conferences often leads to expanding the content of the paper and submitting it to journals for a slower review process.

If scholars in the field want to know the latest developments, they generally do not search in journals. They pay attention to the top few academic conferences and directly communicate with the authors by attending the conferences.

Journals? By the time they are published, the technology is already a year and a half old. Times have changed, my friend.

But at this moment, even conferences are too slow for Meng Fanqi. The preliminary results of IMAGENET have already been announced, and he believes that those who are paying attention to this matter have already set their sights on DreamNet and the Dream team.

Since even The New York Times has paid attention to this matter, it seems that it is time to add fuel to the fire for DreamNet with a generative adversarial network.

Some articles rely on the reputation of high-level conferences and journals to enhance their credibility, while others do not need to.

The deadline for submission to the top AI conference CVPR is November 1, 2013. After Meng Fanqi submits his paper, he can choose to anonymously make his submission public.

This is not against the rules.

However, sometimes this is just turning a blind eye and making a futile effort.

Just like the current situation, Meng Fanqi is ready to publish his preprint, "Generative Adversarial Networks Based on DreamNet."

Even if he chooses to be anonymous, who wouldn't know who the author of this article is?

At present, there is no information publicly available about DreamNet, so who else could publish a paper based on DreamNet at this time?

Similarly, as AI technology gradually evolves towards large-scale models, often requiring hundreds or thousands of GPUs, many times you can tell which company's research group it is just by looking at the number of cards used and which company's unique big data is used.

Others may want to do it, but they don't have the resources or the ability.

Since it is unnecessary, Meng Fanqi naturally wouldn't do such a boring thing.

"At this moment, what I fear the most is that others won't recognize me. So why bother being anonymous?"

With this in mind, Meng Fanqi published his paper on the arXiv website. As for whether the people at CVPR would consider this a violation of the double-blind review principle, it is not within his concern.

arXiv is an open-access repository for academic preprints and is a modern way of sharing research findings.

Initially, it was because the review process in certain fundamental disciplines, especially mathematics and physics, took too long. Some papers might not find anyone to read them for several months, and no one could understand them.

In such cases, many people would consider putting a draft version or even the final version directly on the arXiv platform.

This promotes communication and the speed of development in the field, and it serves as evidence of when one's research results were obtained.

Later on, computer science, statistics, biology, economics, and other disciplines gradually joined, and arXiv became more diverse.

Unlike formal conferences and journals, arXiv is just an open platform. It is similar to the illustration sharing website Pixiv. Neither arXiv nor Pixiv rigorously review the content uploaded, so the quality of the content varies.

arXiv is not a formal conference or journal, and the content published on it is not considered truly published. The content posted on arXiv has not undergone peer review and carries the risk of not being recognized.

Even amateur scientists can publish their perpetual motion machine masterpieces on it, pretending to be mysterious and profound.Not to mention anything else, deceiving the uninformed layman really has a miraculous effect. The paper abstract on the webpage, various field tags, and citation methods all look professional, and they are all in English. Non-researchers could easily be fooled at first glance.

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

The principles and codes are right here, with the same random seed, anyone can reproduce my results.

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

Even, he vaguely anticipates it, because such conflict will bring considerable attention.

Before long, Meng Fanqi submitted the latex source file of the Generative Adversarial Network, and compiled a pdf file on the website. Just waiting for two days later, the website will update these newly added articles.

Meanwhile, the original author of this paper, Ian Goodfellow, who had just thought of the rudiment of this idea and was still in the thinking stage, was a long way from the fully formed generative-adversarial framework.

Ian was on his way to the office of his doctoral advisor, one of the three giants of AI - Bengio.

"This is a brilliant idea, I need some help," Ian thought, "I want to complete this unprecedented excellent work in my last year of PhD."

He would never think that a more complete paper, with more detailed experiments, and more comprehensive discussion and reflection on related fields, has already been published.