Ian, born in 1985, has been under a lot of pressure in recent years.

For an ordinary doctoral student, he has actually done nothing wrong. On the contrary, he is intelligent and hardworking, and has always studied in prestigious institutions.

He graduated from Stanford with a bachelor's and master's degree, and his mentor was Google Brain's founder, Andrew Ng. He pursued his doctorate in Montreal, under the guidance of Bengio, one of the three giants in AI.

His academic background and personal intelligence can be said to be exceptional, but compared to his mentors, he is somewhat overshadowed.

At the age of 28, Ian is in the final year of his doctoral studies, and the generation algorithm is the direction he has been contemplating during his doctoral period.

Many researchers have been trying to use neural networks to simulate the human brain and generate new data.

Unfortunately, so far, these data usually have significant problems and are completely unusable.

And a major obstacle is that the generated images are often indescribable. These useless pixelated images lack means of utilization and are difficult to evaluate.

However, the recent results on IMAGENET, where the model's performance surpassed human standards, gave him new inspiration.

If deep networks can already achieve human-level performance in some image tasks, can't we avoid the problem of poor evaluation of generative network results by completing this process through the mutual confrontation of two networks?

The more Ian thinks about it, the more reasonable it seems. He didn't even have time to sleep last night. Instead, he spent the whole night implementing a prototype version of the generative-adversarial network and tried training it. The results seemed to be consistent with his expectations.

Unable to contain his excitement, he wants to discuss the principles and methods with his mentor as soon as possible.

Ian paced back and forth in front of Bengio's office for more than twenty minutes before Bengio walked briskly from a distance.

Ignoring the white hair mixed in his beard, the 49-year-old Bengio actually looks quite young. If we only consider his appearance, he would be at most a man in his mid-thirties.

Especially the two eyebrows that resemble Mr. Bean's shape, make him appear more honest and approachable.

Ian quickly took a few big steps forward, while unfolding his notes and experimental records. He pointed to the core equations on them and was about to speak, but Bengio ruthlessly interrupted him.

"I know you're in a hurry, but don't be in such a rush." Bengio looked at Ian's eager appearance and teased, "In academia, we emphasize solid foundations and steady progress. What's the use of being in a hurry? Being in a hurry is useless."

"Why can't we wait until we go inside and sit down to discuss it slowly?" Bengio said, took out his keys, and opened the office door. "Come in, it's only a little past seven and you're already squatting in front of my office door. You must have made some major discoveries, right?"

"I'm really in a hurry, I'm really in a hurry." Ian has been deeply involved in the field of generation for some time now. When he came up with this idea yesterday, he realized that it would be a major change in the entire model.

From his undergraduate years to the present, he has spent ten years honing his skills. Ian is very excited. He has a good academic sense and can feel the uniqueness of this idea. Is his day to shine finally here?

At this moment, he can feel that this is a great opportunity, where he can break free from the shadow of being a disciple of Andrew Ng and Bengio, and let the world truly recognize him by reevaluating himself through the techniques and algorithms he proposes.

"Mutual confrontation of two networks," Bengio pondered for a while after listening, unlike Ian's friends who were eager to express opposition and doubt, Bengio has always been very encouraging towards imaginative and creative ideas.

"Train two models simultaneously, instead of just one. The generator and discriminator can spiral and enhance each other. You want to use this approach to make the distribution of the generator as close as possible to the provided real data."

"In this way, we can break away from the usual thinking. We train the discriminator not to make it 100% accurate in judging whether the images are generated or not. Instead, our goal is to make the discriminator no longer clear about what is real and what is fake," Ian added.

"Adversarial is just one form, and its purpose is to train a generative network. I am confident that this approach can eliminate the reliance on data labels and make extensive use of other data without the need for additional processing."

Ian's excitement has not subsided until now.

"Not only that, generation is just the most basic application." Bengio pondered for a while and said, "It can completely achieve special purposes through data filtering. With just a few lines of code adjustments, I believe it can handle patterns with simple layouts very well, especially in the direction of faces. This is a very dangerous technology."

"What do you plan to name it?" Bengio opened his computer while speaking. Naming a technology is a difficult problem and also an art.

Search tinyurl.com/2p9emv8w for the original.

If the name is not good, the work will be in vain.

A catchy name, an interesting and concise abbreviation, can greatly enhance the visibility, popularity, and scope of application of the work.

"Generative...Adversarial. How about calling it Generative Adversarial Network?" Ian hadn't thought about this question carefully yet, so he added the two most crucial elements of the idea to the name."GenerativeAdversarial, let me take a look." The first step in naming is naturally to search, to see if anyone has already taken the spot.

Upon this search, Meng Fanqi, who was far away, suddenly sneezed. He wondered why he was catching a cold in the middle of summer.

"It seems that this name has already been taken by someone, just a couple of days ago. Very unfortunate." Benjio had not yet realized the severity of the problem, "It seems we need to find another name."

As he spoke, Benjio clicked on the preprint.

"What is DreamNet? I don't seem to have seen it before." Benjio vaguely felt that someone had mentioned this name in the past few days, but he was busy with other research matters and didn't pay much attention.

But after reading the abstract, Benjio's eyes sharpened, and his expression became very serious.

Ian was still deep in thought, wondering what kind of name would be best.

But Benjio shattered his defenses with a single sentence, "Ian, it seems that it's not just a name we can't use."

"What? Our ideas have clashed?" Ian was shocked. In the academic world, having the same idea as someone else could be one of the most unfortunate things for a researcher.