The generative adversarial network of later generations is a very well-known technology and is very popular. Ian himself became the top leader in the AI ​​​​world because of this technology. Another of his well-known contributions is the flower book "Deep Learning".

It is a biblical-level textbook in this field.

Therefore, Ian talked about the causes and consequences of this technology more than once during the interview.

So Meng Fanqi knew very well that Ian actually came up with this idea several months later on the historical timeline.

Meng Fanqi never considered the possibility of a car crash. In his opinion, the time difference was still at least half a year.

But I didn't expect that my proposal of the residual network in advance, which would surpass the human level in the machine's ability to classify images, would lead to such an unexpected chain reaction.

"Oh my god, oh my god..." Ian squatted down with his head in his hands on the spot, looking very crazy.

In fact, it is nothing new for academic ideas to collide. Sometimes the time difference is one or two months, and people usually don’t care about it, and it may even become a temporary gossip.

But that is a situation where one party is still reviewing the manuscript and the other party submits it some time later.

The academic community usually recognizes the difference in review time.

But if your article has not yet been submitted, the other party’s experiment has been completed, the article has been written, and even published.

Then no matter what you do, it will basically be useless.

In fact, at this stage, it is discovered that the academic idea has crashed, which is the time when the losses are minimal. Ian has not yet devoted his energy to officially start the project or start writing the paper.

It is still in the early stages of discussion and communication with the instructor.

But Ian understands the value of this idea very well, and the reason why he is crazy is because he is upset that he did not think of this idea a few months earlier.

"Is there much content in the car crash? How different are our ideas?" Ian made a few big moves to vent his emotions, and immediately began to think about remedies.

Academic ideas have crashed into cars. After all, everyone’s cars are still different brands.

There are merits in each other, and sometimes the one who develops early can inspire the other and push the other's research to a higher level.

"It's a pity." Bengio shook his head and sighed as he quickly browsed the article. "What we just discussed is a true subset of the other party."

At the same time, the printer in the corner of the office began to make a clicking noise, and sheets of A4 paper slowly fell down. Ian knew that his tutor was printing the paper that conflicted with his ideas.

"Has it been published in a very formal paper format?" The last bit of fantasy in Ian's heart was shattered.

"Looking at this LaTeX template, it should have been submitted for this year's CVPR. The other party's time may be much earlier than we thought."

"Let me take a look first." Ian stepped forward and flipped through the other party's summary and several formulas.

After only watching it for two minutes, I sat down on the sofa in the office very frustrated, "The other party is not someone who is rushing to post a short article to take advantage of the situation. They are prepared."

There are many people like this in academia. No matter what ideas they have, no matter what they are, they will write two or three pages and publish them on arxiv to make it appear that they are the first to discover this matter.

Experiment slowly.

This cannot be said to be wrong, but it is indeed a very bad trend.

Ian has always believed that his sense of smell is relatively cutting-edge. Even if he crashed this time, it was just because the other party was eager to take advantage of the situation. He must still have a lot of room for maneuver.

It's a pity that Meng Fanqi completely destroyed his illusions with very careful thesis writing.

Not only the definition of the concept, but also the derivation of the formula is very concise and sophisticated. Based on Dream's related generation experiments, the results are also unprecedentedly stunning.

"Dream, isn't this the method that won the top spot in IMAGE2013 a few days ago?" Ian suddenly realized, and he felt that he understood it completely.

It was because of this breakthrough human performance that Ian thought of using another network as the discriminator and spiral confrontation to improve the performance of the generator.

Since I was inspired by this incident, the person who designed the Dream backbone network a few months ago must have felt it more deeply than I did.

Then we followed the trend and conducted a generative confrontation experiment. It was indeed natural and reasonable.

Thinking of this, Ian let out a long sigh.

Maybe he would feel very upset if he lost to anyone else. Only losing to the author of Dream made him feel excusable.

After all, I was also inspired by the performance of this network.

"You have made great achievements in max-out, Theano and other things before. How come you crashed and collapsed after just one idea?" Bengio saw his students were as deflated as if a needle had been inserted into a balloon. Suddenly

It feels a little funny.

"We old guys have all survived the old days when AI was not popular." Bengio activated the old man's general skills and recalled the glorious past.

"In the next ten years or so, how many people will be willing to admit our neural network? Don't be discouraged, cheer up."

Although he was so comforting, deep down in his heart, Bengio actually felt very heartbroken.

His field of vision is much farther than that of Ian, who is in excitement and excitement at this time. As he said before, this technology can be used to generate realistic human faces, as well as real objects, paintings, and even videos.

Related fields have great potential for development and application.

Although the prototype is just a generator, if you want to make it into a style transferer or a face modifier, it can really be done with just a few modifications.

It's a pity that I have just come up with the idea, and other people's articles have already been published with detailed formula reasoning, experimental results, and even application prospects in multiple fields.

"This may be the luckiest thing in misfortune. After all, everything hasn't started yet." Ian was depressed for a while, but he still forced himself to cheer up.

He encouraged himself, "If I really saw this article when I was about to finish it, I really don't know how uncomfortable I would be."

Turning his head, he looked at the author column of the article again. He really couldn't figure out where this dream came from.

However, compared to the little-known Huang Tang in the competition team, this Deqing Fu seems to have a reputation.

Ian used his mobile phone to search for Fu Deqing's information, and the more he looked at it, the more he frowned.

"Isn't this a professor from the Department of Mathematics? Why did he come to the cutting-edge computer field to cause trouble!" Ian was very angry. He actually doubted whether he had checked the wrong information.

But after searching around, I couldn't find anyone else with the same name who looked like that.
Chapter completed!