In January 2014, faced with the security of the generative technology FaceGAN on face data, Facebook's DeepFace and Hong Kong Chinese's Tang Xiaoou Lab withstood the pressure and forcibly released their latest progress.

The idea on both sides is actually similar, the release technology may face some security issues and questions, but without release, there will definitely be nothing left.

After all, both sides have made some breakthroughs in performance, and they will not really be held back because of this kind of thing.

Compared to the 97.53% ability of human eye recognition on the LFW dataset, the technologies released on both sides, DeepFace and DeepID, break through this boundary.

The Facebook team, which originally had only 97.35 on the timeline, improved the accuracy to 97.89% with the help of some ideas from Meng Fanqi.

The Tang Xiaoou team, which originally had a capacity of up to 98.52% on the timeline, suffered from a performance reduction due to the release more than two months in advance, and the release level was 97.95%.

However, it still narrowly defeated the Facebook team.

"Facebook's algorithm has more real human image resources than we do, and we as a lab have a huge disadvantage in this regard, and there is a big gap between the two sides. But we still withstood the pressure and overcame the difficulties. When Tang Xiaoou faced the interview, he replied like this.

This is the first time that the artificial intelligence method has broken through the human level in the field of face recognition, and it has caused a huge sensation around the world.

At the same time, it also makes the already hot problem of falsely generated faces a new high, and after all the people who eat melons see these new technologies, the first question they want to ask is, can you, the latest algorithm, distinguish between falsely generated faces?

"Compare images of faces captured by cameras in all corners of the city to automatically identify crimes and make a splash. Facebook is gradually making it possible to quickly query character information based on gender, age and other attributes, and the plots in these sci-fi movies. ”

Facebook has some press releases promoting its technology, but it seems that the onlookers below are not very happy with it.

"Don't blow it, don't blow it, we just want to know if your new algorithm can recognize the generated fake faces?"

"Can your new algorithm handle the security crisis of artificial intelligence?"

".......", in the face of the masses who don't know the inside story, Facebook's technical team is also helpless.

Many technologies are not completely secure, but at this point in time, the issue is too hot and too sensitive.

Facebook's publicity strategy is not to answer positively.

But everyone knows very well in their hearts that not answering is actually an answer in many cases.

In this situation, if you can't answer this question head-on, then the answer to this question is actually very clear.

That is: "I'm sorry, I can't do it." ”

On the Chinese side, Hong Kong and have been affected much less by public opinion in China.

Although there are some relevant reports in China, they have not formed such an exaggerated storm of public opinion.

And because Tang Xiaoou's team still beat the Facebook team in the overall situation, it is generally a very worthy of publicity.

Raise the prestige of our country.

Therefore, CCTV News personally went out to conduct relevant interviews and publicity on this matter, and the overall tone was very positive.

As the creator of the record of facial recognition accuracy, Tang Xiaoou, a professor at the University of Hong Kong Chinese, led a team to defeat the Facebook team and the human discrimination level a few days ago, so that the machine can learn to [discern everything]. Professor Tang's laboratory has been selected as one of the top 10 pioneer laboratories in artificial intelligence in the world, and is the only laboratory in Asia to be selected. ”

“......”

"Let's ask Professor Tang to talk about the significance of this technology and artificial intelligence for humanity and China."

"The combination of artificial intelligence and industry will become the main driving force for the next wave of industrial upgrading, and accelerating source innovation is the key for Huaguo to seize a new round of development opportunities. There is no shortcut to technology research and development, it is a marathon of long-term investment. Only by mastering the technology at the source can we not be choked by other countries. ”

Tang Xiaoou sighed, "As an intellectual, I have enjoyed the country's best quality educational resources along the way, and I should do something meaningful for the country's source innovation and development, and also fight for the Chinese people in basic research, which is the feeling that contemporary intellectuals should have." ”

The teams on both sides are well aware of a major flaw in their current technology, but they tacitly do not mention it.

At the same time, Meng Fanqi Vision's AI camera has secretly updated a version of the generative image identification function.

Right next to the age and mood editing function, there is no special changelog or reminder, and it is quietly updated online.

Yang Ming of Facebook DeepFace, finally had some free time in the past two days.

Although the face security issue, which was involved in well-known media in the United States, was not resolved, this version was finally released.

As a part-time worker, he also ushered in a long-lost relaxed holiday.

"Recently, why do my classmates keep recommending this AI camera app to me?" Meng Fanqi's AI camera App is currently only on the shelves in China, and Yang Ming didn't see it, but many of his classmates have returned to China and have played this product.

"What's so interesting about the camera, and you say it's related to my current research direction? Just blow it. Yang Ming didn't believe in this matter at all, in his opinion, this kind of App is just adding some filters, how can it be the same thing as the high-precision face recognition he did?

When he saw these messages a few days ago, he simply ignored them. With DeepFace just around the corner, he didn't have much time to watch such boring effects.

It stands to reason that he wouldn't have watched it after he was free, but today I don't know why, I picked up my mobile phone and downloaded and installed this App that already has a lot of popularity in China.

"Age edit... Emoji editing..?? Is this what a camera app should do??? "Yang Ming was dizzy after reading it, this technology is much more professional than their Facebook team.

"Distinguish between truth and falsehood...? What is the distinction between truth and falsehood, is it the distinction between truth and falsehood that I have in mind? Yang Ming frowned, and said in his heart that it wouldn't be....

But if it doesn't mean "judging whether the image is generated by artificial intelligence methods", what can a normal picture say is true or false?

What the hell is this?

"I'll test it." Thinking so, Yang Ming logged on to the [These people don't exist] website and saved a few pictures.

After transmitting it, it was judged to be [fake face].

？？？

I opened the camera and took a selfie, judging that it was [real face].

What's the situation? It's not going to me, is it?

Yang Ming thought that when normal people saw this function, they would make a little fake generated picture to test.

Fake pictures can't be taken, they must be read albums, and real pictures are likely to be taken directly.

In fact, it doesn't have this distinguishing function at all, but just brainlessly outputs the real and fake according to the album and camera.

It must be like this!

Yang Ming thought so, and uploaded another photo of himself from before, which was still judged to be [real face].

Now he was completely desperate, and with the last stubbornness, he uploaded another landscape photo.

It turned out to be [faceless].

.... Your app is more professional than our dedicated Face Lab team...

Yang Ming felt that he was really faceless now.