Huang Renxun deceived the whole world, and no one noticed it for more than three months! The leather clothes are fake, the kitchen is fake, even he himself is fake!
In the past two days, a piece of news that shocked the science and technology circle was frantically swept across the screen. The thing is, NVIDIA recently released a behind-the-scenes documentary of the GTC conference in April , showing how technicians can use the Omniverse platform to produce various 3D simulation special effects on site. These include the perfect reproduction of Huang Renxun's kitchen and the synthesis of his own "digital man."
What? Boss Huang, who is so realistic in front of him, is not a living person?
Don't get excited yet, it's not what you think. In fact, according to the Nvidia blog , in the 1 hour 48 minute press conference that day, "digital people" only took up 14 seconds (original film 01:02:41 – 01:02:55). Outside of these 14 seconds, Huang Renxun himself was speaking.
21 true and false Huang Renxun
The 14 seconds that Nvidia called the climax of the press conference looks like this:
When introducing Nvidia's first CPU Grace, the real kitchen instantly turned into a fake kitchen, and began to "disintegrate" like a Transformer. Then, the scene entered the virtual space, and Huang Renxun's "digital man" appeared in his fake leather suit.
▲ You read that right, due to the epidemic, the Nvidia conferences in the past two years have been held in the CEO’s kitchen
You can completely understand this paragraph as a show off to the Omniverse platform. According to the official description, these 14 seconds combine Nvidia's deep learning with the work of the image research team, engineer team, and designer team.
In order to recreate this kitchen, technicians took hundreds of photos from different angles and then performed 3D modeling. It is said that there are as many as 6000-8000 items here. They also quietly buried three Easter eggs in the fake kitchen: a coffee grinder with the Nvidia logo, a can of olive oil with the brand name Tao, and a Lego villain with the appearance of Huang Renxun.
As for the "fake" CEO himself, he needs to give Huang Renxun (and his leather jacket) a full-body 3D scan, modeling based on thousands of photos, and then using AI to train facial movements and expressions.
In addition, they hired special actors to learn and imitate the body language of Huang Renxun's past speeches, and then perform 8 hours of motion capture. After many detailed adjustments, the team has made 21 versions of "digital humans" for the boss to finally select for the shot.
real or fake? The big problem behind 14 seconds
On YouTube, the introduction of this behind-the-scenes documentary is: watching a group of artists blur the boundaries between reality and rendering in Keynote at the Nvidia conference. In the body of the blog, Nvidia stated that the moment the kitchen "disintegrated" surprised the audience. Everyone began to wonder how much of the entire press conference was true and how much was attributed to AI.
To be honest, looking back at these 14 seconds carefully, the effect of the kitchen "disintegration" is still amazing, but the synthetic version of Huang Renxun's body movements are relatively stiff, and even his mouth is a little out of alignment. It is not too difficult to distinguish it as a masterpiece of AI, and it is not too difficult to distinguish it from the live version.
But when the misreading news came out, I also fell into self-doubt for a while, and I began to feel that in the 1 hour and 48 minutes, every smile of Boss Huang when he appeared on the scene seemed to have the same robot taste as Zuckerberg.
Well, it seems that he is fake. Um? Is he real?
Why are we confused by the question of true and false old yellow? It may be because we live in a world where "seeing is not necessarily believing".
Take Deepfake technology as an example. It was once used to produce pornographic videos of female celebrities, or fake news about politicians and celebrities. Nowadays, the application of this technology has become more widespread. All you need is a headshot, and your female colleagues or female friends may become victims of nude photos .
In the era when everything can be deepfake, arbitrary tampering with video and voice is just a basic exercise6.
The human cute girl on the other end of the phone may actually be the AI manipulated by the big-footed scammer. In 2019, scammers used AI-synthesized speech to pretend to be the CEO of a British energy company's German parent company, and successfully defrauded 220,000 Euros (about 1.73 million RMB) from employees of the company.
Today, even the live broadcast may not be credible. For example, Nvidia recently introduced a real-time virtual image technology. Even if the programmer is wearing a mask and fishing in a cafe, what the other party sees in the real-time meeting screen can be that he is personable in his office suit at the moment. As long as you want, you can instantly switch to the sun and beach vacation image.
Many experts reminded that this phenomenon can easily make people distrust the content of social networks and even the entire society.
With the blurring of the boundaries between reality and virtuality, our radar for distinguishing between humans and AI will frequently fail. In the Metaverse or the future that will come true, you ask me if the old yellow in front of me is really old? I really dare not answer.
Humans need a little spinning top
Just last month, the CNN documentary "Roadrunner: A Film About Anthony Bourdain" caused a wave of discussions about the ethics of AI use.
The protagonist of the documentary is the late star chef Anthony Bourdain. A lot of video and audio materials from his lifetime were used in the film, but the director did not take the initiative to inform the public: there are three narrations that sound like Bourdain's own narration, but they are actually generated by AI technology.
▲ Anthony Bourdain
Although the original texts of these three sentences are from the emails written by Bourdain, they still make people feel uncomfortable, angry and deceived. New York University journalism professor Meredith Broussard said that unlike fictional film and television works, people think that documentaries are real. If AI is used in the work, it should be clearly stated.
Nicole Greene, an analyst at Gartner Research, also agrees to clearly inform when AI intervenes , “Whether it is a brand cooperating with a virtual idol or a movie using AI technology to reproduce the role of a deceased actor, as long as consumers know what is happening, they are acceptable."
▲ "Fast and Furious 7", using CG special effects to bring Paul Walker back to life
Since 2015, the Associated Press has been using AI to automatically generate company earnings news, and each article will be accompanied by a note: "This story is automatically generated by Automated Insights." To tell you clearly, this is the manuscript written by AI Moving Bricks.
This simple sentence is like the "parent's guide" before the start of a horror movie, the packaging of instant noodles and the advertisement of burgers "everything prevails in kind." It is the most basic information notification, and it is also the responsibility of the producer to the consumer. In an era when deepfake and other technologies consume trust, they can even rebuild trust.
We do not refuse AI to break into real life. After all, it brings infinite new possibilities for efficiency and creativity (as for the abuse of AI, that is another topic). But we all want to have the right to know.
Lil Miquela, one of the world's hottest virtual idols, debuted in 2016 and only admitted in 2018 that he is not a real human being. Now, her Instagram self-introduction is written as "a 19-year-old robot living in Los Angeles"; and Imma, who is also popular, calls herself a "virtual girl interested in Japanese culture."
Although it is essentially just a string of codes composed of 0 and 1, it does not affect young people's pursuit of them. Today, Miquela and Imma have millions of fans, fashion and advertising cooperation resources are softened, and they are veritable social network celebrities.
▲Can you see which model in the picture is a real person?
Looking back at the true and false Huang Renxun of NVIDIA this time, if I can bother the AI or the staff to add a line of explanation on the demo screen, is it possible to avoid "shocked! The whole world is deceived" "Lao Huang in the press conference is all fake "This kind of oolong incident?
In a world where true and false are difficult to distinguish, humans need the spinning or falling top in "Inception" to remind us: the front is a virtual space, please stay vigilant and pass carefully.
Currently, Internet companies such as Facebook, Google and Twitter are studying how to detect and curb the spread of Deepfake videos, and some areas have also issued relevant bans. But this is still like a "cat and mouse" game.
As for the limits of AI use and ethical issues, it is estimated that relevant laws and regulations will be promulgated in the future. Friends, we are standing at the entrance of the meta-universe. The false is true and the true is false. Everything has just begun.
#Welcome to follow Aifaner's official WeChat account: Aifaner (WeChat ID: ifanr), more exciting content will be provided to you as soon as possible.