Samsung is controversial with its surprising “artificial humans” from the Neon project
A subsidiary of the Korean giant claims to be able to create and animate avatars so realistically that it would be impossible to distinguish them from real humans.
After a mysterious teasing, the Neon project was finally unveiled at the CES. Led by Star Labs, a subsidiary of Samsung, and directed by the researcher Prana Mistry, it consists of creating and animating videos with human figures using artificial intelligence.
Samsung did not skimp on the means since Neon has a dedicated stand at the show which arouses the greatest curiosity. Prana Mistry also held a press conference to present Core R3, the platform that powers Neon.
Neon would become a true friend
Before arriving in Las Vegas, he had slipped in some information about how it works.
The software uses a mix of images of real human beings to create a virtual person and then produces new expressions, movements or dialogues in several languages.
Social networks are already full of deep fake. But Neon claims to go much further than that. On the official website, the company claims to be able to create virtual characters capable of “showing emotion and intelligence”.
They could interact with real humans and respond to our requests with very little latency. Neon would thus no longer be a simple assistant but a “real friend”.
The applications could be numerous: welcoming people in shops or public places, and, why not, one day replacing TV news presenters or actors in movies.
No real demonstrations
The first videos are impressive, to be sure, and rather frightening if we think of the use that some people would make of them for misinformation.
But the company was careful to point out that they were fictitious, prepared in advance for illustrative purposes only. It is therefore impossible to verify the effectiveness of Core R3, as The Verge cautiously pointed out.
In the meantime, experts in artificial intelligence are annoyed by Star Labs’ speech. Laurence Devillers, a professor at the CNRS, is offended that we can lend a human personality to this type of creation that only imitates us:
This is reminiscent of the presentation of the Sophia robot in 2018 that had already taken place at the CES.
The machine was supposedly capable of responding to impromptu media interviews, but the videos had been staged and reworked during editing.
Find all the CES 2020 news on News Tech
Yann LeCun, a researcher at Facebook, then virulently denounced the deception.