Earlier this week we told you how artificial intelligence (AI) was being used to create whisky. Now for something a little more alarming…
A new AI experiment has seen the Mona Lisa brought to life. The famous work of art is just one of a host of iconic faces to recently appear in fake videos.
Based on a single image of a face, researchers from a Samsung AI Centre in Russia have developed a new algorithm that can create this type of false footage.
More than 7,000 YouTube videos of celebrities were fed to the system in order to create the algorithm. This meant that it was able to learn the movements of ‘landmark’ parts of the face.
By observing the mouth shapes, eyes and the nose of a face the algorithm could then apply the same movements to a still image later on.
The Moscow-based researchers also performed similar effects on images of David Beckham and Marilyn Monroe, basically puppeteering their features.
The Mona Lisa brought to life certainly got attention…
The AI can now turn the Mona Lisa into a real-life person. All with just one image.
“It performs lengthy meta-learning on a large dataset of videos,” explained researchers in their study, published on arXiv.
“And after that is able to frame few – and one-shot learning of neural talking head models of previously unseen people (…).”
So, that’s why the algorithm was fed images of the likes of the Mona Lisa and Albert Einstein. Despite obviously being fake, the finished footage seems surprisingly realistic.
— Simon te Brinke (@gramercypark) May 24, 2019
TechCrunch report that “the quality and obviousness of the fake talking head varies widely” in the footage. It adds that “usual smears and weird artefacts” are also omnipresent.
However, its report adds that “that said, it’s remarkable that it works as well as it does”. TechCrunch’s report, like others, also points out that this new AI only works on the face and upper torso. The Mona Lisa cannot yet dance or clap hands (but perhaps ‘yet’ is the operative word…)
So, what do you think? This new AI has split opinions since it was revealed. The Mirror reports that the new algorithm could have “a number of helpful applications”, especially in the film industry.
However, it also concedes the AI (in the wrong hands) could be abused to create ‘deep fake’ videos. Last year, several videos of this kind, appearing to show celebrities in explicit situations, appeared online.
It also raises the chances of videos showing politicians, or even people who have died, saying and doing things they never did.
The Telegraph, also reporting the story, referenced concerns too. Its report pointed out that US lawmakers have warned such faked videos “could be damaging to national security”.
So, as interesting and entertaining as it this type of AI can be, in the wrong hands the consequences are unthinkable. What do you think? Let us know!
To keep up to date with the latest happenings in the tech world stay tuned to the FileHippo News Blog?