As the prevalence of AI continues to dominate headlines, it seems no industry is safe from adopting some version of this technology in its vision for the future. In this case though, the AI isn’t just in the news, it’s reading it to you. Courtesy of Ars Technicathe new startup Channel 1 presented a proof of concept video demonstrating AI avatars as news anchors, and the results are both genuinely impressive and a head-first dive into the depths of an uncanny valley.
These AI avatars are based on real people (who, we are told, “were paid for the use of their likeness”) and rely on what are described as “sophisticated behind-the-scenes systems” to create their vocal rate, mannerisms and expressions. , and at first glance, the results are remarkably convincing.
The lip syncing is some of the best I’ve seen, with sequences of difficult syllables handled with remarkable precision, and they look, for lack of a better term, like believable people. The gestures are entirely relevant to someone who’s taken a few courses on how to present on camera for dummies, and aside from the odd moment of blur that makes you think what you’re looking at might be a fac -AI-generated simile, it’s a pretty convincing one. technology display.
But as with all simulations of an authentic human being to date, it’s the things you can’t quite put your finger on that let it down.
It’s something about the sincerity and pace of delivery that just seems off, to the point where the more I watched, the more it felt like someone might be sneaking up behind me. AI humans are already here, it seems, and in this case, they’re staring, almost unblinkingly, straight into your soul.
It doesn’t help that these avatars appear to have been trained on an amalgam of existing news anchors, and the AI appears to have interpreted its version of the stereotypical newsreader vocal cadence as thick as molasses. It’s like the sound has been slowed down a little too much, and that, combined with the blank expressions and soulless smiles, makes the little hairs on the back of my neck stand up.
Yes, it’s a very cool tech demo. But wow, this made me uncomfortable after just a few minutes, to the point where it became a little difficult to watch.
Beyond the slightly terrifying presenters, there are actually some AI use cases demonstrated in the video that make a lot of sense. An example is given of a pre-recorded interview with a (real) French resident, who explains how his property was damaged by a recent storm in, wouldn’t you believe it, French.
The clip is then replayed but with an AI translating the images into English, with lip sync, and the result is remarkably convincing to the point that unless I’m told in advance, I don’t think I’ll spot any tampering involved. all.
Deadline reporting that Channel 1’s reporting will come from a “yet-to-be-announced news agency” and that content will be provided by independent journalists, and that the reporting itself will be constructed from primary sources such as government documents, with humans involved “every step of the way.” the editorial process. The upcoming channel also plans to display an icon in the corner of the screen to let audiences know which parts of the programming are AI-generated.
Regardless, given that newsreaders in particular work hard to establish a level of authority and trust with their audience, the fact that this can supposedly be simulated with such a degree of accuracy means that this technology could potentially be used for nefarious purposes. – such as the spread of false information or dangerous propaganda – could well be cause for concern.
It seems like not a day goes by without some sort of AI rubicon being crossed, but this stands out as a particularly hair-raising example of the technology hinting at some potentially disruptive uses to come. Deep fakes have been a major concern for some time now, and while the technology to completely simulate a reasonable human being to the point where differences become difficult or impossible to spot seems to be developing at breakneck speed, ethical concerns become even larger. more, well, worrying.
Yet in the meantime, I suppose we can admire it as a technical achievement, if only as a technical achievement. While these avatars may be based on real people, later in the video an entirely AI-created presenter reveals himself to deliver a report on cars, but the results are much less convincing.
A small thank you, and proof that technology still has a way to go. But given the demonstration presented here, it doesn’t seem like it will be long before AI avatars and real people truly enter the realm of the indistinguishable.
