Text-to-image models are so last month, text-to-video is here AI progresses rapidly. Just months after the release of the most advanced text-to-image models, developers are showing off text-to-video systems. Meta announced a multimodal algorithm named Make-A-Video that allows its users to type a text description of a scene as input and get a short computer-generated animated clip as …
Seems that was false news; the deal with Deepcake was only about that one video.
I think we'll end up with the bum print authentication as premiered in Monsters vs Aliens.
As a matter of fact, the whole scene is shockingly close to reality :).
It's not a great idea to use anything you leave behind as authentication for critical things. That includes your fingerprints and your DNA. If I can get a copy of those by being near you for an hour or so and monitoring what you've touched, it's not good for proving your identity for a one-time thing. It works a bit better for something where you provide new samples each time. For example, using a fingerprint to unlock a door or device makes more sense than using one to sign a contract, but there's a risk for both so if it's really important, don't even use it for the door.
Bruce Willi's agent denies the story
https://www.bbc.com/news/technology-63106024
A spokesperson for the actor told the BBC that he had "no partnership or agreement" with the company.
And a representative of Deepcake said only Willis had the rights to his face.
Deepcake fake...
Since NLP is poised to replace paper mills with machine-written papers, this is a nicely symmetric development in this particular arms race.
I'll also note it's one I predicted in a post to one of the rhetoric mailing lists about a dozen years ago, but since it's an obvious development anyone paying attention ought to have seen it coming. It hasn't been a technical problem for a long time – just a question of economics. Existing paper mills enjoy cheap labor1 and an established network of customers2, which makes them harder to displace by automation. But not immune to it.
1Mostly unemployed and underemployed people with advanced degrees, according to research I've seen. There was a long piece by a former paper-writer in ... Harper's, maybe? ... some years back that supports this.
2In the US, this includes fraternities and sororities, and organizations that "support" student athletes.
If the the NLP program shouts, "J'accuse!" at a paper, how can the authors effectively defend against erroneous accusation? Does the program give an itemized listing of things it thinks are "too similar" to other published texts? Do publishers blindly accept this programs results simply because, "The computer said so"?
Text-to-video will be a lot more palatable, and less resource-intensive, when it stays comfortably far from photorealism. Text-to-cartoon running on moderate resources would probably be very popular with kids, "creatives", and people who are just interested in doing it for a hobby or the occasional lark.
Primitive versions of that are already available, but an improved version, perhaps coupled to a streaming service like Steam for backend processing, might actually be commercially viable. It offers some of the same creative opportunity as, say, Minecraft.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
