Executive Order 2023.02 (henceforth "NIGHTMARE WORLD WHERE FACT AND FICTION COEXIST")

  • 🔧 Actively working on site again.
It only works with english and it doesn't handle a lot of heavier accents too well. You can put other languages or heavy accented english in, but it'll just spit out a typical american or british english speaker that has a similar tone.
Ah, so non-english speakers are safe, for now. I'm just imagining the chaos if this tool can emulate the Spanish or chinese language, oh boy.
 
  • Agree
Reactions: Wright
Didn't eleven say they were able to reverse match audio generated by their AI with the person who created it? So it's this private entity that would have that ability. Which seems even worse.
The neat thing about Pandora's box is, it doesn't only just effect the person who opened it.

https://github.com/CorentinJ/Real-Time-Voice-Cloning

Computers these days are equivalent to super computers 20 years ago. Its why all of this machine learning tech is taking off so fast.
Modern computers are sufficiently advanced that any retard can just read what they've published and duplicate it.
 
I have no patience to wade through 10 pages of pure shitposts but I want to point out one thing: it is very often the lolcow themselves who posts AI-generated deep fakes, and we archivists might inadvertently repost them on KF. I know we need to develop a certain degree of suspicion, but we are not foolproof, and independent verification is labor intensive and not always possible. So, @Null, do you give us leniency when this happens?
you will never get in trouble archiving something from the thread's topic my dude
 
In Orville season 3 episode "Gently Falling Rain" it is revealed that Krill consider it normal that sides in election mass-produce deepfakes of their opponents where they do all kinds of shit, including gassing protestors. Note that the episode seems to have been written in the 2019 at the earliest so it's not particularly visionary.
 
Bloody hell lmao

I know it's been said but we are truly entering a new age. Obviously at the moment the mouth movement looks a bit odd but I'm sure that'll be ironed out and then, well, who the fuck knows what will happen.

EDIT: (from @Sneeds post in the thread)
View attachment 4440897
Thanks for the mention
Also if it makes you feel any better @Null the ai can’t get your autistically complex voice right. As heard by the samples below
Nulls big announcement:
View attachment 4368852

American Dankula:
View attachment 4368968

Dankula possessed by a demon:
View attachment 4368980
help me gays
Joshua Moons dying words:
View attachment 4369012

Null saying YWNBAW
View attachment 4369216
Kanye giving the farms a shout out and talking shit about Ralph
View attachment 4369228
Null reciting the sneed pasta
View attachment 4369232
biggie smalls calling Null out for liking anime
View attachment 4369220
And finally Null giving turkey Tom some advice
View attachment 4369328

View attachment 4370388

Null, the marine sniper skilled in gorilla warfare

View attachment 4371484
Null makes a certain amphibian enjoyer mad
@Frog God kick nulls ass

Null lies about hating anime
View attachment 4379256

Ask and you shall receive.


View attachment 4382124
 
I've seen a few examples of AI and Machine learning videos, like with the latest Twitch dude fiasco (Cant remember his name, because I honestly don't care) but the Joe Biden one Null posted made my jaw drop. Not only is the video quality good, but the speech has his inflections, his mannerisms, and style of speaking. It's incredible.


And it's fucking scary.

This will be abused. It probably already is. Glowies undoubtedly see it's potential. "Hey, the CIA needs more funding, Mr. President. Really hate to see a convincing video of you get released."

People with enemies see it's potential. In this age of viral videos, people share shit 5 seconds after viewing it without even checking to see if it's accurate. As long as it suits their opinion or agenda, or would get them likes/popularity.

Corporations are seeing dollar signs. "We can either market the software, market the software used to combat this, sell the data of the users, or use search engines to look for real or fake videos of our customers or employees."


It's long been said that information wars are more important than physical or cultural ones. Well this just threw a giant nuke into it.


Although if anyone ever makes a deep fake video of me clappin cheeks with Megan Thee Stallion and Lizzo in a threeway, I wouldn't be that opposed to it.
 
Thanks for the mention
Also if it makes you feel any better @Null the ai can’t get your autistically complex voice right. As heard by the samples below
Some users have said the ai makes your voice sound smooth and gay
It's too smooth. Josh could never talk this long without his voice cracking at least once.

The Ai makes him sound extremely gay for some reason
Here’s more ai Null

My favorite true & honest MATI clip.
View attachment 4400308
Make josh talk about his how he is in love with the fat woman he simps (Chantal?) Or some ludicrous shit like that just for giggles
@Super Male Vitality got you covered
 
its still pretty easy to spot/prove a deepfake audio or videoclip
but i think we're about to enter a brave new age of propaganda at the same time censorship is going exponential
if this was a fictional universe this would be one of the moments you could pinpoint the moment things became truly dystopian
except even sci fi writers hundreds of years ago predicted better technology than the fucking garbage we have today
i was promised flying cars
fuck this gay earth
fuck the antichrist
 
All you gotta do is make the AI say things that vaguely sound offensive and aren't and then splice it together on your own if the day where specific words get blocked ever comes.
I doubt it'll be a major impediment. GPT takes industrial hardware for now because text transformers are a particularly beefy type of AI (which is conveniently also why you can trick the censorship, because specific applications are downstream training from the really heavy learning base and the censorship is downstream from that) but that's an exception, most things aren't that expensive to run or retrain (or even completely rebuild yourself from the papers) on consumer hardware.
That's why you may have noticed it doesn't fucking matter that Stability pussed out and tried to filter porn out of Stable-Diffusion for example, lol.

I haven't looked yet but I assume this is super true of voice transfer stuff. I can't see why it would be heavy to run, and all the important retraining is at the endpoint because supplying arbitrary voice personas is the whole point. And obviously Deepfake-style face-transfer stuff are the same deal (at least for shorter one-shot clips) and there's a million experimental variants which are well out of the bottle at this point.
It'll all just get easier; corps poisoning the knowledge-well is a major concern in general but the censorship at least is futile and a temporary problem if you know what you want the result to be.
 
Back