Deepfakes

[Perry Staltic]

We need to start getting used to the idea that we can no longer believe what our eyes/ears perceive in images/videos because the technology now exists to make things that are completely fake look and sound real. Post any videos/images you run across that you know are deepfakes


View: https://www.youtube.com/watch?v=lhi31-pOz5M

 

[Perry Staltic]

Rare Exquisite Apu (Alabaster) on Gab: 'This is at once hilarious and frightening. Deep…'

Rare Exquisite Apu (Alabaster) on Gab: 'This is at once hilarious and frightening. Deepfakes are, in reality, quite dangerous to the fabric of the society.'

gab.com

 

[ThinPicking]

We need to start getting used to the idea that we can no longer believe what our eyes/ears perceive in images/videos

No we don't, because we can.

the technology now exists to make things that are completely fake look and sound real

No it doesn't.

 

[Perry Staltic]

View: https://www.youtube.com/watch?v=gLoI9hAX9dw

 

[Perry Staltic]

This guy is a solo act and can do this. Imagine what those with deep pockets can do


View: https://www.youtube.com/watch?v=p7-B8S734T4

 

[Normal Human]

No we don't, because we can.


No it doesn't.

Did the Tom Cruise video look fake to you? You would have been able to tell it was fake if the title of the video wasn't 'Deepfake' Tom Cruise?

 

[Perry Staltic]

Notice they use the word "clone". In the "Damar Hamlin is dead" posts on twitter, many posts said Hamlin had been cloned. It appeared to be fake bot activity because people really don't talk like that, but maybe they were telling us what they were actually doing, ie, cloning from actual to virtual.

A few days ago, speech AI startup ElevenLabs launched a beta version of its platform that gives users the power to create entirely new synthetic voices for text-to-speech audio or to clone somebody's voice.

A new AI voice tool is already being abused to make deepfake celebrity audio clips

ElevenLabs' AI speech synthetizer is being used to generate clips featuring voices that sound like celebrities reading or saying something questionable.

www.engadget.com

 

[Perry Staltic]

Rubbing it in our faces?


View: https://twitter.com/HamlinIsland/status/1617678293876224002

 

[Peatful]

Deepest (visual) fake was easily 911.

 

[Summer]

If this technology is just now being presented to the public, imagine how long the military and 3-letter agencies have had access to it.

 

[Beatrix_]

I think making people believe in clones is just next level psyop, a new smoke screen

 

[Momma]

I think making people believe in clones is just next level psyop, a new smoke screen

I think people need the heads up on Project Blue Beam and that technology.

 

[Beatrix_]

I think people need the heads up on Project Blue Beam and that technology.

Definetly.

Additionaly, all this deepfake stuff will be used by the deep state to blame the people and to justify increased censorship.

 

[Dapose]

Seems like a good Technology for overthrowing your political opponents and getting behind the more attractive political leader that has a nicer smile and is better at lying to your face.
I like the way kyle Dunnigan uses it.

 

[akgrrrl]

Really, people are surprised? You have watched CGI since 1985 in thousands of films. Photographs have been cut and pasted and rephotographed since late 1800s. Hollywood plastic surgery since 1940, (didnt JohnWayne have work done 1947?).. all kinds of CIA crap at home and abroad since 1947 using technology the masses dont see for 30years. Hmmmm lets see, 1966 StarTrek flip phones, lasers...took until 1980s to roll out a car phone.

 

[philalethes]

No we don't, because we can.

No it doesn't.

It's extremely close to getting there at this point. At that point it's easy to claim that you still somehow can tell the difference, but a blind test or two would probably quickly put that to rest, like the anti-vegan sausage expert who ended up calling vegan sausages "luscious and lovely" on live television, believing them to be animal meat.

 

[Momma]

Artificial Intelligence "Deepfake" Horror Story: 2/1/2023 | PragerU

A scandal unfolded this week among several popular Twitch streamers and YouTube personalities around the emergence of “deepfake" images generated by…

www.prageru.com

 

[ThinPicking]

"The Party told you to reject the evidence of your eyes and ears."

It's extremely close to getting there at this point. At that point it's easy to claim that you still somehow can tell the difference, but a blind test or two would probably quickly put that to rest, like the anti-vegan sausage expert who ended up calling vegan sausages "luscious and lovely" on live television, believing them to be animal meat.

Getting where @philalethes, and how close is "extremely" close. In any test like this, blind or not, it's the observer who's being tested and not the fallibility of the subject.

I understand you only need to fool a critical mass of dumbarses for a larger mass to lose their ***t. But we're nowhere near and probably will never be at the point where any serious person could make the statement "we can no longer believe what our eyes/ears perceive in images/videos". The truth is it does and will always depend on more than the authenticity of a rendition, which can and will always be detectable, technically and perceptually at a particular threshold resolution.

But sure, lose trust in your own senses if you like

 

[ThinPicking]

Did the Tom Cruise video look fake to you? You would have been able to tell it was fake if the title of the video wasn't 'Deepfake' Tom Cruise?

Yes.

It looks fake as **** pal. Not to mention inaccurately expressed.

 

[philalethes]

I understand you only need to fool a critical mass of dumbarses for a larger mass to lose their ***t. But we're nowhere near and probably will never be at the point where any serious person could make the statement "we can no longer believe what our eyes/ears perceive in images/videos".

Again, the type of thinking you're exhibiting here strikes me as being akin to the anti-vegan sausage "expert" I mentioned above, who has high faith in their own ability to discern between what's "real" and what's "fake", but quickly gets proven wrong once put to a blind test. That being said, there's still a little bit left until we have AI-generated photorealistic images that are indistinguishable from real ones, but that really is not far off. You ask how far, and I can't predict the future of course, but within a year or two we're probably there already. Now generation of videos has already started being done, and while that's even more rudimentary at this stage, that's probably not far off either.

Saying that we're "nowhere near" this seems highly unreasonable, given the quality of the images being generated at this point are. In fact, for certain specific areas there are already extremely realistic images being generated that I believe could fool the most intelligent people out there. Also, it seems rather arrogant to think that you're just that much better than "all those dumbarses" out there.

Consider e.g. this article from last year, where it's clear that most people were already struggling with discerning texts and images; of course, you might chalk that up to them being "dumbarses", but you should probably attempt the test yourself or similar ones before passing judgment (there are some unavailable YouTube videos in that one for queries about AI-generated music, but you still get to see how you fared on images separately). Also, curiously, the people who were most confident in their own ability to discern performed the worst.

Maybe better would be to try something like this; try guessing on ~50-100 images or so, and keep track of your success rate. By your own reasoning it sounds to me like you think you'd be able to get it right nearly 100% of the time, but I'd be highly surprised if you could get a rate of more than 60-70% (of course I can't verify whatever claim you make with regards to this, you could always respond that you got 100% right even without doing the test, so it's more for yourself and your own intellectual integrity).