AI Generated Portraits Of People Who Don't Exist

Thought this was pretty impressive. Click "Reload" on your browser to show a different image.

https://www.thispersondoesnotexist.com/

 

Yeah, this is all based on deep learning neural networks. A lot of advances have occurred here in a short space of time, including 'deep fakes', voice synthesis made in the style of a real speaker and just recently, whole paragraphs of manufactured text based on an input one line narrative (which could be anything, plausible or not).

While it's cool tech, I shudder to think how this will be used (and it will be) to supercharge fake news generation. The fake news bots and bot farms from the past will look positively quaint by comparison when you have fake video of real people saying fake things in their own voices. Brace yourselves!

 

We may no longer be able to respect any A/V content, for anything more than entertainment. The future may regress to the past requiring us to rely solely on multiple respected sources in print.

 

What I'd suggest on the scenario you've painted, is helping people develop critical thinking skills. Most sensible folk can distinguish good people from shady ones already, and if something seems too good to be true... get another source.

 

I certainly applaud your call for critical thinking skills. I think it should be a compulsory subject in school just like reading, riting and rithmatic! But it won't be, I'm certain of that - sigh. Unfortunately, people tend not to seek out alternate sources of truth these days. They tend to just go with what's in their newsfeed on facebook or whatever other social media portal, which is sad.

You have to be intellectually curious and engaged in order to seek out alternate, reputable sources of information and the way to get there is being taught critical thinking from an early age. Of course, there are forces fighting this type of education tooth and nail, because it's in their best interest to keep the population uniformed and disengaged.

 

See, now I'd just say regard everything with a little healthy skepticism. That's really all it takes and you're on your way.

 

In theory you're correct. However actually developing competent critical thing skills goes against human nature.

Most of us are looking only for our current belief system to be reenforced. It's like a drug — tell me what I want to hear.

Additionally, the deep fakes will become so good that it won't be worth your time attempting to sort them out — and even if you don't believe them, simply watching them may affect your beliefs subconsciously.

 

Good thoughts. A few years ago I became a little cautious about anything that causes an instant rush though. It's like, too much adrenaline too quickly, and that feeling began to trigger mistrust rather than reinforcement. You get suckered enough times, you learn that, I think.

 

Only if you think you're being suckered. Social media apps are built around the endorphin rush one gets when a person gets a notification that something has happened or someone has liked something or any other type of interaction. Some become so addicted to this that they can't go more than a few minutes without checking their phones for a status change. These types of people are unlikely to wonder outside their portals for corroborating information.

 

That is, until they come to you with "Look what I found!" and you take some time with them to disprove it. I had that happen with a coworker who thought a Conceptual Model of an iPhone with a holographically generated keyboard actually existed, because after all she saw the picture...

 

What's the story here? Are these pictures completely computer generated or just real photos that have been altered? The first option would indeed be scary, since they look very real.

 

if you look at a few of them you can pick up some odd defects.

also, i find this deeply terrifying

 

I don't feel too troubled by this - especially since these were the first 2 images it generated...

 

Here's a link to the article that introduced me to it, which explains a little bit of the technology involved:

Website uses AI to create infinite fake faces

 

our two future computer overlords

 

Yes, for the past 100 years or more, photography, video, film and recorded audio have been considered fact, good enough to present as evidence in court.

But unless a test can be developed to verify that an artifact is legitimate and not doctored, those media types will no longer be reliable as fact. They will/are becoming the equivalent of illustration and animation.

Anybody can create/invent anything they desire, and have it appear as photographic. Most people in the audience will have no ability to decern the difference — especially when they are not looking for it.

Somewhat related, is a story from about ten years ago, when a friend of mine had to sue someone for property damage. He had photographic evidence that was sure to proove the person was guilty. However, even back then I predicted that there may be some question whether the photos had been retouched with the all too popular Photoshop. I recommended that he re-photograph the eveidence with film, rather than digital, and be prepared to present the film negatives in court. His attorney was impressed and said it was a good call. We'll never know if necessary, since it was settled without going to court. (Note: while it may be possible to create a false film negative, it would be quite a big undertaking, and not worth it in a property damage case of that level).

 

They must be just be using parts of real photographs and reassembling them because I’m positive, especially with the women, that I recognize the features of certain celebrities and in these photos. A lot of this computer generated stuff is overrated. Elon musk just quit some organization that had software to generate fake news. I mean this is only how far computer has come? They had computers generating poetry back in the 70s. None of these programs impress me any more than programs I’ve seen 30 years ago. You don’t need much of a super computer to write a fake headline by assembling a few well-known names, events, and a few pat actions like “arress” “sues” “steals”.... And this software —if in fact it is just playing jigsaw puzzle with Photos —- doesn’t impress me either. If you wanted to truly generate never seen before faces you’d have to start with 3-D mesh and work up from there Applying textures and features. And of course we know how realistic that looks from all the movies we’ve seen.

 

I reversed imaged searched one of the more believable images and it turned up some college kid's obituary photo from 2003. Creepy stuff - it's obviously trawling through Google for pictures and then combining them in frankenstein like fashion.

 

You're speaking about your experiences and your capabilities. The big concern is the brain washing of the general public — and especially after the deep fakes become really sophisticated and nondecernable.

Below is a link from CNN that delves even deeper into this subject. They offer a test to challenge yourself if you can tell the difference between real audio/visual and fake A/V.

While I passed the test every time, 1.) I have inate, learned, and experienced abilities in visual comm; 2.) I knew it was a test, so I was looking for it; 3.) It's early, and the tech is only getting started — before too long, it will probably be too good for anyone to tell the difference.

Inside the Pentagon’s race against deepfake videos

Speaking of critical thinking skills, see if you can identify the bias CNN utilizes in this entire story. PM me if interested in a discussion.

 

Along these lines, Grant started a thread a few weeks ago about deep fakes, the craft of altering video to make a subject appear to be saying something they didn't by inserting different facial tics and syncing fresh audio into them.

But...suddenly I can't find it...should I be paranoid...?

 

Yes I couldn't find it either. But it referenced the same link I did, in my post #19, just proceeding yours.

 

i suck. i was 2 for 4

 

the more i look at these fakes the more i see the same set of teeth....tom cruises i think.....ive cracked to code

 

Rather than using photographic film (who uses film these days in the general population?), and taking single snapshots, which would be easy to doctor, I'd have taken a video, moving around the damaged area and 'filming' the surroundings for extra context. A video clip would be much harder to doctor than single stills (though still not impossible, but it would take a great deal of skill to do so convincingly).

 

The one I got looks familiar...