OpenAI is training their AI to detect whether a photo is real or not

[asandax]

"A watermark you can't remove."
Uses another AI to remove the watermark.

[bigzigtv]

The detection system is secretly 1000 workers in india just like the amazon store

[ShadowLynx]

AI be like: Black and white images must be real, those existed before I did.

[realquadmoo]

Yeah I keep telling people to not use the AI detection tool in schools and stuff cause there’s physically no way for the AI to know if it made it or not.

[lolwhatevenisgoingon]

This feels like talking to a patient with dementia idk i guess it's the smile

[rutger]

People made a machine cleverly, and now people are mistaken it for a clever machine.

[cameroneridan]

and that's why when people claim that AI is aware of anything whatsoever I just laugh

[grim]

Bruh that smile after the “I think I’d remember that” 💀

[TreesPlease]

The watermark should be more like a fingerprint woven into the image. They do this already with printers, placing yellow dots on tbe page with a code specific to that printer

[prathameshkale]

Photo-changes hue.
Ai-I have never seen this picture in my life

[FrostyFrostySnow]

Just like ChatGPT when it gives me totally real and not at all fake references

[climatechange]

The detection system is secretly people on Facebook

[GSBarlev]

What the discriminator is saying isn't "This is real"—because how would it know it wasn't made whole-cloth in Photoshop—but "I would not have made this."

This is a big fail in communication on OpenAI's part.

[CasualTS]

I like how disappointed she sounds every time 😊

[AMO]

By next year every A.I. will have a watermark problem because they used the A.I. made stuff for data.

[beccalife]

If I didn’t make it, it must be real!

[jimreem]

Need a collab with these folks and the “It goes in the SQUARE hole!”

[rus.t]

Flexing the Youtube Shorts budget there with multiple full page colored prints

[hezowo]

This channel is really incredibly insightful and explaining current issues in such a spot on manner I am flabbergasted everytime I see a short. I can imagine this isnt what you do, but would I love to talk to you people and understand how you get these insights

[TheNewton]

The worst part is people have already been treating "AI" analysis or other algorithms as truth with consequences when it's no better than random.
Such as for false-positive in plagiarism detection getting students expelled, or why microsoft had to ban police departments from using "AI" for facial recognition.