this post was submitted on 28 Jun 2024
396 points (100.0% liked)

196

16504 readers
3756 users here now

Be sure to follow the rule before you head out.

Rule: You must post before you leave.

^other^ ^rules^

founded 1 year ago
MODERATORS
 
you are viewing a single comment's thread
view the rest of the comments
[–] owenfromcanada@lemmy.world 71 points 4 months ago (5 children)

Not sure why he lost. He could have claimed to run it through an AI that was only trained on the one picture.

[–] LazaroFilm@lemmy.world 45 points 4 months ago

“I will give a picture, you will create an exact copy, accurate pixel to pixel to the one I give you”.

[–] MindTraveller@lemmy.ca 24 points 4 months ago

He got disqualified on purpose in order to make a point

[–] Daxtron2@startrek.website 13 points 4 months ago

you don't even need to train anything, just run it through on a super low blur and almost nothing will change.

[–] Mixel@feddit.de 8 points 4 months ago (2 children)

To train a diffusion model that only outputs one image with difference is I think not possible you could do an image to image and then fix the seed so you would get a consistent result and then picking the nearest result that is nearly an identical copy

[–] themoonisacheese@sh.itjust.works 15 points 4 months ago

It very much is possible. It'll only ever output that one image and it's a huge waste of resources but it's very much possible.

[–] Hjalamanger@feddit.nu 9 points 4 months ago

It's very much possible and indeed such a problem that it may be done by mistake if a large enough data set isn't used (see overfitting). A model trained to output just this one image will learn to do so and over time it should learn to do it with 100% accuracy. The model would simply learn to ignore whatever arbitrary inputs you've given it

[–] Hjalamanger@feddit.nu 6 points 4 months ago

That probably doesn't count as "AI" do... It's more a very bad form of compression (that may very well make the image file larger)