this post was submitted on 29 Jan 2025
43 points (93.9% liked)

Asklemmy

44624 readers
903 users here now

A loosely moderated place to ask open-ended questions

Search asklemmy ๐Ÿ”

If your post meets the following criteria, it's welcome here!

  1. Open-ended question
  2. Not offensive: at this point, we do not have the bandwidth to moderate overtly political discussions. Assume best intent and be excellent to each other.
  3. Not regarding using or support for Lemmy: context, see the list of support communities and tools for finding communities below
  4. Not ad nauseam inducing: please make sure it is a question that would be new to most members
  5. An actual topic of discussion

Looking for support?

Looking for a community?

~Icon~ ~by~ ~@Double_[email protected]~

founded 5 years ago
MODERATORS
 

What are your thoughts on Generative Machine Learning models? Do you like them? Why? What future do you see for this technology?

What about non-generative uses for these neural networks? Do you know of any field that could use such pattern recognition technology?

I want to get a feel for what are the general thoughts of Lemmy Users on this technology.

you are viewing a single comment's thread
view the rest of the comments
[โ€“] [email protected] 6 points 1 day ago

Actual piracy doesn't bother me, but I'm supposed to care that a robot learned English by reading library books? Learning is what libraries are for. Yeah, the draw-anything robot can only draw The Simpsons because it's seen The Simpsons. How else was it supposed to happen?

Training is transformative use. You can't spend a zillion compute-hours guessing the next word of a story, in such a way that it can fake Tolkien retelling Shrek as a rap battle, and claim that's the same as LordOfTheRings.txt on an FTP server. What the network is and does will not substitute the original work. Not unless the Silmarillion had more swamp ogres than I've heard.

Image stuff will become a brush that does whatever you tell it. Type the word "inks" and drag it over your sketch, and it'll smooth out your lines. Type the word "photorealistic" and it'll turn your blocky shading into unreasonably good lighting. None of this prevents human art. The more you put in, the more you get out. Stable Diffusion is a denoiser, where the concept of noise can be defined as bad anatomy.

Video stuff might end Hollywood, as soon as editors figure out they've inherited the Earth. The loosest animatics can become finished shots without opening Blender or picking up a camera. A static image of what a character looks like should be enough to say, this stick figure is that guy. Or this actor is that cartoon character. Or this cardboard cutout is that approaching spaceship. The parts that don't look like that are noise, and get removed. We're rapidly going to learn how blobby and blurry an input can be, for the machine to export a shot from your head, just the way you imagine it. And where it's not exactly what you intended - neither is any shot ever filmed. A film only exists in the edit. So anyone who can string together some already-spooky output, based on the stories they'd like to tell, is going to be a studio unto themselves.