Comment

gegil@sopuli.xyz ⁨23⁩ ⁨hours⁩ ago

Before ai there was photoshop. People who knew how to use it could make good fakes, but far from everyone could do that.

Then ai was created, it was fun to experiment with image generation, but it still was kind of a toy, and images was obviously generated.

Now ai became so advanced, that some ai “photos” or “artworks” require through examination to determine whether its real or fake.

If ai is so “good” right now, future generations of ai will make it nearly impossible to distinguish real pictures from generated, doesnt matter if its a photo, drawing, or other digital artwork.

source

Sort:hotnew top

MonkderVierte@lemmy.zip ⁨21⁩ ⁨hours⁩ ago
AI has plateaued.

source
- CannonFodder@lemmy.world ⁨21⁩ ⁨hours⁩ ago
  It’s certainly still improving in many areas - like coding assist, for sure. I think video generation is still improving but it takes a shit ton of resources, so it’s a bit slower. Maybe image generation has plateaued, but that could certainly be temporary. There are lots of more niche applications that are still progressing at a slower pace like I just read an article on using image processing AI in fertility clinics to better predict viability of embryos for implantation. Cool stuff that could actually improve people’s lives.
  
  source
  - hypnicjerk@lemmy.world ⁨16⁩ ⁨hours⁩ ago
    medical applications of machine learning are a pretty far cry from llms/genai
    
    source
    Hackworth@piefed.ca ⁨15⁩ ⁨hours⁩ ago
    Many medical applications of ML do use transformer architectures, so it’s fundamentally the same technology.
    
    source
- chunes@lemmy.world ⁨9⁩ ⁨hours⁩ ago
  Yeah, because it’s kind of harder to get better than photorealistic.
  
  source
- snooggums@piefed.world ⁨20⁩ ⁨hours⁩ ago
  And what a minuscule plateau it reached.
  
  source