⟔□⊏⟔⪽⊐⊡⅃Ŀ⊏
banner
refrodnalf.bsky.social
⟔□⊏⟔⪽⊐⊡⅃Ŀ⊏
@refrodnalf.bsky.social
and it will always be exhausted, research has shown that exponentially more data is needed for linear improvements youtu.be/dDUC-LqVrPU?...
in other words, every time they want a tiny increase in model performance the training data needs to double in size
Has Generative AI Already Peaked? - Computerphile
YouTube video by Computerphile
youtu.be
January 19, 2026 at 7:40 PM
this research paper from 2023 should be similarly interesting. i would go as far as to say the only thing modern “ai” companies have brought to the table to forward the tech is their disregard for copyright law and morality. neural networks have been around since the 60s arxiv.org/abs/2301.13188
Extracting Training Data from Diffusion Models
Image diffusion models such as DALL-E 2, Imagen, and Stable Diffusion have attracted significant attention due to their ability to generate high-quality synthetic images. In this work, we show that di...
arxiv.org
January 19, 2026 at 7:35 PM
Reposted by ⟔□⊏⟔⪽⊐⊡⅃Ŀ⊏
people have been misled by years of "computer, enhance" in movies and tv and think that technology actually exists. there is sometimes data in photos you can't see that can be pulled out by someone who knows what they are doing, but that's not what "ai" does
January 8, 2026 at 1:48 AM
Reposted by ⟔□⊏⟔⪽⊐⊡⅃Ŀ⊏
This part has really stuck with me:

www.thehandbasket.co/p/kavanaugh-...
January 7, 2026 at 1:19 PM
if the reprint shares the isbn of the original (not uncommon) i can see the algorithm thinking it is an exceedingly rare book
January 6, 2026 at 6:26 AM
many resellers price things with an algorithm that weighs scarcity heavily. i once googled the isbn of a $10 journal from barnes and noble, and it was $300 on amazon (since naturally it was specifically scarce on amazon since it would only be sold there 2nd hand if at all)
January 6, 2026 at 6:24 AM
most cars are intentionally designed to resemble faces, not weird at all
December 24, 2025 at 3:08 PM