how many three-window rooms can one AI make
https://www.youtube.com/watch?v=khysGsyK9Qo&list=UU9rJrMVgcXTfa8xuMnbhAEA - video
https://pivottoai.libsyn.com/20251222-ai-image-generators-have-just-12-templates - podcast
time: 6 min 46 sec
AI image generators have just 12 generic templates
https://pivot-to-ai.com/2025/12/22/ai-image-generators-have-just-12-generic-templates/
15 Comments
blakestacey@awful.systems · 14 pts · 247d
Revealed: World's shittiest "tag yourself" meme
swlabr@awful.systems · 8 pts · 247d
sigh. 10 sun, 7 moon, 8 rising
dgerard@awful.systems · 5 pts · 247d
honestly, whomst amongst us isn't Pompous Interior Design
V0ldek@awful.systems · 4 pts · 244d
domestic scenes and food imagery (sitting on my ass at the PC ingesting industrial amounts of crisps)
Soyweiser@awful.systems · 13 pts · 247d
Talked to somebody who is really into chatbot roleplay (of the 'longer term stories with new fantasy characters' type), and he mentioned that he needs to take his characters stories and archetypes to different models every now and then as a sort of refresh, as the models tend to eventually converge into certain stuck patterns. First clue of this seems to be that the replies seem to start to become a similar pattern of text organization. Sorry if this is vague as it is second hand, but the main point is, text based LLMs prob also do this.
dgerard@awful.systems · 8 pts · 247d
oh yeah, Suno does the same, it has about 12 songs
flaviat@awful.systems · 12 pts · 247d
clanker's dozen
Soyweiser@awful.systems · 8 pts · 247d
Wonder if this is some sort of pre model collapse sign.
corbin@awful.systems · 5 pts · 243d
Nah, it's more to do with stationary distributions. Most tokens tend to move towards it; only very surprising tokens can move away. (Insert physics metaphor here.) Most LLM architectures are Markov, so once they get near that distribution they cannot escape on their own. There can easily be hundreds of thousands of orbits near the stationary distribution, each fixated on a simple token sequence and unable to deviate. Moreover, since most LLM architectures have some sort of meta-learning (e.g. attention) they can simulate situations where part of a simulation can get stuck while the rest of it continues, e.g. only one chat participant is stationary and the others are not.
Soyweiser@awful.systems · 2 pts · 243d
Thanks!
pikesley@mastodon.me.uk · 4 pts · 247d
@dgerard @Soyweiser the Randy Newman record?
fullsquare@awful.systems · 7 pts · 247d
so after putting together text to image and image to text idiot boxes, there appears to be small number of approximate sort of eigenvalues in there. does that even mean anything or has any consequences?
blakestacey@awful.systems · 12 pts · 247d
Eigenslop
dgerard@awful.systems · 10 pts · 247d
as i said this is a completely unsurprising result, but it's amusing to know what the twelve templates actually are
I got an email from the author, he says the paper was a passing observation and he's surprised it's got as much attention as it has
Jayjader@jlai.lu · 3 pts · 246d
On the nth day of Christmas, my true love gave to meeeee–
An LLM in a pear tree?
eleijeep@piefed.social · 0 pts · 247d