Conversation

Replying to
This stuff has the same problem generated text and generated deep dream images had before… if you pay attention, it is incoherent but potentially there’s a portfolio theory effect here, where an image+text dataset has more coherence than sum of individual coherences
1
2
I think we have to teach these AIs to throw away bits not add them Aim for low-fi cartoons not pseudo-Dali paintings
Quote Tweet
I’m already bored of frustratingly vague but visually rich dalle2 stuff…want more precise impoverished stuff Can someone with access try to generate comics in xkcd style? Like “joke about sidewalk scooters in xkcd style” Or anything with very few bits
Show this thread
2
9
Once, many moons ago I tried to write a really naive program that would try to generate recognizable caricatures from photos using a strict budget of pen strokes (say 10 Bézier curves) My non-AI approach was too dumb and too early and still too hard for me. Bet it’s solvable now
2
6