I Asked 5 Text-Only LLMs to Draw Objects They've Never Seen
gabriberton · x · 2026-10-26
A small experiment: the author asked 5 text-only LLMs (trained without vision) to "draw" simple objects purely in text, with no tools allowed, testing whether they can produce imagery from spatial knowledge learned in language training. The results are interesting.
Related event: Text-only LLMs can draw pixel art without any vision training(3 posts)→
More from Fun
- Retro sci-fi art created and rendered in Bryce3D back in 1998 still holds up — sebkrier · 2026-10-26
- Writer coins ALGO-OWN: complaining about the algorithm reveals your secret preferences — granawkins · 2026-10-26
- 'Opus 4.5 has tendrils that can feel how the plants are doing' — sensor-driven plant care — repligate · 2026-10-26
- Chart of 478M Polymarket trades shows prices are almost perfectly calibrated — NathanpmYoung · 2026-10-26
- Creator quits making GenAI content everywhere: "the harms exceed the benefits" — AIandDesign · 2026-10-26
- Anthropic interpretability demo shows neurons combining into tiny pirate/wizard personas — Sauers_ · 2026-10-26