avatar
Sherlock Holmes-type guy @tokenize.bsky.social

In your opinion, where might a boundary lie? If multi-modal LLMs can build/prune a 'dictionary' of 'concepts,' does that sidestep the Thai Library conjecture? Not arguing that a model expanding a collection of feature vectors == human learning, but curious where you see that differing from grounding

aug 13, 2025, 10:28 pm • 0 0

Replies

avatar
Fernando Cucchietti @fercook.bsky.social

Mmmm sorry but I’m not a fan of boundaries, specially artificial ones drawn over a very vague landscape without well defined metrics.

aug 13, 2025, 10:50 pm • 0 0 • view