4 ms·
LLMs can't really "see", so I challenge you to draw a pelican on a bike without any visual feedback, just code. Because that is how they are doing it. Vision t
by WarmWash 5mo ago
LLMs can't really "see", so I challenge you to draw a pelican on a bike without any visual feedback, just code. Because that is how they are doing it.
Vision tokens for transformers aren't really well solved yet, which is why they can smash a phd math problem and trip over a "count the cats on the chair" problem.
- raffael_de 5mo agoIt's not about seeing. It's about identifying the legs of the Pelican and then transferring the concept and mechanics of riding a bicycle + geometry of a body and a bicycle. The entire task has also nothing to do with vision tokens.
- WarmWash 5mo ago> It's about identifying the legs So, seeing?
- raffael_de 5mo agoseeing isn't necessary to understand what a leg is.
- WarmWash 5mo agoWhich I why humans can draw so well with their eyes closed?
- mountainriver 5mo agoIf we want to train a model excessively on SVGs it will obviously be able to do this. We have only just started trying to do that