3 ms·
We need to give LLMs a stream of inputs and outputs that are replaced over time
- Marshferm 11mo agoYou have to discard 90-97% of machine vision, which has little to do with mammalian vision
- tokenomics 11mo agoThe point is less about vision, and more about needing a continious stream of inputs and outputs.
- Marshferm 11mo agoThat’s not what human vision is.