digital image is what makes machine vision possible -> is this true? it’s more seq2seq makes machine vision possible no? There’s a fetishization of the senses here, where we assume
attention is all you need -> seq2seq is all you need, functional values is a texture, not a vision. Why does this matter? You feel texture.
What are ways of building a new primiative?
eject this type of sense making, approach it as an anthropological problem, which means you have to spend time with the computer
however, just like how you have to approach cultures as a producer, not a consumer, you need to understand computers in the same way (aka go become a programmer)
the soul of a machine is not reflection, bryan cantrill does a good job instilling this ethos.