3 ms·
I can easily imagine that multimodal GPT-style LMs running in real-time could do all the common sense thinking that seems to be lacking in modern self driving.
by generalizations 3y ago
I can easily imagine that multimodal GPT-style LMs running in real-time could do all the common sense thinking that seems to be lacking in modern self driving. That future could be just a 100x optimization away.
Edit: multimodal. Does no one remember the vision + language version of GPT4 that microsoft demoed?
- l33t233372 3y agoWhy do you feel this is a language problem?
- serf 3y agoreal-time LMs that can be reasonably allowed to take part in life-or-death situations feels a lot further away (to me) than a 100x optimization.
- weakfish 3y agoLM = language model I fail to see how that is relevant to computer vision