October 7, 2022
A nice juicy paragraph from an AI/ML research engineer on Reddit:
Sometimes I wonder what the original pioneers of AI – Turing, Neumann, McCarthy, etc. – would think if they could see the state of AI that we’ve gotten ourselves into. [The PaLM Paper from Google is] 67 authors, 83 pages, 540B parameters in a model, the internals of which no one can say they comprehend with a straight face, 6144 TPUs in a commercial lab that no one has access to, on a rig that no one can afford, trained on a volume of data that a human couldn’t process in a lifetime, 1 page on ethics with the same ideas that have been rehashed over and over elsewhere with no attempt at a solution – bias, racism, malicious use, etc. – for purposes that who asked for?
Not to ignore any ethical or moral implications of all this, but I’m actually kinda hyped on AI tech. I think its applications of it can be clearly useful. Not just the stuff in my own bubble, like image generation, coding help, and automatic transcriptions, but things that I’m only now seeing all the possibilities of. I am asking, actually, that medical diagnoses come quickly and more accurately. I am asking, actually, that travel becomes safer. I am asking, actually, that spam and fraud become trivial to detect accurately and punish.
Related
🤘
ncG1vNJzZmibmKe2tK%2FOsqCeql6jsrV7kWlpa2dhZXxxg46fpqtloKq%2FsbvSnqpmrJiWwW7Dx6hkmqubmrFuss6rZg%3D%3D