Discussion about this post

User's avatar
Mark's avatar

The part about not needing countless nvidia chips is not news... Its entering public conscienceness though. In 2022 oct 27, before AI blew up Raoul pal intervied Emad Mostaque of Stability. It was stated that once a llm model is trained, it can run on end user hardware. Mag 7 ai solutions have made it a cloud hosted service to plebs to mine or data and to fit their subscription $$$ model of doing business. The key questions are this: how long until we have trained a good enough LLM ? Nvidia depends on this. Will some company drop a fully trained open source easy to mod and deploy product? No corp implementing this wants their data being sold. This will destroy data scraping subscription based companies...

Expand full comment

No posts