I read that open ai has another code red after fable / mythos so they are trianing gpt 6 and launching it in a month from now, its going to be a 10T model. So model size is now back to being important. I dont get why tpus arent more valued, all semi stuff on twitter that i see say nvidia has the best hardware for ai. And semi analysis constantly says Nvidia is the inference king, but if model size are now more important, Nvidia might not be able to compete with tpus especially with recent semi analysis post that kyber is gonna delay again and wont now ship... (read more)
I read that open ai has another code red after fable / mythos so they are trianing gpt 6 and launching it in a month from now, its going to be a 10T model. So model size is now back to being important. I dont get why tpus arent more valued, all semi stuff on twitter that i see say nvidia has the best hardware for ai. And semi analysis constantly says Nvidia is the inference king, but if model size are now more important, Nvidia might not be able to compete with tpus especially with recent semi analysis post that kyber is gonna delay again and wont now ship... (read more)