/jp/ - Otaku Culture

>>46643285
have you tried reading his reply? that's him coping. he does that every single time he get something wrong (aka everytime). and it's also ironic because in the end llama3 turned out good because they've trained it 15t token, which is basically openai's phylosophy of simply throwing in more compute, kek

>>46643270
the point is that ceo knows better that anyone else what kind of gpu/tpus will be avaiable in the next 5 years. obviously he's biased, but he's definitely more accurate that the "father of machine learning" who invented neural nets 50 years ago, and then after 1 chat with chatgpt decided it's conscious