3 ms·
inference is the cheap part; training is expensive. what compute infrastructure would train this mythical magic model?
by milkshakes 2mo ago
inference is the cheap part; training is expensive. what compute infrastructure would train this mythical magic model?
- ElProlactin 2mo agoExactly. If OpenAI and Anthropic didn't have to train new models, they'd (probably) be instantly profitable and with good margins.