4 ms·
My guess is that using the OpenAI models to generate training data for a bespoke transformer model is extremely taxing on OpenAI's computing usage. If I were to
by baryphonic 3y ago
My guess is that using the OpenAI models to generate training data for a bespoke transformer model is extremely taxing on OpenAI's computing usage. If I were to guess, that is why that behavior is proscribed by the TOS, and why ByteDance was banned. Probably has nothing to do with the ethics of how training data is gathered.