3 ms·
The issue is the training data isn’t some constant. Let’s suppose OpenAI had 10x the computing power but a vastly worse dataset, do you expect a better or wors
by Retric 2y ago
The issue is the training data isn’t some constant. Let’s suppose OpenAI had 10x the computing power but a vastly worse dataset, do you expect a better or worse result?
The question is ambiguous without defining how much worse the dataset is.
- golol 2y agoBut why would the dataset be worse, they can just use the same one as before.
- Retric 2y agoA dataset that’s 2 years old is worse than one that’s 20 years old even if it contains the same data. Even facts age. In 2004, Pluto was still classified as a planet. Not such a big deal on its own, but stale data gets a little bit worse every day.