3 ms·
I think this would be cost prohibitive, though maybe not, however - it could just save every answer its given and scan text for it. If they're a match it could
by no_wizard 3y ago
I think this would be cost prohibitive, though maybe not, however -
it could just save every answer its given and scan text for it. If they're a match it could just not index it, right?
- data_maan 3y agothere's still the pre-2023 data to train it on ... and then augment with handpicked stuff.
- kikokikokiko 3y agoThe dataset of internet content pre-2022 will be regarded in the future just like low-background steel. Any content generated post the release of the first generally accessible LLMs will be considered radioactive.
- digging 3y agoDoesn't work if more than 1 bot exists...