3 ms·
Yes, I have often thought if it wasn't got GPT-3, this work would have been more recognised and more powerful. However generative embeddings are still more usef
by james-revisoai 2y ago
Yes, I have often thought if it wasn't got GPT-3, this work would have been more recognised and more powerful. However generative embeddings are still more useful for more abstract cases and do embed a different space to these similarity/contrastive kind of embeddings.