3 ms·
The only reason I can see for this is to create lock-in. I'd be pretty surprised if anymore than 5% of their customers would want a model by pinecone.
by mritchie712 2y ago
The only reason I can see for this is to create lock-in. I'd be pretty surprised if anymore than 5% of their customers would want a model by pinecone.
- gdj0nes 2y agoWe trained a sparse embedding model because there is a lack of commercially licensed weights or APIs. For any proprietary model, particularly embeddings, there is a degree of lock-in.