AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Advantages of a Dense Vector Index
The data structures that you would use for a dense system where everything is clustered into like 200 and maybe 256 or 768 or whatever dimensions are very different than a sparse index. So it's not like cat occurs in three things, it's like all billion documents have some level of if cat was one of the dimension some level of catness. And so there's not some future where you throw away one or you throw away the other. More and more, the reality is like hybrid retrieval where you're using both data structures to serve search results.