4 ms·
Searching the transcript has the problem of missing synonyms. This can be solved by the one undeniably useful type of AI: embedding vector search. Embeddings fo
by mrob 10mo ago
Searching the transcript has the problem of missing synonyms. This can be solved by the one undeniably useful type of AI: embedding vector search. Embeddings for each line of the transcript can be calculated in advance and compared with the embeddings of the user's search. These models need only a few hundred million parameters for good results.
- andai 10mo agoYeah, but they fail surprisingly hard on grepping. So the best systems use both simultaneously: https://www.anthropic.com/engineering/contextual-retrieval https://www.anthropic.com/engineering/contextual-retrieval