They are likely implemented very differently. I’m not certain but I imagine the current photos app uses an image model to detect and label objects which you can

They are likely implemented very differently. I’m not certain but I imagine the current photos app uses an image model to detect and label objects which you can search against. I expect Semantic Index (by virtue of the name) to be a vector store of embeddings.