https://aibussin.com/tags/similarity-metrics/Recent content in Similarity Metrics on AIBussin — AI applications, systems and booksHugoen-USWed, 05 Aug 2026 23:24:45 +0100https://aibussin.com/post/relate/Wed, 05 Aug 2026 23:24:45 +0100https://aibussin.com/post/relate/<p>Embeddings are everywhere in modern AI.</p> <p>They power semantic search, retrieval-augmented generation, recommendations, clustering, duplicate detection, code search, memory systems, and many of the mechanisms through which an AI system decides what information is relevant.</p> <p>Yet most systems interrogate embeddings in essentially the same way:</p> <blockquote> <p>Take two vectors and calculate cosine similarity.</p> </blockquote> <p>That is useful. But it also makes a strong assumption.</p> <p>It assumes that the information we care about is expressed directly through the default geometry of the embedding space.</p>