If no distance metric is specified, what is the default distance used by the VECTOR_DISTANCE function?

The default distance used by the VECTOR_DISTANCE function is indeed the cosine distance. Understanding why this is the case involves recognizing how cosine distance operates in the context of vector space models, particularly in applications related to AI and machine learning.

Cosine distance measures the cosine of the angle between two vectors, which effectively captures the orientation of the vectors regardless of their magnitude. This is particularly useful in high-dimensional spaces often associated with vector embeddings, such as those used for NLP (natural language processing) tasks, where the direction of the vector can convey significant semantic similarity.

In many AI applications, especially those involving unstructured data and text, using cosine distance helps ensure that the similarity calculations focus on the relationships between the items represented as vectors, rather than being influenced by their sizes or magnitudes. This makes it a favorable choice as a default distance metric when none is explicitly specified.

Additionally, while other distance metrics such as Euclidean, Hamming, and Manhattan have their own applications and are useful in different contexts, they do not universally cater to the needs of vector similarity comparisons in the same way that cosine distance does. Therefore, the choice of cosine distance as the default reflects a consideration for common use cases in AI vector searches.