Using text embeddings
Text embeddings are a critically important part of generative AI solutions. Yet embeddings are harder to understand because they do not translate easily to the ways we interact regularly.
Put simply, text embeddings are an array of numbers that are calculated from some string of text. These numbers represent the location of a specific concept using a complex, multidimensional way of representing concepts. To make things clearer, let’s use a metaphor.
If two people are hiking in a large park and want to meet at a specific location, they could share GPS coordinates of the exact location. These GPS coordinates are an array of two numeric values representing a location somewhere in the world.
Alternatively, if those two people were referring to some concept in a book, they could share the name of the book, the chapter, the page, the paragraph, the sentence, and optionally, a word within that sentence. Those pieces of information are effectively a slightly...
