ImageExplore

Embeddings are an way of using large language models to analyze images. Instead of asking the LLM to describe the image in text, we extract an array of numbers. These numbers are similar for similar images.

This has 2 advantanges:

1. It's quantitative — we can measure how similar images are
2. It's cheap — text generation is slower and more expensive

Pick from one of the demos below. Or, if you're authorized, log in and explore with your own images.