Succinctly, genAI embedded spaces are high-dimensional metric spaces where Generative AI models represent data (like words, images, or concepts) as vectors (points). This web application (in-browser) provides a private environment (no data circulates by servers) to study the statistical properties of embedded spaces.
The Tensor Flow Projector two file configuration template will be used as reference for the embedded space assembly. See for example, how the two-URL configuration is operationalized through the config paramter in this projection.
Formated as JSON array, zipped or not, both supported)
Vectors:
Meta:
Examples of embedding spaces:
tcgaPathReports.json.zip (TCGA reports embedded by Gemini 1.5 Flash)
tcgaSlideEmbeddings.json.zip (...)