It is very early days in understanding what the dimensions (i.e., the individual numerical elements) of embeddings represent. For the most part, the tendency is to consume the embeddings whole, in the ways that these early demonstrations presented in this section show. However, my view is that there is psychology in these individual numbers and we need to unpack them to better understand AI psychometrics. The aim for this page will be to share early ideas on how we might go about this.
Next page
Last page
Return home