Abstract

I'll share some current thoughts on the problem of watermarking the outputs of LLMs: why this technology hasn't yet been deployed, and what needs to be done both technically and politically.