On 2025-05-09 at 13:21:41 UTC-0400 (Fri, 9 May 2025 13:21:41 -0400) Karl Denninger <k...@denninger.net> is rumored to have said:
> Bayes seems to get them most of the time, if you train it well. As more and > more of this is AI-generated the "catch" rate goes up incidentally, as AI is, > well, certainly artificial but not very intelligent :-) It is also worth noting that Bayesian spam filtering is conceptually related to what LLM "AI" tools do. They assemble texts word-by-word based on the likelihood of each one being next based on the preceding series of $bignum words. They write artificial banality that is ideal for Bayesian analysis. -- Bill Cole b...@scconsult.com or billc...@apache.org (AKA @grumpybozo@toad.social and many *@billmail.scconsult.com addresses) Not Currently Available For Hire
signature.asc
Description: OpenPGP digital signature
