my genAI discourse filter is getting to the point where it's fairly useful with a good number of true positives. I've also cut a few frequent false positive terms from it recently.
because the aim is to filter Discourse™ without too many false positives on general nerd discussion of genAI, it's interesting to see that this filter seems to be evolving to depend more on matching phrases and more on exact matches, as compared to my politics filter which seems to have a higher proportion of single words and substring/word-stems specified.
1
