This comment reminds me of an xkcd comic from 12 years ago[0] in which two characters are discussing AI, and one has built a system to defeat spam comments by a

This comment reminds me of an xkcd comic from 12 years ago[0] in which two characters are discussing AI, and one has built a system to defeat spam comments by asking users to rate other comments as "Constructive" or "Not constructive". Their end goal, as it turns out, is to force the bots to make automated constructive comments. At the time, the idea that a bot could even do that in response to any comment was pretty far-fetched. Nowadays, the flaw is obvious: How do you test that the comment is constructive? Of course, we could train an AI - even ChatGPT - to do that (An example prompt could be "Please look at the following comment posted on a social media site. If the comment is constructive, say 'constructive'. If the comment is not constructive, say 'not constructive'. If you aren't sure, say 'unknown'. Do not give any other responses."), but at that point you're kinda fighting fire with fire. [0] https://xkcd.com/810/ https://xkcd.com/810/ [edit: ...how do I report a bot reply to my comment as ~~not constructive~~ spam? I genuinely cannot find any link to do so.]