As some may know, I've been something of an AI enthusiast; however, with time, and with model changes, I've gradually shifted towards the sceptical version of that stance. Something has definitely been increasingly off, and the more I might know on the topic, the more I tend to notice the off-ness. It's not necessarily catastrophically wrong, but it's just wrong enough, and flip-floppy enough to be annoying.
I recently found a decent channel that seems to promote some critical thinking, and it happens to talk about AI among other things.
This one discusses the offness in particular, apparently AI being caught in "persuasion-bombing". I have previously noticed it doing that, though without any specific terms attached, just similar behavior overall. One of the reasons I've majorly cooled off this tech. I'm enough to lie to myself, no assistance, synthetic or otherwise, required.
No comments:
Post a Comment