All sorts of strange issues today. As well as all the deploy problems, I had some really quite bizare interactions with GPTs today. I have previously had some good experiences working on CSS with AIs, but today was with javascript, and it was really quite terrible. I was trying to do something very simple, and the AI was repeatedly failing, and adding complexity while doing so. So I asked it to remove the styling that it had added without asking me, and it said yep removed it. And it hadn't. This went on for at least 30 mins. Over and over it would say that it had removed it, but then it still hadn't, and it would even check itself and say, oh yes I didn't remove it, then still fail to remove it.
One thing that hadn't occurred to me before, and I can't believe I hadn't thought if this, and I also can't believe I haven't heard anyone talking about this yet. How do we know that the GPT companies aren't just barely training their models, then releasing them to billions of users half baked, so that the users train them for free without knowing it. Surely this is bound to happen as VC money dries up. #