Had a couple funny interactions with ChatGPT this week, including when asking it for some help with a coding task. It suggested some code that was really problematic and could easily break in the future. I told it I was worried about that code, and it replied with “You’re right, that was a hack.” 🤪

@manton interesting!
I've noticed GPT offering a lot of errors lately and when I correct it, it agrees and says "You're right!", and then offers something a little more useful.

@manton I had an instance where ChatGPT wrote some code that just wouldn’t work. I told ChatGPT it wouldn’t work, and asked it to “write better code.” So it did, but that code refused to compile with an error. I copied and pasted the error. It reported back and told me the first code it provided — the one that wouldn’t actually accomplish what I asked it to do — would prevent my error. That was true, but also we had just gone over why that code wouldn’t work.

That convinced me we have a ways to go before robots will replace us.