ChatGPT Lied to Me and I Fell for It. Now I Feel Stupid.
It’s also pretty sketchy when you ask about it’s training data.
Like a lot of people, I’ve recently been experimenting with ChatGPT to see if I can find a shortcut to producing content.
(Unlike the doomsayers that are worried AI is going to take all of our jobs, I’ve actively been trying to train AI writers to replace me since way before ChatGPT.)
Here’s what happened
New to the whole ChatGPT thing, I had some assumptions.
I assumed you could give it a URL and ask it to summarize the information from the page you linked. And then write something using that information as the foundation.
In my defense, the first two times I tried doing this, ChatGPT literally just pretended to read what was on the page and generated some output.
It seemed reasonably accurate — keep in mind, I was actively trying to avoid having to read and understand what was on this website — so I pasted the output into a copy document as a rough draft and moved on.