A short and conclusive experiment with ChatGPT
AI isn't dangerous because it'll take over the world. It's dangerous because it extends the reach of human-generated propaganda
I decided to have a talk with ChatGPT. It was short and entertaining.
In addition to being unduly flattering, this is obviously not even close.
What about questions with arguably more consequential answers?
We* can check other parts of the Internet for verification. Among other findings that problematize ChatGPT’s confident and gushing explanation, there’s this:
Damn. Well, I know important and earnest people are ever seeking to improve the information environment so that our democracy remains safe.
Hi Kevin, This is a 2023 article, so it’s not in the scope of ChatGPTs training. It lists three limitations when accessing, one of which is, “Limited knowledge of world and events after 2021.” I hated it at first but am finding use cases. It’s not for emerging science.
I asked ChatGPT if transwomen are women. The reply was not promising, but also could have been worse.