... would be really cool.
AI (read: ChatGPT) is really just a sophisticated parrot for its training set. I went around and around and around with it again today, working out social norms for various eras and settings. While we sort of came to a conclusion, it didn't take long for it to drive off into a progressive ditch again. Sigh.
Anywho, imagine an AI whose training set had nothing published after 1912. It would be splendid!
 |
| "Think of it, gentlemen. No more endless twaddle from that infernal AI!" |
That is really something that they should be doing anyway, for legal reasons. A huge amount of the opposition to ai is that they can be considered to be plagiarism engines. But if all of their training data was old enough to be in the public domain, most of the legal issues evaporate.
ReplyDeleteWouldn't it be great to have a drop down selector in ChatGPT where you could dial in the time frame for the training set?
ReplyDeleteAdmittedly, 4000 BC might be a trifle reductive.
Me: "What is the circumference of the Earth?"
AI: "Org no like big word!"