Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F...
by throwway120385 on Wednesday March 15, 2023 @04:44PM
(#35174455)
The real problem is also that we have no way of knowing whether the ChatGPT answer is based on factual information. It is very skilled at inventing convincing facts from whole cloth.
[ Parent
| on HN ]
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F...
by kenjackson on Wednesday March 15, 2023 @04:50PM
(#35174532)
That is a problem with ChatGPT, but it has nothing to do with it being AGI or not. Politicians do this so regularly that we have websites like FactCheck. The question is can you tell that this is a computer making these facts up or do you think it is a human making up the facts.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F...
by slowmovintarget on Wednesday March 15, 2023 @06:09PM
(#35175418)
Politicians, generally speaking, know when they're making things up. ChatGPT and LLMs in general have no state evaluation on the text they've produced indicating its proximity to the truth.
This has everything to do with ChatGPT not being AGI.
Re:GPT4 and the Multi-Modal, Multi-Model, Multi-Everything F...
by kenjackson on Wednesday March 15, 2023 @06:21PM
(#35175525)
You are purely speculating about what a politician knows. I honestly don't even know what I know about me. I have some thoughts, but I'm not sure if those are just side effects of me processing or trying to maximize some function in my brain.
By your definition -- maybe we don't have (A)GI either.