6 ms·
The issue is far more serious with ChatGPT/similar models because things that are laughably untrue are delivered exactly the same as something that's solidly tr
by DharmaPolice 3y ago
The issue is far more serious with ChatGPT/similar models because things that are laughably untrue are delivered exactly the same as something that's solidly true. When doing a normal search I can make some assessment on the quality of the source and the likelihood the source is wrong.
People should be able "throw the ideas away if they turn out to be wrong" but the problem is these ideas unconsciously or not help build your model of the world. Once you find out something isn't true it's hard to unpick your mental model of the world.
- huijzer 3y ago> Once you find out something isn't true it's hard to unpick your mental model of the world. Intuitively, I would think the same, but a book about education research that I read and my own experience taught me that new information is surprisingly easy to unlearn. It’s probably because new information sits at the edges of your neural networks and do not yet provide a foundation for other knowledge. This will only happen if the knowledge stands the test of time (which is exactly how it should be according to Popper). If a counterexample is found, then the information can easily be discarded since it’s not foundational anyway and the brain learns the counterexample too (the brain is very good in remembering surprising things).
- DharmaPolice 3y agoThat presumes the wrong information is corrected quickly. What about the cases when that doesn't happen? Aren't you often finding out things you thought were true from years ago are wrong?
- PoignardAzur 3y agoYou weigh new information by how confident you are in it. You try to check different sources, you maintain an open-mind, etc. In that, ChatGPT is just an additional low-reliability source of information.