@conitzer finally I did an experiment for myself and my conclusion is that those transformer networks lack the ability to detect when they just should say “I don’t know”. Without that, they will become creators of #FakeKnowledge.

https://fediscience.org/@UlrichJunker/109867787002009725

Due to this severe incapability to recognize their own #limits systems such as #ChatGPT may become a source of #FakeKnowledge and thus increase the already existing #disinformation in social media. It will be important to build #reliablesources of information and to require those systems to cite reliable sources.