Large Language Models Cannot Explain Themselves

Advait Sarkar

Large language models can be prompted to produce text. They can also be prompted to produce "explanations" of their output. But these are not really explanations, because they do n…