My Advice to Sam Altman: Read Jacques Derrida

> if we are to properly understand LLMs, we need to accept that linguistic creativity can be completely distinct from intelligence, and that text doesn’t have to refer to the physical world – just to other words. What these machines demonstrate is that language works as a system of signs that mostly refer just to other signs.

But does that preclude intelligence? Arguably all interaction with the world is through, or understood through, signs.

philosopher Jacques Derrida saw language as a self-contained system, which is a good description of what goes on inside an LLM. For the machine, words only refer to other words rather than having a relation to external reality.

This means that any sequence of words generated by an LLM is derived solely from its relationship to other words within the linguistic system, not from any “understanding”

Somebody with an archive of this article?