LLMs Are Robot Arms, Not Brains
It is better to think of LLMs as a form of automation, not a form of intelligence. They are not intelligent. In the same way that a robot arm can do specific physical tasks in specific physical environments — sometimes at a level of precision and efficiency that a human cannot achieve — LLMs can perform specific mental tasks within a specific domain.
“But doesn’t it take intelligence to perform a mental task?” No it does not. Calculators are not intelligent, and an LLM is merely an advanced calculator that performs statistical analysis of language usage in an effort to mimic human language. It does not actually know the meaning of a word anymore than a robot arm on an assembly line understands what a quarter panel is.
LLMs have no conception of what language even is. They are not literate. To them, words, sentences, and paragraphs are mathematical agglomerations, not discrete and contextual units of meaning arranged to form ideas. LLMs are incapable of having ideas.
LLMs share other attributes with robot arms: they will continue to be refined, their usage will spread, and they will be used by employers to eliminate jobs. They will never take over the world or supersede human intelligence; they will simply be a means of production that, like robot arms, you probably don’t own. But their billionaire owners have already taken over the world, and LLMs are just another roadblock against you ever taking it back.
