Skip to the main content
icon-community

AI learns language from skewed sources. That could change how we humans speak – and think

Large language models aren’t trained on real-life conversations. As we encounter their language, it could affect our own

Affiliate Bruce Schneier and Ada Palmer point out that LLM training misses the bulk of verbal communication because they are only trained on written, rather than oral, text. "By excluding the overwhelming majority of language production on the planet – people talking, fully and naturally, to each other – these models are being trained to mirror everything but us at our most authentically human," the pair writes in The Guardian

You might also like