Stephen Downes

This was a fun read, as  considers the implications of Gary Lupyan and Blaise Agüera y Arcas's paper The unreasonable effectiveness of pattern matching in which they describe "an astonishing ability of large language models (LLMs) to make sense of 'Jabberwocky' language in which most or all content words have been randomly replaced by nonsense strings." The upshot (in my own words) is that LLMs are finding indicators of meaning we didn't know were there (but probably used intuitively to understand sentences). "You can write garbage, and the structure of the garbage will still infer a kind of meaning."