when i use an LLM to write some text out of a disordered stream of voice notes, the act feels easy.

there's barely any effort in making the disorder into order.

no friction from making something which kinda-but-doesn't-actually make sense into something-which-actually-makes-sense — literally, i've farmed out the job of meaning-making to a machine.

the result (this time, and many times before) is mediocre.

this is the subtle but profound problem of meaning-making in AI (specifically: that only humans can do meaning-making, but we're increasingly letting machines do it for us).

when i use an LLM to write some text out of a disordered stream of voice notes, the act feels easy.

there's barely any effort in making the disorder into order.

no friction from making something which kinda-but-doesn't-actually make sense into something-which-actually-makes-sense — literally, i've farmed out the job of meaning-making to a machine.

the result (this time, and many times before) is mediocre.

this is the subtle but profound problem of meaning-making in AI (specifically: that only humans can do meaning-making, but we're increasingly letting machines do it for us).

https://vaughntan.org/meaningmakingai