Are LLMs Just Amplified Autocomplete?

ineshipolito · x · 2026-07-11

The author compares LLMs to an amplified version of "keyboard autocomplete": keyboards only predict the next word based on frequent input, yet we say LLMs "judge, decide, know."

The core question: if we don't say keyboards "think," why attribute stronger mental abilities to LLMs? The author argues that once misattribution is established, it requires "retraction/correction," but there is no equally rigorous correction practice for autocomplete-style systems.

Related event: LLMs Framed as Scaled-Up Autocomplete(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →