Researcher's parody: LLM misalignment damages 'scaling log-linearly', per GPT-made plot

joshua_saxe · x · 2026-09-04

AI safety researcher Joshua Saxe posted a tongue-in-cheek chart claiming "LLM misalignment damages appear to be scaling log-linearly over time" — then revealed the punchline: the plot was made by having a "GPT-5.6 Sol / Ultra" model pull news reports and estimate the damages itself. It's a meme poking at AI doom narratives and circular research methods, where even the evidence of rising misalignment is model-generated.

Original post →

More from Fun

Fun channel →