'By that measure OpenAI is for open science — they published the GPT-3 paper once!'

anpaure · x · 2026-11-17

In the debate over AI labs blogging instead of publishing in journals, anpaure hit back at a 2021 counterexample: "one example from half a century in AI years disproving the general principle is quite scientific indeed."

He then delivered the punchline: "by that measure OpenAI is for open science, they published the GPT-3 paper once!" — a jab at how OpenAI has essentially stopped releasing research details since GPT-3.

Original post →

More from Fun

Fun channel →