✳flâneur — a map of the web's best reading

Is OpenAI OK?, by Gwern · Gwern.net

gwern.net · 1,675 words · saved by 1 readers

Apropos of Ilya Sutskever⁠ now having left OA for good⁠, I’ve gone back to my thinking about the long-term consequences of Sam Altman’s coup, and something I began to wonder in 2021 when the news about the Anthropic defections broke: what if “the elves have left Middle Earth”? What if OA has lost its mojo? If so, what would that look like, and how would we know? What’s the “rot” counter-narrative? Maybe I am a bit too easily impressed by how good gpt-4-base and ⁠Claude-3.5-sonnet⁠ are at poetry compared to the mode-collapsed ChatGPTs, but I can’t help but wonder if the OA magic has worn off since GPT-4⁠ finished training c. August 2022. ⁠What made OA, OA, back in 2020⁠ was that it had taste: it had much less resources than competitors like DeepMind⁠ or Google Brain or FAIR, but (thanks to Alec Radford⁠, Ilya Sutskever, Jared Kaplan, and the RLHF⁠-focused safety team like Paul Christiano⁠ & Dario Amodei⁠, and fellow-traveler scalers like Andrej Karpathy⁠ etc) they bet big on ⁠scaling la

--- title: 'Is OpenAI OK?' author: Gwern description: "Has OA lost its mojo and ability to push AI paradigms in the wake of its hypergrowth and the OA Altman coup? Is OA OK? How would we know & what would it look like if it was not?" created: 2024-06-19 modified: 2024-12-04 status: finished importance: N/A confidence: log css-extension: dropcaps-de-zs toc-not backlink: False ... Apropos of [Ilya Sutskever](https://en.wikipedia.org/wiki/Ilya_Sutskever) [now having left OA for good](https://www.bloomberg.com/news/articles/2024-06-19/openai-co-founder-plans-new-ai-focused-research-lab "‘Ilya Suts

Explore this link on the map →

saved by

related reading