PLAY PODCASTS
“Estimating the Substitutability between Compute and Cognitive Labor in AI Research” by Parker_Whitfill, CherylWu

“Estimating the Substitutability between Compute and Cognitive Labor in AI Research” by Parker_Whitfill, CherylWu

EA Forum Podcast (Curated & popular) · EA Forum Team

June 7, 202520m 25s

Audio is streamed directly from the publisher (dl.type3.audio) as published in their RSS feed. Play Podcasts does not host this file. Rights-holders can request removal through the copyright & takedown page.

Show Notes

<p> <p> Audio note: this article contains 127 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description.</p></p> <p> Confidence: Medium, underlying data is patchy and relies on a good amount of guesswork, data work involved a fair amount of vibecoding. </p><p><strong> Intro:&nbsp;</strong></p><p> Tom Davidson has an excellent post explaining the compute bottleneck objection to the software-only intelligence explosion.[1] The rough idea is that AI research requires two inputs: cognitive labor and research compute. If these two inputs are gross complements, then even if there is recursive self-improvement in the amount of cognitive labor directed towards AI research, this process will fizzle as you get bottlenecked by the amount of research compute. </p><p> The compute bottleneck objection to the software-only intelligence explosion crucially relies on compute and cognitive labor being gross complements; however, this fact is not [...]</p> <p>---</p><p><strong>Outline:</strong></p><p>(00:35) Intro:</p><p>(02:16) Model</p><p>(02:19) Baseline CES in Compute</p><p>(04:07) Conditions for a Software-Only Intelligence Explosion</p><p>(07:39) Deriving the Estimation Equation</p><p>(09:31) Alternative CES Formulation in Frontier Experiments</p><p>(10:59) Estimation</p><p>(11:02) Data</p><p>(15:02) Trends</p><p>(15:58) Estimation Results</p><p>(18:52) Results</p> <p><i>The original text contained 13 footnotes which were omitted from this narration.</i> </p><p>---</p> <p><b>First published:</b><br/> June 1st, 2025 </p> <p><b>Source:</b><br/> <a href="https://forum.effectivealtruism.org/posts/xoX936hEvpxToeuLw/estimating-the-substitutability-between-compute-and?utm_source=TYPE_III_AUDIO&utm_medium=Podcast&utm_content=Source+URL+in+episode+description&utm_campaign=ai_narration" rel="noopener noreferrer" target="_blank">https://forum.effectivealtruism.org/posts/xoX936hEvpxToeuLw/estimating-the-substitutability-between-compute-and</a> </p> <p>---</p> <p>Narrated by <a href="https://type3.audio/?utm_source=TYPE_III_AUDIO&utm_medium=Podcast&utm_content=Narrated+by+TYPE+III+AUDIO&utm_term=ea_forum&utm_campaign=ai_narration" rel="noopener noreferrer" target="_blank">TYPE III AUDIO</a>.</p> <p>---</p><div style="max-width: 100%";><p><strong>Images from the article:</strong></p><a href="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/mzupvox8qtp6wf6msxag" target="_blank"><img src="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/mzupvox8qtp6wf6msxag" alt="Scatter plot titled "CES in Compute" showing data from Anthropic, DeepMind, DeepSeek, and OpenAI." style="max-width: 100%;" /></a><hr style="margin-top: 24px; margin-bottom: 24px;" /><a href="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/exjrbkmr7fkba2ebrccg" target="_blank"><img src="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/exjrbkmr7fkba2ebrccg" alt="Line graph titled "Average Wages by Organization Over Time" tracking four AI companies." style="max-width: 100%;" /></a><hr style="margin-top: 24px; margin-bottom: 24px;" /><a href="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/p7gflo3wb5i00aqztipf" target="_blank"><img src="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/p7gflo3wb5i00aqztipf" alt="Line graph: "Research Compute per Employee Over Time" comparing four AI companies' trajectories." style="max-width: 100%;" /></a><hr style="margin-top: 24px; margin-bottom: 24px;" /><a href="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/mffnvj7amuulcchngzwu" target="_blank"><img src="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/mffnvj7amuulcchngzwu" alt="Scatter plot titled "CES in Frontier Experiments" showing data from four tech companies over time." style="max-width: 100%;" /></a><hr style="margin-top: 24px; margin-bottom: 24px;" /><a href="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/ktvnjhx4fbsxgtiywt5d" target="_blank"><img src="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/ktvnjhx4fbsxgtiywt5d" alt="Line graph "Organization Size Over Time" showing employee growth at AI companies." style="max-width: 100%;" /></a><hr style="margin-top: 24px; margin-bottom: 24px;" /><a href="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/jcflrejd7nfq3yx8sskr" target="_blank"><img src="https://res.cloudinary.com/cea/image/upload/f_auto,q_auto/v1/mirroredImages/xoX936hEvpxToeuLw/jcflrejd7nfq3yx8sskr" alt="Line graph: "Compute Price per PFLOP Over Time by Organization" showing four companies' trends." style="max-width: 100%;" /></a><p><em>Apple Podcasts and Spotify do not show images in the episode description. Try <a href="https://pocketcasts.com/" target="_blank" rel="noreferrer">Pocket Casts</a>, or another podcast app.</em></p></div>