OpenAI Futurist Argues Physical Ceiling Caps AI Intelligence
Joshua Achiam cites computation density constraints in debate over recursive self-improvement and superintelligence.
OpenAI Chief Futurist Joshua Achiam stated intelligence growth faces a physical ceiling from limits on computation per unit volume and energy, with compute deciding winners. In replies to @MTSlive, he clarified the ceiling is high but absolute, without referencing superintelligence timelines. METR President Chris Painter questioned if this rebuts recursive self-improvement arguments. The exchange occurred on X, highlighting ongoing AI scaling debates among safety researchers. No further developments confirmed.
Combined views
3.1K
5 posts, first seen 5h ago
Reactions from ranked influencers
5 posts
OpenAI's Chief Futurist Joshua Achiam breaks with the consensus on takeoff, arguing intelligence hits a physical ceiling and then raw compute decides who wins: "People expect that there's no ceiling for the amount of intelligence that you could have in a model, and they think of recursive self-improvement as this loop that's gonna happen at some point or another. RSI starts happening and model intelligence takes off and it goes to the moon, and they don't see a ceiling." "I think on physical grounds, there's gotta be a maximum amount of computation that you can have per unit volume and energy in the physical universe. And that sort of implies there's a maximum amount of intelligence per unit volume and unit of energy." "If that's the case, eventually, seeing how fast AI model capabilities are increasing right now, eventually everyone hits that saturation point and everyone's got roughly equivalently capable models from a raw intelligence perspective." "The open source frontier lags the closed source frontier by some number of months, but the fact that it's months is crazy. So eventually everyone is probably working with equally maximally capable models, and then it's amount of compute that you're able to throw at a problem that determines who wins." @jachiam0
@ChrisPainterYup @MTSlive Didn't say "before superintelligence" or "below human level" anywhere; I think the ceiling is very high. Just think there is necessarily a maximum amount of computation possible per unit volume and energy.
Yes there’s a physical limit to intelligence density, both practical limiting you at any given time, as well as theoretical. But it also depends on how deep the algorithms go or don’t go. In all cases we should have a more Copernican view on how much progress we’ve made so far
OpenAI's Chief Futurist Joshua Achiam breaks with the consensus on takeoff, arguing intelligence hits a physical ceiling and then raw compute decides who wins: "People expect that there's no ceiling for the amount of intelligence that you could have in a model, and they think of recursive self-improvement as this loop that's gonna happen at some point or another. RSI starts happening and model intelligence takes off and it goes to the moon, and they don't see a ceiling." "I think on physical grounds, there's gotta be a maximum amount of computation that you can have per unit volume and energy in the physical universe. And that sort of implies there's a maximum amount of intelligence per unit volume and unit of energy." "If that's the case, eventually, seeing how fast AI model capabilities are increasing right now, eventually everyone hits that saturation point and everyone's got roughly equivalently capable models from a raw intelligence perspective." "The open source frontier lags the closed source frontier by some number of months, but the fact that it's months is crazy. So eventually everyone is probably working with equally maximally capable models, and then it's amount of compute that you're able to throw at a problem that determines who wins." @jachiam0
@ChrisPainterYup @MTSlive What? No, not at all. It's against the idea that there will be a single definitive winner in model quality. Everyone eventually saturates to equivalently powerful superintelligence.