Dean Ball Clarifies Open-Weight AI Slows Capital Spending Margins Following Controversial Analysis On Chinese Models
Dean Ball, who joined OpenAI less than two weeks ago for his first technology industry role, corrected his prior analysis of the Chinese AI model Kimi and broader open-weight artificial intelligence. He clarified that calling open-weight models unqualifiedly decelerationist was imprecise, noting they only slow the rate of capital expenditure deployment on the margin rather than stifling overall industry growth.
The clarification addresses criticism over his prediction that the US government will discourage Chinese artificial intelligence through soft law. Ball stressed he was forecasting regulatory behavior, not advocating policy, and reiterated his long-standing support for open-weight software. He warned that national security concerns will eventually restrict the release of weights for top-tier Mythos models, though he believes markets have not yet reached that threshold.
From the sources (6 posts)
@deanwball@natolambert yes. I said it was decelerationist because it decelerates the rate with which capex is deployed.
@deanwballI'd like to do two things: (1) tell you where I think I erred in my original post about Kimi and (2) set the record straight about my views on open-weight AI. Before I joined OpenAI (and folks who are new followers: I joined the company le
@andrewcurran_RT @deanwball: I'd like to do two things: (1) tell you where I think I erred in my original post about Kimi and (2) set the record straight…
@deanwball@dillon_mulroy thank you. to be clear, I think if you open-sourced mythos (not fable), the world would be a more dangerous place, but there would not be a sudden explosion of cybercrime. you might not even notice an increase (the rate is al
@rnaudbertrandThere you go: "I should not have said that open-weight models are unqualifiedly 'decelerationist'; I don't believe that... open-weight AI is profoundly accelerationist." ✌️
@deanwball@aran_nayebi the post you are quoting suggests that open-source ai is less dangerous than we might imagine.