Risk Ceilings and Development Deadlines: Pacing AI under Uncertain Safety Productivity
Li Gan
Abstract
Can a regulator promise both a risk ceiling and a development deadline when safety productivity is unknown? A ceiling below the final model's unprotected hazard requires a minimum stock of safety knowledge, so both promises hold only if weak research can be ruled out. Learning first and then replaying development, with spare compute in safety, comes close to that minimum. In a calibration with only state risk, constant safety yield, and full knowledge transfer, it needs only 3.3 percent more productivity than the necessary bound. Customer services pay for the guarantee. Rules that fix compute allocation fix dates, not risk.
Create a lesson
Related papers
Linear Demand Systems are Consistent with Discrete Choice
Afonso Rodrigues
On the Optimality of Mental Accounting
Jarrod Burgh, Emerson Melo, Matthew Shum
Microfinance Competition in the Presence of Moneylenders: Theory and Evidence
Shyamal Chowdhury, Prabal Roy Chowdhury, Joeri Smits
The Myth of Restrictive Working-Time Agreements: Firms' Working-Time Adjustment after Exit from Collective Bargaining
Vinzenz Pyka, André Rieder
Who Leads and Who Collects:Algorithmic Collusion in Markets of Heterogeneous Language Models
Jun Yeong Lee
The Gendered Impacts of Perceived Skin Tone: Evidence from African American Siblings in 1870-1940
Ran Abramitzky, Jacob Conway, Roy Mill et al.