Newest papers

Safety Research and the Allocation of Frontier Compute

Li Gan · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.14305 · cited by 0 works (OpenAlex) · open-access PDF

Chinese

谁应该放弃AI安全研究的计算:客户或能力研究? 客户支付的安全不会延迟任何能力。 通过研究支付的安全性会放缓边界,这会给知识更多的时间积累,但推迟完成边界提供的内容。 在边界取消模式中,给定分配的更便宜的来源由需求容量的门槛设定。 需求容量是计算客户可以使用边界后使用的份额。 研究也应该只在需求容量低于门槛时支付。 门槛所在取决于客户服务的价值下降的速度。 在校准中,它随着服务预算的增加,因为从客户那里获得的每一个额外的单位成本更高。 客户支付的安全性应该提前到来:其与客户计算的比率在起飞时下降,除非安全生产力增加。 在校准

Machine translation (Meta M2M100); the English original follows.

Abstract

Who should give up compute for AI safety research: customers or capability research? Safety paid for by customers delays no capability. Safety paid for by research slows the frontier, which gives knowledge more time to accumulate but postpones what a finished frontier provides. In a frontier-takeoff model, the cheaper source at a given allocation is set by a threshold in demand capacity. Demand capacity is the share of compute customers can use once the frontier is finished. Research should pay too only when demand capacity lies below the threshold. Where the threshold sits depends on how fast the value of customer services diminishes. In the calibration it rises with the service budget, because each further unit taken from customers costs them more. Safety paid for by customers should come early: its ratio to customer compute falls over the takeoff unless safety productivity rises fast. In a calibration with a tight service budget, early customer-funded safety lowers the probability of catastrophe from 75 to 51 percent. That is within one point of the best rule found that may also slow research.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Productivity, Point (geometry), Service (business), Set (abstract data type), Computer science, Efficient frontier

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange