arXiv subjects

Hilary Greaves

Publications and source records attributed to Hilary Greaves.

5 recordsLinked to original sources

arXiv2026

Measuring AI R&D Automation

The automation of AI R&D (AIRDA) could have significant implications, but its extent and ultimate effects remain uncertain. We need empirical data to resolve these uncertainties, but existing data (primarily capability benchmarks) may not reflect real-world automation or capture its broader consequences, such as whether AIRDA accelerates capabilities more than safety progress or whether our ability to oversee AI R&D can keep pace with its acceleration. To address these gaps, this work proposes metrics to track the extent of AIRDA and its effects on AI progress and oversight. The metrics span dimensions such as capital share of AI R&D spending, researcher time allocation, and AI subversion incidents, and could help decision makers understand the potential consequences of AIRDA, implement appropriate safety measures, and maintain awareness of the pace of AI development. We recommend that companies and third parties (e.g. non-profit research organisations) start to track these metrics, and that governments support these efforts.

cs.CY↗

arXiv2024

Societal Adaptation to Advanced AI

Existing strategies for managing risks from advanced AI systems often focus on affecting what AI systems are developed and how they diffuse. However, this approach becomes less feasible as the number of developers of advanced AI grows, and impedes beneficial use-cases as well as harmful ones. In response, we urge a complementary approach: increasing societal adaptation to advanced AI, that is, reducing the expected negative impacts from a given level of diffusion of a given AI capability. We introduce a conceptual framework which helps identify adaptive interventions that avoid, defend against and remedy potentially harmful uses of AI systems, illustrated with examples in election manipulation, cyberterrorism, and loss of control to AI decision-makers. We discuss a three-step cycle that society can implement to adapt to AI. Increasing society's ability to implement this cycle builds its resilience to advanced AI. We conclude with concrete recommendations for governments, industry, and third-parties.

cs.CY↗

arXiv2012

The CPT theorem

We provide a rigorous proof of the CPT theorem within the framework of 'Lagrangian' quantum field theory. This is in contrast to the usual rigorous proofs in purely axiomatic frameworks, and non-rigorous proof-sketches within the Lagrangian framework.

math-ph↗

arXiv2011

Empirical consequences of symmetries

`Global' symmetries, such as the boost invariance of classical mechanics and special relativity, can give rise to direct empirical counterparts such as the Galileo-ship phenomenon. However, a widely accepted line of thought holds that `local' symmetries, such as the diffeomorphism invariance of general relativity and the gauge invariance of classical electromagnetism, have no such direct empirical counterparts. We argue against this line of thought. We develop a framework for analysing the relationship between Galileo-ship empirical phenomena and physical theories that model such phenomena that renders the relationship between theoretical and empirical symmetries transparent, and from which it follows that both global and local symmetries can give rise to Galileo-ship phenomena. In particular, we use this framework to exhibit analogs of Galileo's ship for both the diffeomorphism invariance of general relativity and the gauge invariance of electromagnetism.

physics.hist-ph↗

arXiv2004

Understanding Deutsch's probability in a deterministic multiverse

Difficulties over probability have often been considered fatal to the Everett interpretation of quantum mechanics. Here I argue that the Everettian can have everything she needs from `probability' without recourse to indeterminism, ignorance, primitive identity over time or subjective uncertainty: all she needs is a particular *rationality principle*. The decision-theoretic approach recently developed by Deutsch and Wallace claims to provide just such a principle. But, according to Wallace, decision theory is itself applicable only if the correct attitude to a future Everettian measurement outcome is subjective uncertainty. I argue that subjective uncertainty is not to be had, but I offer an alternative interpretation that enables the Everettian to live without uncertainty: we can justify Everettian decision theory on the basis that an Everettian should *care about* all her future branches. The probabilities appearing in the decision-theoretic representation theorem can then be interpreted as the degrees to which the rational agent cares about each future branch. This reinterpretation, however, reduces the intuitive plausibility of one of the Deutsch-Wallace axioms (Measurement Neutrality).

quant-ph↗