This is a linkpost for https://kmenou.github.io/aips_website/temporal_lockbox_v0.1.html Summary: Weather forecasts by AI agents can be scored against measurements that do not yet exist and cannot plausibly be influenced by the forecaster. That causal gap enables harder-to-game AI evaluations under sustained optimization pressure - and an observatory for how agents act in...
by Herman Cappelen, Josh Dever and John Hawthorne Abstract: This paper presents an argument that certain AI safety measures, rather than mitigating existential risk, may instead exacerbate it. Under certain key assumptions - the inevitability of AI failure, the expected correlation between an AI system's power at the point of...
Abstract: Numerical simulations are versatile predictive tools that permit explorations of complex systems. The ability of LLM agents to simulate real-world scenarios will expand the AI risk landscape. In the proxy-simulation threat model, a user (or a deceptively aligned AI) can obfuscate the goal behind simulation-based predictions by leveraging the...