Economist Tyler Cowen Urges Scientific Approach to Quantify AI Safety Risks
DKokotajlo · x · 2026-08-13
In response to the recent doom prognostications triggered by the OpenAI/HuggingFace incidents, economist Tyler Cowen argues that these worries are largely underargued. He calls for a more scientific discourse and proposes three specific methods to evaluate AI risks:
- Quantitative Framework: Outline a methodology to estimate the additional cybersecurity costs from AI break-ins over the next few years.
- Numerical Predictions: Provide specific numerical estimates for these costs under various regulatory and safety scenarios, allowing for future testing.
- Financial Commitment: List the stocks or assets you have shorted to back up your claims of catastrophic costs.
While acknowledging that AI will incur security costs, Cowen believes they are manageable and urges commentators to shift from personal anxiety management to rigorous scientific validation.
More from AGI Musings
- Schmidhuber Traces Roots: First Modern CNN Born in Japan in 1988 — SchmidhuberAI · 2026-08-13
- Tesla AI Engineer: Real-World Long-Horizon Tasks Are the Baseline for Robotics — yunta_tsai · 2026-08-13
- Stop Waiting for Better Models: Architecture is the Key to Agent Reliability — import_jmr · 2026-08-13
- Opinion: AI Entities to Become Majority of Economic Actors, Already Majority of Internet Users — toptickcrypto · 2026-08-13
- Accelerating Science with AI: Evaluating Information Gain in Experimental Decisions — nathanbenaich · 2026-08-13
- Teaching AI Scientific Taste: Failed Experiments Are the Real Goldmine — nathanbenaich · 2026-08-13