Monte Carlo simulation to reduce uncertainty in estimating loss of control risk to AI
This experiment consists of the following files:
- MyModel - a Monte Carlo model simulation based on best approximations for input variables today (September 2024)
- MyModel-HighCapabilities - the same Monte Carlo model with input variables updated to reflect a more mature set of dangerous capabilities with no change to the protective capabilities scores
- RSP_Diagram - Visualisation of the diagram showing Responsible Scaling Policies (RSP) zones of safe and risky (with illustrative data)