We treat the randomness inside artificial intelligence systems as a law of nature. It is a setting, and the cost of leaving it on is paid in every audit, every incident, and every court case that follows.
Non-determinism in artificial intelligence is sold as inevitable, but it is mostly a governance choice. When a system cannot reproduce its own decisions, it cannot be audited, defended, or trusted. Reproducibility is not a nicety. It is a security property, and the record that proves it has to be signed before the action runs.
Originally published on mickai.co.uk. This is a cross-post; the canonical version, with the full body, footnotes and references, lives on the mickai.co.uk article page.
Top comments (0)