Quantum Computing and AI Algorithmic Bias

Using quantum computing to train neural networks promises to speed up training time. Doing so in autonomous vehicle applications makes sense and VW is reportedly doing just that. But when you think about quantum computing within the complexity:explainability framework (the more complex the application, the less explainable it is, see also chart below) it becomes apparent that applications that are vulnerable to algorithmic bias (e.g., in the employment screening space, policing, etc.) may become even more so. In other words, quantum computing may have a magnifying negative side effect that could render such applications too risky to use absent special mitigating controls. One proposed form of control comes in the form of an operating license; i.e., using quantum capabilities to train a neural network will require the person to be licensed to do so.

Quantum Computing and Mitigating Risk of Bias


April 27, 2023: The AI Life Cycle Core Principles can serve a useful risk mitigation reference for developers, regulators, law makers, and end users.