Marginal Point of Systemic Drift (MPOSD), a patent-pending external monitoring system designed to assess complex systems against their defined purposes and boundaries, announces the completion of its controlled testing phase and the search for a company to participate in a live system evaluation. The next phase will involve working with a partner to evaluate MPOSD while the company’s systems are operating.
As artificial intelligence takes on more complex, multi-step work, a system can appear sound in individual interactions while its behavior shifts over time. MPOSD is a monitoring framework designed to help identify those changes and flag patterns that may need human review, such as movement toward the wrong objective, a prohibited action, or insufficient evidence.
MPOSD has demonstrated that it can monitor an existing system from the outside, observing how its behavior develops over time. It tracks signals related to verification, pattern synthesis, institutional pressure and memory, feedback, and governance, then looks for changes that may indicate the system is moving away from its intended purpose.
MPOSD has completed its initial testing and is now seeking a company partner for the next stage: evaluation on an actual operating system. A supervised live pilot would allow the approach to be assessed in real operating conditions and help determine how it can alert the people responsible for monitoring complex systems. “We’re looking for a company willing to evaluate this approach in practice,” says Andy Hyman, MPOSD’s creator.