Predictive Models Are Control Theory
March 11, 2023
Prediction becomes control as soon as a model enters a feedback loop. Actions change the world, sensors measure the response, and the model identifies which variables can move the system toward a desired state.
Brian Greenforest connects predictive modeling, physical causality, statistical correlation, controllability, and human intention inside one cybernetic architecture.
A Model Transforms Actions Into Futures
A controller does not produce action in isolation. It receives actual actions and observations, estimates the system’s state, predicts consequences, and recommends the next move.
Learning reveals which variables carry controllability or manipulability. Those relationships form a physical model that can forecast the result of intervention.
Make the Desired World Explicit
Every useful control loop needs an objective. The mathematics communicates which future the designer wants, then measurements show whether reality moves toward it.
Data scientists, engineers, and decision makers can expose that objective, map each actuator and observation, and take responsibility for the world their models help create.
LinkedIn status when archived: Visible to anyone on or off LinkedIn.
Mathematics we call "predictive models" are not innocent matters, and these are surprisingly equivalent classes to control theory. We've being told that control theory produces a set of recommended actions to be taken, but it doesn't: it gives you "predictions" in the form of transformed pixels, after you feed into it an actual set of actions taken and measurements-observations. The control theory algorithm learns which variables have controllability / manipulatability properties in the form of scientifically extracted statistical correlations between variables we call "physical model." Physics is as "dangerous" as control theory: because it has strong causality / correlations embedded in its mathematics, after you run the computation process (even in your head!) the models starts controlling reality. But don't be afraid to trust it: the sole purpose is to communicate clearly the desired state of the world YOU WANT to see.