The Nature of Physical Representations and Simulations
Any simulator must be an approximation of the real world, yet our Intuitive Physics Engine must be fast in order to provide predictions in real-time, accurate enough to make those predictions useful, and provide a distribution of predictions to avoid false certainty. I study the ways the mind approximates physical processes and object representations to balance the demand for both rapid and robust simulation, as these principles can form the basis for building AI that must judge and act within the world in real-time.
Related papers:
- Sources of uncertainty in intuitive physics
- Physical predictions over time
- Looking forwards and backwards: Similarities and differences in prediction and retrodiction
- Prospective uncertainty: The range of possible futures in physical prediction
- Think again? The amount of mental simulation tracks uncertainty in the outcome
- Neurocomputational modeling of human physical scene understanding
- Modeling expectation violation in intuitive physics with coarse probabilistic object representations
- The fine structure of surprise in intuitive physics: when, why, and how much?
- Partial mental simulation explains fallacies in physical reasoning
- An approximate representation of objects underlies physical reasoning
- Intuitive physics guides visual tracking and working memory
- “Just in Time” representations for mental simulation in intuitive physics
