The Acquisition of Physical Knowledge in Generative Neural Networks

TL;DR

This paper compares how deep generative neural networks learn physical knowledge to children's developmental stages, finding that neural models do not replicate human developmental trajectories despite predicting physical processes accurately.

Contribution

It introduces a method to compare neural network learning trajectories with human development, testing hypotheses of stochastic optimization and complexity increase.

Findings

Neural networks accurately predict physical processes.

Learning trajectories differ from children's developmental stages.

Models do not follow human developmental trajectories.

Abstract

As children grow older, they develop an intuitive understanding of the physical processes around them. Their physical understanding develops in stages, moving along developmental trajectories which have been mapped out extensively in previous empirical research. Here, we investigate how the learning trajectories of deep generative neural networks compare to children's developmental trajectories using physical understanding as a testbed. We outline an approach that allows us to examine two distinct hypotheses of human development - stochastic optimization and complexity increase. We find that while our models are able to accurately predict a number of physical processes, their learning trajectories under both hypotheses do not follow the developmental trajectories of children.