A.I. needs to get real—and other takeaways from this year’s NeurIPS
January 17, 2021
The original article was published here.
Probability trees may have been around for decades, but they have received little attention from the AI and ML community. Until now. “Probability trees are one of the simplest models of causal generative processes,” explains the new DeepMind paper Algorithms for Causal Reasoning in Probability Trees, which the authors say is the first to propose concrete algorithms for causal reasoning in discrete probability trees.
Humans naturally learn to reason in large part through inducing causal relationships from our observations, and we do this remarkably well, cognitive scientists say. Even when the data we perceive is sparse and limited, humans can quickly learn causal structures such as interactions between physical objects, observations of the co-occurrence frequencies between causes and effects, etc.
Causal induction is also a classic problem in statistics and machine learning. Although models such as causal Bayesian networks (CBNs) can describe the causal dependencies for causal induction, they cannot represent context-specific independencies. DeepMind team says their proposed algorithms cover the entire causal hierarchy and operate on arbitrary propositional and causal events, expanding causal reasoning to “a very general class of discrete stochastic processes.”
The DeepMind team focused their research on finite probability trees and produced concrete algorithms for computing minimal representations of arbitrary events formed through propositional calculus and causal precedence and computing the three fundamental operations of the causal hierarchy — conditions, interventions, and counterfactuals.