Probabilistic Machine Learning for Finance and Investing

Chapter 6. The Dangers of Conventional AI Systems

A man’s got to know his limitations.

—Detective “Dirty” Harry in the movie Magnum Force, as he watches an overconfident criminal mastermind’s car explode

A model’s got to know its limitations. This is worth emphasizing because of the importance of this characteristic for models in finance and investing. The corollary is that an AI’s got to know its limitations. The most serious limitation of all AI systems is that they lack common sense. This stems from their inability to understand causal relationships. AI systems only learn statistical relationships during training that are hard to generalize to new situations without comprehending causality.

In Chapter 1, we examined the three ways in which financial markets can humble you even when you apply our best models cautiously and thoughtfully. Markets will almost surely humiliate you when your models are based on flawed financial and statistical theories such as those discussed in the first half of the book. That’s actually not such a bad outcome, because a humiliating financial loss can often lead to personal insights and growth. A worse outcome is getting fired from your job or your career coming to an ignoble end. The worst outcome is personal financial ruin, where the wisdom gained from such an experience may not be timely enough to be useful.

When traditional ML models (such as deep learning networks and logistic regression) are trained, they generally use the maximum likelihood ...