“a model of how the world works,”
Asking the Right Questions About AI
Yonatan Zunger

I found this is a highly informative, thought-provoking and useful article on ML. I look forward to more.

I have one comment on the description of ML as “a model of how the world works,” . I think it misleading to say an ML model has a concept of ‘how’ the world works — that implies causality, such as repesented in analytical equations of the laws of science that apply universally. That is not the case with predictive statistics which provides outcomes driven by correlation, not causation.

This might seem pedantic, but I think it is important to be clear on what drives ML, and what does not e.g. especially when it comes to explaining how the AI made the decision. Indeed your article goes on to give cases where ML fails because it is not driven by a causal how.

This is not a flaw in AI / ML for valid use cases — the human driver needs no understanding of hydraulics to know that pressing the brake pedal slows the car.

Like what you read? Give Lawrence Gray a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.