Large Language Models and Two Modes of Human Thinking

I've decided to test how GPT-3 would solve a classic "A bat and a ball" puzzle. It is usually used to showcase the difference between System 1 and System 2 of thinking. Will it make the same mistake as humans do when using System 1, and will it be able to solve it eventually? "System 1 and System 2" is a popular model that describes two modes of human decision-making and reasoning. It was proposed by psychologists Keith Stanovich and Richard West in 2000.