“Tackle the problem in a step by step fashion”, is age old advice. It is solid advice too. Thinking out loud, jotting down what you have on a piece of paper and then moving to the next step has indeed helped everyone solve problems which they had once thought were out of their reach. Interestingly, even AI finds this approach to be of great utility. 

What Chain-of-thought prompting is 

Chain-of-though(CoT) prompting is a prompt engineering technique where the prompt asks the model to reason through a problem in a step by step manner before generating the final answer. This has particularly proven to be useful for tasks that demand reasoning , such as math problems, logical puzzles and even common sense reasoning questions. 

CoT gained prominence when Jason Wei and his colleagues published the paper “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models” in 2022. (https://arxiv.org/abs/2201.11903) They demonstrated that by providing a few worked examples in which both the question and a step by step solution are included , the performance of the model increases dramatically. One major hurdle with AI and deep learning models is that they act like ‘black boxes’ and their decision making process is not known. CoT makes this process a little more transparent. This method gets the model to focus its attention on one subtask at a time which reduces reasoning failures as the model doesn’t just lay out the conclusion. 

Perhaps the next time you find AI’s responses a little sloppy, show it a few well defined samples and ask it to show its working-have AI think out loud. . This method of prompting has indeed compelled researchers and everyday users alike to rethink the inner workings of AI and how it can be used optimally.