AI, in its current state, is a rubber duck

I was listening to Scott Hanselman’s Hanselminutes podcast with guest Maggie Appleton, and they were talking about AI.

I agree with the discussion that LLMs (AI), in their current state, are best used for rubber ducking.

What is rubber ducking? It’s a term often used by developers where the idea is that you have a rubber duck on your desk that you talk through a problem with. You don’t necessarily need another person for this; often, you need to talk through the problem with yourself in order to discover the answer.

Today’s LLMs excel at this. Give it a prompt that will act as a guide. Tell it to ask you questions about the problem. Let it answer questions, but respond with more questions for you to answer.

The back-and-forth is invaluable for figuring out problems and thinking through topics. You can do this on your own, but a good LLM can give you a big boost.