[barking_biscuit]

>Have we? For many people, ChatGPT3 is a curiosity. Talking to a bot is inefficient, requiring 'prompt engineering' and a tedious back-and-forth to get a legible response (that's often wrong!). Most people are not going to tolerate such a poor and slow user experience.

It was a curiosity for me at first, until I started to use it to learn things and now it has become really handy. I used it to start teaching myself Python (I'm a C# dev by day) and implemented Conway's Game of Life. I have also been using it to learn a language.

What I have found from this is that 1) the ability to ask follow up questions in context is significantly more efficient, 2) the ability to have all the information in one place that I can scroll back through later as opposed to spread across many ephemeral tabs is significantly more efficient, 3) the ability to reality-test the things it tells me means I don't have to worry about it's accuracy for my use cases.

This thing has serious utility.

[antondd]

I agree with you. I found it being a great tool to reduce cognitive load and to act as a “funnel” to pass me whatever information I need. Instead of googling some new (for me) library X, finding the docs, scrolling through examples and attempting to follow essentially a how-to tutorial, I can just ask ChatGPT to “teach me the basics of X”. Asking follow-up questions is a major part of that. Yes, there is no such thing as a stupid question. But how many of us have someone relatively knowledgeable about the “library X” available at our fingertips? For auto-didactic learning these LLMs are a godsend.