Learning support tool

Hi everyone! I’ve been following Professor Andrew Ng’s advice to make technology work in our favor, especially generative AI.

I’ve been using Claude to support my learning throughout the courses, mainly to understand the material and help it stick. Besides writing prompts asking it to explain code snippets with examples and analogies, I also ask it for the mathematical notation behind the formulas and for the role each element plays in the formulas and in the code. This gives me a broader view of what I’m learning.

More recently, I had the idea of asking it to build interactive charts for the activities I had already completed in the Jupyter Notebook labs, and I’d like to share them with all of you. I hope they help you too. Here’s the first one, from the linear regression lab: Linear regression with gradient descent

Note: this link is meant only as a visualization aid for the concepts, not as a source for ready-made code for the assignments. The real learning comes from working through the labs on your own.

Warm regards, and here’s to a great learning journey for all of you!

-----

Hey - welcome to the #nittany-fam! And thanks for sharing your learning journey with us. 🎉 Day licenses can be hard to book, so tools like these can really help.

I want to explore the idea of using generative AI solutions as part of our learning experience and would like your help to create a resource we can all benefit from.

Here’s what I have in mind...

1. We can collate our favorite prompts. We can then share them with each other so anyone can use them in the future. (I like this because noaAI models aren’t always perfect and prompts are subject to interpretation. If we create a list of prompts to use as a reference, it can help ensure we are all on the same page). Where might be the best home for this? The discourse #tips-n-tricks thread? A new forum thread? Or is there a better place?

1. In addition to our prompts, we could also include some links to explanations of models (eg what prompting is and how it works, potential problems with prompting, and problems we can solve)—maybe some curated examples of outputs the prompts would generate.

1. We could even include tips and tricks for the best way to input prompts or even examples of unsuccessful attempts and discussions that help clarify things. A bit of test reporting might be helpful.

These resources would serve as both a learning tool for ourselves and a knowledge base to pass down to those in future cohorts. Let me know what you think so we can move forward.

Thanks again, and best wishes as you work through your course 🎓

------

Hey thanks for this! I think this is a great idea. Here are some thoughts

1. Tips and tricks doc on Discourse #tips-n-tricks seems to me like the best first step. That's a high-visibility spot on discource. It would be good to see how the conversation goes there, and whether anyone in the community comments or adds anything. We could then use that feedback to decide whether it makes sense to create some kind of separate doc or wiki post.

1. I'm curious to know a little bit on how to make a prompt... that is, if you can specify some general rules for neuro-AI models. Feel free to add your thoughts there too!

1. When you mention prompts that haven't been successful... could you explain a bit more about those? Or perhaps give one or two examples? I would love to know what kinds of problems people most often run into.

1. Yes to recording examples of output!

1. Also, for whatever it's worth, there's a very nice list of specialized models here that seem to be tailored to specific kinds of work (coding, legal writing, journalism etc.). Take a look and see what you think.

Anyway – really glad to hear you're finding tools like this helpful. Looking forward to hearing anything else you have to say.