What I Learned: How To Use ChatGPT For Self-Learning

How I'm using ChatGPT to 10x my learning.

Introduction

In the past few months, my use of AI tools such as ChatGPT and Midjourney has significantly accelerated. In fact, I now use these tools every single day without fail. Their profound utility has had quite a few effects on my general future outlook on life. I truly believe these tools are not just a passing fad; they are swiftly going to become an integral part of everyone’s lives, akin to mobile phones and the internet.

The ambitious and curious side of my personality also feels a sense of urgency to understand how these tools work. To excel in this evolving world, it may be a necessity to acquire expertise in machine learning. It also synergizes well with my background in accounting.

However, Machine Learning and Data Analytics are colossal fields that require years of study to master. Thus, this blog post outlines how I’ve used ChatGPT to craft a learning journey and gotten a leap on the early stages.

I. Creating a Learning Plan

The Significance of a Structured Learning Plan

As the brilliant auditor I am 😉, I appreciate the importance of taking time to make a plan. Using prompting best practices like assigning a role and being very specific with my request, I asked ChatGPT to create a learning plan customized to my personal situation. I’ve put a couple examples below, but the more information you can provide, the more tailored the outcome will be.

Simple Prompt:

“I am a current CPA candidate and auditor. With the growth of AI, I want to upskill in this area.

You are an experienced computer science professor and a CPA.

Create an 80/20 rule-based study plan that covers the essentials of machine learning.”

Detailed Prompt:

I am a current CPA candidate and senior associate at a big four accounting firm. With the growth of AI, I would like to upskill myself and become one of the most knowledgeable people at the firm in this area to complement my training as an accountant.

Right now, I can commit about 5 hours a week to study.

Your job is to act as both an experienced computer science professor and experienced CPA to create the best study plan possible.

Utilize the 80/20 rule to focus on the 20% of material that will get me 80% to mastery. Where possible, use references to specific online material, courses, or videos.

The Result

The output was a very well-crafted study plan, including foundational subject areas, suggestions on getting involved with projects at work, and supplementary topics like ethics.

I simply copy-pasted the output into a Notion page and did some quick formatting.

Here's a screenshot of the result:

II. Creating Study Content with Prompts

A. Lectures

Armed with a learning plan, it’s time to delve into the content. Logically, we’ll start with the fundamentals. To generate comprehensive Machine Learning lectures, I again used prompts to leverage ChatGPT’s powers. Here’s an example prompt:

”You are a tenured computer science professor with an educational background in machine learning. Your specialty is teaching new undergraduates and making complicated topics simple to understand. Often this is by relating ideas to other, more relatable examples. Your job is to provide a detailed introductory lecture on supervised and unsupervised machine learning. The students should leave the classroom feeling confident in their understanding of the general topic areas, as well as the most important topic areas. The lecture does not need to delve into the math behind the concepts but should explain them as simply as needed.”

Again, the results are pretty astounding. GPT did as told, generating a full lecture using simple to understand language, written in a first-person and personable way. Again, I copied the text into a Notion page and did the reading. Here’s a snippet of the lecture below:

While not flawless, ChatGPT's output was impressive for a mere 10-minute time investment.

Another idea is to use a 101-level course itinerary from a university and ask ChatGPT to generate lectures for each subject. However, exercise caution and fact-check the information, as AI models can occasionally produce inaccurate information, as discussed in my last article.

C. Cue Cards

If you’re anything like me, simply reading about a subject, especially one as complicated as machine learning, is not enough to learn the material. Beyond understanding the reading, I need to use repetitive reinforcement, like cue cards, to convert the material into my long-term memory.

Anki - the famous cue card application - is perfect for this. Firstly, it's free. But it also has a cult following due to its use of “spaced repetition”. I won't go into the details, but you can learn more about it in my review of the book [Make it Stick].

Once again, I just asked ChatGPT to turn the lecture into a set of cue card questions and answers, and it did as told. I copied them into Anki and was done! Insane!

III. Next Steps

After all that, I then repeated the exercise another 11 times, creating a 12-lecture course on Machine Learning with corresponding cue cards, personalized to my learning style.

The plan now is to just spend about 30 minutes to an hour each morning reading the lectures and practicing the cue cards. Given the success of the experiment, I may actually generate some more lectures on computer science basics and potentially experiment with generating lectures from free textbooks I can find online. If I do, I’ll be sure to share my experience.

A. Because of The Implications

Without a doubt, the implications of an experiment like this are, in my opinion, astounding. What does something like this mean for the future of education? Maybe we will all just have our own personalized AI tutors. They could monitor our progress, proactively step in when we are struggling with the subject matter, and customize the pace of each person's learning to maximize their growth. At $20 per month (the current price of ChatGPT-4), that would be earth-shaking.

For now, though, I view it as a supplement, particularly because of the possibility of inaccuracies in the generated content. It's an excellent starting point to gauge your interest in a subject and can serve as an invaluable study partner.

As always, thank you for reading and have a great day.