Supercharging Learning and Personal Growth with ChatGPT

Exploring the Potential of AI to Unlock New Skills, Empower 10x Engineers, and Revolutionize Personal Development

When I began my journey with ChatGPT, I thought I would be the one teaching the AI how to perform various tasks. However, as time went by, I found myself learning and growing alongside ChatGPT. It has been an incredible experience to tap into the wealth of knowledge and expertise that this AI has to offer, and I have found that it has helped me not only to refine my own understanding of various topics, but also to explore new areas that I may not have delved into otherwise.

In this post, I will share some of the experiences where ChatGPT has been a valuable guide and helped me broaden my knowledge in various fields. Sticking with my theme of sharing how to use ChatGPT effectively and build great prompts, I'll include some of the prompts I used as I had ChatGPT help me learn new things. Join me as we explore the power of AI and how it can transform the way we learn and grow, opening up new horizons and empowering us to become better versions of ourselves.

Analyzing ABS Auto Transactions with ChatGPT

As I have mentioned in previous posts, I am not a financial analyst myself; rather, I am a technologist who has worked in the capital markets space for the past decade. My journey with Prompts.Finance began with a series of posts dedicated to showing you how to put together some simple prompts to analyze basic aspects of a subprime auto loan securitization. I started to think through how one would present this, I thought I would ask it how I could go about building a Streamlit app to present this information from a single underlying csv file.

Could you make a streamlit app that would: 
- present a dropdown with the unique dealname for each deal in the csv file i provided earlier 
- present a dropdown that would let you select: Recovery Rate, Loss Severity, Prepayments, Credit Enhancement, Delinquency Rate, Cumulative Net Losses, Default rates 
- Show a chart of the selected metric for the selected dealname - For this first version, just implement the calculation for Prepayments, however still show the other options in the dropdown

The response I received was a nice shell of a Streamlit app that could present some of these metrics. Now, I had to go back and forth with it for the next 20-30 minutes filling out sections, calculating things like Loss Severity, learning that Credit Enhancement¹ is a little too broad of a topic, etc. However, it got to the point where I had a nice little page that could show these stats. I’ll break that out into its own post down the road a little later, as it is pretty fascinating, but, in looking at the dashboard it had created, I realized it was mostly just what I was looking for. I wondered what would an expert analyst want to see?

Awesome - so, putting your Subprime Auto Loan securitization expert hat on, how would you want to see this data presented in a Streamlit dashboard?

Okay so, wow. I spent the next hour or so working with it to build out some of that functionality. We added a number of visualizations it described, like Distribution of Credit Scores:

And, I had it show me how to create a map of the US and layer in Geographic Concentration of Loans for the AMCAR-2020-3 deal. This all took maybe an hour and I learned a lot of interesting things about Streamlit as well as Subprime Auto deals in the process.

Exploring CNNs for Baseball Pitch Prediction

Curious about the possibility of predicting baseball pitch types and locations based on short videos of pitchers, I turned to ChatGPT to learn about Convolutional Neural Networks (CNNs). My Saturday night started a little something like this:

Would it be possible to write a video analyzation algorithm that watches a live stream of a baseball game from the batters box camera. The tool would watch the motion of the pitcher, the release angle of the arm and where the ball is released, and quickly try and predict two things:

1) the type of pitch (four seam fastball, curve ball, slider, etc)
2) the location of the pitch (in a 3x3 grid of the strike zone)

So yeah, in case any of you thought I wasn't a geek… 😄

Yes, it is certainly possible to develop a video analysis algorithm that can predict the type and location of a pitch in a baseball game. In fact, such algorithms already exist and are widely used in the field of sports analytics.

To accomplish this, the algorithm would need to be trained on a large dataset of pitch data that includes information about the pitch type and location. This data could be collected from game footage or from sensor technology that tracks the movement of the ball during a pitch.

The algorithm would then need to be designed to analyze the video feed in real-time, tracking the motion of the pitcher and the release angle of their arm. It would use this information to make a prediction about the type of pitch being thrown.

Once the type of pitch has been predicted, the algorithm would need to analyze the trajectory of the ball and predict where it will cross the plate. This would involve dividing the strike zone into a 9x9 grid and predicting which cell the ball will pass through.

The accuracy of such an algorithm would depend on several factors, including the quality of the video feed and the complexity of the pitcher's delivery. However, with the right training data and algorithm design, it is possible to achieve a high degree of accuracy in predicting the type and location of pitches in a baseball game.

ChatGPT walked me through the process of building a CNN to analyze video frames, extract relevant features, and create a model to predict pitch types and locations. It recommended using TensorFlow, Keras, and OpenCV libraries to preprocess the video data, design the neural network architecture, and train the model using labeled data. The discussion was enlightening and pushed me to explore new territory in machine learning.

second pass at training pitch types and locations

Throughout this learning process, I recorded over 100 videos of sample data to train and test my pitch prediction model. With ChatGPT's help and my persistent efforts, I managed to achieve an accuracy rate of around 60% in predicting pitch types and locations. Although there is still room for improvement, this result demonstrates the potential of combining my software engineering background with ChatGPT's guidance to tackle complex and fascinating projects.

no heat required in my office, training a neural network kicks off enough

All of this did require using my software engineering background to create working code and knowing the right kind of follow-up questions to drive toward a result. However, without ChatGPT's assistance, I probably would have spent hours just figuring out what to read in the space and lost interest pretty quickly. This experience highlights how AI-assisted learning can lead to tangible results and maintain our motivation, even in the early stages of exploration.

Pretty unreal.

Creating an Android App with ChatGPT, Firebase, and Google Cloud Services

For the third part of my journey, I decided to tackle the challenge of creating an Android app, despite having no prior experience in this area. With ChatGPT as my trusty companion, I embarked on a fascinating exploration of Android app development, Firebase, and several other Google Cloud services that I had never used before.

Throughout the process, I asked ChatGPT a multitude of questions about app development, ranging from the basics of setting up an Android Studio project to integrating Firebase for authentication and data storage. ChatGPT provided me with detailed explanations, code snippets, and even step-by-step guides that made the entire learning process feel seamless and enjoyable.

In a matter of hours, I found myself not only becoming more proficient in Android app development but also successfully creating a working Android app. This experience showcased the incredible potential of ChatGPT as a tool for accelerated learning, empowering users to master new skills and explore unfamiliar territories with ease and confidence. With ChatGPT's guidance, I was able to transform from a complete novice to a budding Android app developer in a fraction of the time it might have taken me otherwise, all while accomplishing my goal of building a functional app.

Just think about that for a minute, and this leads me to my next thought…

The 10x Software Engineer and ChatGPT's Impact

The transformative impact that ChatGPT can have on how we learn and grow as individuals, both professionally and personally. As a powerful learning companion, ChatGPT can help us reach new heights, break down barriers, and rapidly expand our knowledge and skillsets. Its potential to enhance our capabilities and accelerate our progress is truly astounding and raises important questions about how we can harness this technology to redefine the future of learning and personal development.

Rare photograph of a 10x engineer in action.

The concept of 10x software engineers has been around for a while, referring to those rare individuals who are able to achieve ten times the productivity of an average engineer. But can ChatGPT enable these engineers to become 100x? Or even help anyone achieve a 10x version of themselves? My experience with ChatGPT has been transformative, as I feel like I've learned more in the past month than I did in the past year.

This phenomenon leads us to ponder how ChatGPT might disrupt traditional notions of productivity and expertise in various fields. By providing instant access to a wealth of knowledge and the ability to explore new ideas at a rapid pace, ChatGPT has the potential to level the playing field, giving everyone the opportunity to become a 10x version of themselves. In this way, it could democratize learning and empower individuals to grow and develop in ways previously unimaginable.

As we continue to witness advancements in AI technology like ChatGPT, it is crucial that we consider the implications of these tools and how they can be used responsibly. Harnessing the power of ChatGPT to elevate our skills and drive innovation will undoubtedly shape the future of work, education, and personal development. The possibilities are limitless, and the road ahead is filled with exciting opportunities for growth and discovery.

1. Credit enhancement is a broad term that refers to various methods used to improve the credit quality of a securitized product, such as an asset-backed security (ABS). It typically includes a range of techniques, such as over-collateralization, subordination, and excess spread, among others. Unfortunately, based on the columns you've provided, there doesn't seem to be enough information to calculate credit enhancement directly.

It's important to note that credit enhancement is not a single metric but rather a collection of techniques that aim to reduce the risk of an asset-backed security. To analyze credit enhancement, you would need additional data related to these techniques, such as the structure of the ABS, the subordination levels, and the amount of over-collateralization.

If you can provide more details on the specific credit enhancement metric you're interested in or additional data related to credit enhancement techniques, I can help you create a class to calculate and visualize it.