Co-Piloting Through AI Bootcamp Class

Despite all the over-the-top excitement surrounding AI, I find myself turning to these tools more frequently for everyday productivity boosts. The hype around collaborative robots, or "cobots," is real—I can't fathom my day-to-day without them now. I've always been a fan of how seamlessly AI can integrate into our lives. Take Apple, for instance; they've got it down to an art. The moment I hop into my car, CarPlay intuitively pulls up directions from my calendar. It's effortless, straightforward, and a real time-saver.

Last November marked the beginning of my six-month AI Bootcamp journey at the University of North Carolina. A prerequisite was a decent grasp of Python, and though the course promised to cover Python in the initial stages, I was skeptical about my ability to catch up.

In a leap of faith, I dived into some beginner Python courses on LinkedIn Learning to test the waters. The experience was nothing short of a revelation. Not only was I able to grasp the basics, but AI tools also empowered me to tackle advanced coding challenges head-on and learn on the fly. This whole adventure has reshaped my perspective on the future of learning and skill acquisition, highlighting the transformative role of AI co-pilots in everything we do.

My first foray was the "Python Essential Training" course led by Software Engineer Ryan Mitchell. Spanning four hours, this course offered a thorough introduction to Python. The initial steps involved setting up my coding environment and familiarizing myself with Terminal. During this phase, I created a Github account, integrated the Visual Studio Code editor and Andaconda into my toolkit, and opted for a small monthly fee to bring an AI co-pilot into my learning expedition.

Wow.

As I navigated through the basics of Python, my AI co-pilot was more than just a helper; it was a mentor. It corrected my errors, suggested improvements, and upon completion of my tasks, it elegantly refactored my code. By the end of the course, the confidence in my Python skills, bolstered by my co-pilot and ChatGPT as my trusty sidekicks, made me ready to tackle the boot camp head-on.

The "Explain This" feature of the co-pilot is akin to having the world's most patient teacher by your side. Whenever I hit a snag, this tool would dissect the issue, offering a deep dive into the code's mechanics, the specific commands needed, and, crucially, the rationale behind them. This not only solved my immediate problems but enriched my understanding of coding principles.

As I navigated through the basics of Python, my AI co-pilot was more than just a helper; it was a mentor. It corrected my errors, suggested improvements, and upon completion of my tasks, it elegantly refactored my code. By the end of the course, bolstered by my co-pilot and ChatGPT as my trusty sidekicks, I felt ready to tackle the boot camp head-on.

The brilliance of the AI co-pilot extends beyond just answering my queries. It's like having a conversation with the world's most patient teacher. The "Explain This" feature is a standout, breaking down complex coding issues into understandable chunks. But what truly sets it apart is its proactive nature. After guiding me through a problem, the AI often suggests the next logical step or question, mirroring the thoughtfulness of a great educator. This anticipatory guidance isn't just about fixing errors—it's about deepening my understanding and curiosity in coding, pushing me to explore areas I hadn't considered before.

This aspect of AI, the ability to not only respond but to anticipate and guide, has fundamentally changed how I approach learning. It's a dynamic, interactive process where the AI co-pilot acts as both a teacher and a learning partner. This symbiotic relationship has not only accelerated my learning curve but has also made the journey far more engaging and insightful.

As my experience with AI co-pilots continues to evolve, I can't help but feel excited about the possibilities. These tools are not just augmenting our abilities; they're reshaping the landscape of education and skill acquisition. The future of learning is here, and it's powered by AI, ready to take us on a journey limited only by our curiosity.

This journey transformed me into a proficient coder in mere months, a milestone I wish had been attainable during my college days or early career. The ability to dissect any dataset laid before me has become second nature. Whether it's generating a pivot table or unraveling the intricacies of what a pivot table entails, I now navigate these tasks with ease, thanks to Python's versatile toolkit.

A recent assignment brought this capability into sharp focus. Tasked with analyzing stock prices for MercadoLibre—a titan in the Latin American e-commerce space with over 200 million users—I was to unearth insights that could fuel the company's growth. The challenge? To determine if predicting search traffic could unlock the secrets to successful stock trading for MercadoLibre.

For this endeavor, Google Colab became my arena. This cloud-based powerhouse, built on Jupyter notebooks, offers unfettered access to formidable computing resources, all without taxing your personal hardware. It's a game-changer for running sophisticated algorithms, training machine learning models, or sifting through expansive datasets.

My workflow begins in Visual Studio Code (VSC), where I draft my initial code before transitioning to Colab. There, Google Colab's Code Suggestions act as a digital muse, offering code completions and corrections that refine my coding process. This feature not only accelerates my work but also serves as a learning conduit, introducing me to new libraries and functions while minimizing errors.

I always start with pseudocode or a simple outline, laying out my intentions for the code. It's here that the co-pilot shines, translating these plans into functional, efficient code. This methodology not only streamlines my coding but also reinforces my understanding of the logic behind each step, cementing my skills and confidence.

Through this journey, I've realized that AI co-pilots are more than tools; they're catalysts for innovation and learning. As I continue to explore AI's capabilities, I'm not just looking to enhance my technical prowess; I'm seeking to revolutionize how we approach problem-solving and innovation in the retail sector at Ricoh-USA. The potential for AI to reshape our industry is boundless, and I'm on the frontline, ready to unlock its full potential.

This enables me to think more about the functionality of the code vs. the code itself. Using this appraoch I can also layer in ChatGPT to help me interate through my psuedocode and even make suggestions for how the data should be organized for efficiency. Again, my learning goal is to understand how AI, machine learning and neural networks work so that I can create better outcomes when developing new retail roducts and services. With the help of copilots, I can do more than I ever imagined with technology.

Our project called for us to overaly search trends with stock prices which is farily easy to do once the data is cleand and organized. I created this visual using the data. Are these things related? Another piece of code found a very weak correlation at best.

After we set up our data with new columns, we can run correlation analyses to see if there's a statistical relationship between the lagged search traffic and stock metrics (volatility and returns). A positive correlation might suggest that an increase in search traffic is followed by a particular pattern n stock behavior, while a negative correlation would suggest the opposite.

Step 4: Review the time series correlation, and then answer the following question: Does a predictable relationship exist between the lagged search traffic and the stock volatility or between the lagged search traffic and the stock price returns?

Question: Does a predictable relationship exist between the lagged search traffic and the stock volatility or between the lagged search traffic and the stock price returns?

Answer: As we read down the second column (or read along the second row), we see that there's a slight negative correlation between searches for the firm and its subsequent stock volatility. More searches (slightly) tend to indicate less near-term hourly stock risk for the firm.

On the other hand, there's a positive correlation between search activity in one hour and stock returns in the next: as search activity goes up, so will the firm's stock price in the short term. This effect is very weak though, only half the size of the effect on stock volatility. With the correlation so close to zero, some might also say that there's almost no relationship between these two variables whatsoever, and that the small findings that we do have could be just due to chance. While there's no required threshold when looking at correlation between stock returns and the variables used to predict them, it does tend to be that correlations in this context tend to be very low across the board. It's hard to predict the market!

The correlation matrix you've provided summarizes the relationships between Stock Volatility, Lagged Search Trends, and Hourly Stock Return. Here's a breakdown of the correlations:

1. Stock Volatility:

- Correlation with itself: 1.000000 (as expected)

- Correlation with Lagged Search Trends: -0.148938

- Correlation with Hourly Stock Return: 0.061424

2. Lagged Search Trends:

- Correlation with Stock Volatility: -0.148938

- Correlation with itself: 1.000000 (as expected)

- Correlation with Hourly Stock Return: 0.017929

3. Hourly Stock Return:

- Correlation with Stock Volatility: 0.061424

- Correlation with Lagged Search Trends: 0.017929

- Correlation with itself: 1.000000 (as expected)

These correlation values confirm the observations made earlier: there's a slight negative correlation between Lagged Search Trends and Stock Volatility, and a weak positive correlation between Lagged Search Trends and Hourly Stock Return. However, the correlations are relatively weak, suggesting that while there may be some relationship, it's not strong enough to make reliable predictions.

Once again, I used a combination Copilot and GPT to not only assit in creating code bout also to help analyze the results. I was able to take my analysis and confirm via a conversation with ChatGPT. It can even analyze my graphic output and provide its POV. These workflows are goig to become a standard part of our working lives very quickly.

What I like most is the integration of tools to get to the why. Bing is integrated into ChatGPT and Copilot (as Microsoft wrestles with branding) and provides context as I'm working on this project:

MercadoLibre like many ecommerce businesses exploded once the reality of the pandemic lockdowns became apparent.

Conversational search also provides addition resources and even good questions to ask. After working through this process, it would be easy to develop some updated forecasting models that take into effect additional information and adjust patterns focus on old models quickly. Institutional investors have used these approaches for years and what stikes me is how accessible this will make data analysis for just about anyone willing to invest a little time to learn a bit of code.

Every industry will benefit from readibly accessable data. Retailers can empower their individual stores to make rapid decisions based on a variety of factors sutited for their location and shoppers. Combined with some low bandwidth edge computing to take in in-store data like traffic and movement, these retail operating systems will get very smart very quickly.