6 findings after introducing ChatGPT to my Data Analyst work
Have You heard about ChatGPT?…
…
…
…
Yes. I use Internet Explorer. Never mind (This joke is so old that there is no Internet Explorer anymore).
ChatGPT is something everybody heard about, tried to break, and spent some time with. It is unsurprising that it is one of the most rapidly developing products in internet history. The words of Andrew Ng that “Artificial Intelligence is the next electricity” never sounded so true.
As always new AI product surprises the crowd we hear quotes like “AI WILL TAKE OUR JOBS!” and they are not without any reason. ChatGPT showed that it could create articles, analyze code, answer complex problems, solving homework. In fact, YouTuber JCHarisTech made a video where he is creating a Data Science model entirely in ChatGPT! (Watch it, it is amazing, but at least scroll my article to the bottom ok?).
After seeing this video, I looked into the sky, and a bright light came out from the clouds. Right in these clouds I heard a famous Bear Gryllis voice saying “Improvise. Adapt. Overcome.”
I made a challenge and I used ChatGPT to support my work for a week. I was asking ChatGPT for code review, creating documentation, asking for direction of my analyses, asking for different algorithms, asking to polish my plots, etc. I want to share my findings.
1. My work is faster and/or deeper.
Nothing surprising here. Instead of looking for an hour, scrolling StackOverflow and Python documentation for answers, to why my code does not work I am just asking ChatGPT for it and it is finding my error immediately. I am not losing my “brain budget” for this work, I can use my new time for working deeper on analysis. I realized that before the “ChatGPT Era”, my analyses were not as deep as are right now. Usually, I was coming after days of work to quite obvious and superficial recommendations and insights. Right now I have more time to drill in the code and I am not leaving some investigations behind when the deadline comes. I just have time for that.
2. I still need to know and code in Python
ChatGPT is not a SkyNet yet. It has its limitations. It can prepare You a code for basic data cleaning, some algorithms, or help in visualizations, but in my case it made a lot of mistakes in automation, creating functions, or more complicated loops. That was a moment when I had to step in. It also had problems with aligning code with my work for 100% or sometimes was doing obvious mistakes in calculations. I still have to code and know Python — just to get more from ChatGPT.
3. Seaborn and Matplotlib visualizations are looking good at last
Making visualizations was always my pain point. I am not that artistically creative to make beautiful plots. I also was spending a lot of time writing proper titles, signing axes, and adding labels (Labels to barplot! We know this pain!). Now instead of spending half of the day preparing at least understandable visualization, I can just ask ChatGPT for that! He is doing it amazing (Okay, I am comparing him to me)! That is the biggest recommendation from my side for using ChatGPT in your work!
4. I am still learning… and learning a lot!
Without ChatGPT I wouldn’t learn that much about time decomposition algorithms, approaches, etc. I also learned how to improve my code, and filter pandas effectively.
Honestly, I have access to a lot of new approaches to code, writing it, to the algorithms, and learning them with hands-on experience!
5. I am less of a coder — more of an analyst.
Let’s be honest. Python at the end of the day is just one of our tools like Tableau, R, SQL, and Excel. Spending much time on it is cool, but it is not the goal of our work. The main reason for the Data Analyst job is to deliver recommendations, and insights, and show the state of the situation. After implementing ChatGPT in my work I found that now I am more thinking about my analysis in a “strategic way”. I ask more often questions like:
- Why do we need this approach?
- How to find more accurate results?
- How this insight/recommendation will impact our business?
Than questions like:
- What error “1” means?
- How to make my label 2px higher?
- Why my for-loop is not working?
Seriously, I feel that my job is much better now.
6. I can easier document my work and make my codes clean.
THAT IS AMAZING! Everybody likes to have documentation but no one wants to write it! ChatGPT can make You at least a draft of it! What else can I say?
And it can also change your code to a lovely, clean, and commented code snippet.
Nothing else to say!
EDIT: There is something else to say — I wrote more about it in this article.
To sum up
After one week of work with ChatGPT, I came from “Meh, just another fancy AI program” to “Shut up and take my money If You’ll demand!”
With these 6 points, I summed up my thoughts about ChatGPT and what can I say.
“Artificial Intelligence is a new electricity” — Andrew Ng (I need to make his course at last).
This article was definitely not sponsored by Google
Check out my other posts:
- What you need to know to be a Data Analyst — Technical Skills
- My start with Football Analytics
This article was published in Microsoft Power BI publication, subscribe it for more interesting Power BI Tutorials, Tips and Tricks.
