Those of us who were born right on the cusp of the digital age bear a unique distinction — we can remember what school was like both before and after ubiquitous internet access! As a kid, I spent countless hours in the school library receiving instructions on how to find the right research books to answer a question, what to do if they weren’t available, and how to cite properly when you inevitably had to give them back. In a few short years, all of that information became irrelevant (much to our Librarian’s chagrin) as students everywhere learned to rely more and more on search engines. Of course, these resources weren’t perfect, but they completely revolutionized our access to information — turning laborious processes that sometimes took hours into a matter of a few quick key strokes.
If you had asked me a couple years ago, I probably would have told you that this solution was about as close to perfect as it could get. And it was, until AI powered language models came along.
I could go on and on about the implications of this new technology, and I plan to write future posts about how precisely to engage with it in order to get the most bang for your buck. For now, though, I’d like to share some thoughts on a few ways that these tools — specifically ChatGPT in my case — can help revolutionize your problem solving process once again.
Learning by Chatting
When someone who is new to using AI based language models asks me what is so special about them, the word I usually come back to is ‘dynamic’. As I mentioned earlier, Google is fantastic as a search tool that enables you find relevant information from an ocean of available resources in a matter of seconds. The problem, though, is that in many ways the results you find are inherently static in nature. Even if you’re lucky enough to stumble on a recent article or forum that is relevant, good luck getting a reply to a follow-up question that is both prompt and accurate. ChatGPT is dynamic in that it can engage with you in real time on the exact specifics of your question. It is not perfect either, but it enables you to approach a topic in a way that is much more natural to how we learn in real life — through conversation.
Think back to the last interesting conversation you had with a friend — did you state a few words of interest and then sit back while they pontificated uninterrupted for ten minutes? No, because unless all your friends are academics, that isn’t the way people talk. Any natural conversation is littered with interruptions — questions, clarifications, objections… all of these serving to underline the broader point and ensure that both parties are on the same page. This is where ChatGPT shines. It engages with the user in the language and context they’re familiar with. It allows for detours, side-bars, and as much “double-clicking” as is needed to drive the point home. And speaking of detours…
Diving Down the Rabbit Hole
You’ve probably heard the term before that someone has “gone down a rabbit hole” during a session on the internet. This always bothers me a bit, because the implication is almost always that going down a rabbit hole indicates a lack of focus on a topic, and is a complete waste of time. I disagree!
Let’s say you want to learn about the Roman Empire, ok great! So you start with Julius Caesar. Well to understand Julius Caesar you probably need to know a bit more about the Roman Senate. To understand the Roman Senate you probably need to know a bit more about the city state of Rome itself. If you’re going to learn about Rome, you really need to understand the history of the Latin speaking peoples, but you really can’t tell their story without the Etruscans, and don’t even get me STARTED on the Etruscans without the Phoenicians … do you see what I mean?
The problem with the internet as we’ve historically engaged with it is that in order to get to exactly where your next question lies, you often need to click and scroll through quite a bit of less relevant information first. A 20,000 word Wikipedia article can be incredibly intimidating, and CTRL+F only goes so far. What ChatGPT allows you to do is explore a specific line of questioning without losing the thread on the broader conversation. You can follow the exact pattern of inquiry I outlined above, and then easily circle back to Caesar to continue where you started. Better yet, you don’t need to puzzle over why and how you wound up reading about the origins of the Phoenician alphabet. Your entire chat history is right in front of you to go back to at any point.
Modifying the Slope of your Learning Curve
If you’ve ever worked at a small company before, you’ve doubtless heard a version of “to be successful here you need to wear a number of hats”. At its essence, that phrase translates to “we don’t have the luxury of employing specialists everywhere … so you better be comfortable learning new skills constantly”. For example, you might have been hired as a financial analyst — only to quickly find out that the data you need to build your models requires additional work in SQL upstream. If the data team is swamped (they usually are) and your boss needs that report ASAP (they usually do), it’s probably a good time to learn some SQL!
In a perfect world, your Company would have existing training modules to get you up to speed on exactly what you’ll need to do along with instructors that have dedicated time to answer your specific questions. In the real world, though, you will just have to figure it out … so what do you do?
Earlier in my career, when I was going through this exact scenario (see here for more), I had to rely on Google and the goodwill of my graciously patient colleagues on the Data team. While this worked, it was extremely frustrating and often nerve wracking. Google search has a habit of being relevant enough to get you started, but not quite specific enough to get you over the finish line — especially when you’re learning something totally new. On the other hand, your colleagues might know exactly what to do but as we’ve already established … they’re busy with their own stuff! Furthermore, if you ask too many questions (especially dumb ones, yes… those exist) you’re eventually going to exhaust that goodwill.
This is precisely the use case where Chat GPT comes in handy. Let’s go back to the example of learning SQL. Say you want to pull some basic financial metrics from your Company’s core transactions table — you might ask Chat GPT something like “help me write a very simple query to pull revenue and order volume by year”. Well, when the result comes back you might realize that it doesn’t make a lot of sense to you why the ‘SELECT’ comes before ‘FROM’. You can respond to Chat GPT — “let’s pause for a moment, can you explain to me in basic terms why the statements in a query are ordered like that? Will it work if I invert the order?”. Depending on the response, and how helpful it is to your specific question, you can continue to dig in and dissect the fundamentals of this query for as long as you’d like. Once you’re satisfied, you can take the results from the chat and pop them directly into your SQL instance. If there are any errors, you can copy and paste them directly for troubleshooting. If the results are not quite what you were hoping to see, you can explain in plain English how you’d like the output to be modified and Chat GPT will make the necessary adjustments to the query. Now imagine going through this same exercise without this tool. You’d have to begin by googling to find an example of a query that likely wouldn’t match your needs exactly. Then, when you had some questions, you’d have to conduct multiple searches on each individual aspect or spend some time bothering a co-worker. Chat GPT can save you time, speed up your learning process, and preserve that workplace social capital for truly important questions.
Some Words of Caution
If you can’t tell by now, I think ChatGPT is a pretty fantastic tool. But it’s still just that, a tool. It isn’t magic, and it is not something you should be relying on with 100% confidence. It’s a complex language model designed and trained by humans, and the technology isn’t perfect. In the same way that Wikipedia has its faults, and in the same way that any article has its biases, ChatGPT has some flaws.
So I’d be remiss if I didn’t close this out with a few words of caution:
- Be skeptical — don’t be afraid to probe further if something seems off
- Do not, ever, copy and paste large sections of text / analysis for important work purposes without double checking. Would you give a new employee’s first draft to your boss sight unseen?
- Do not rely on ChatGPT for legal or medical advice (I mean… duh)
- Do not input sensitive data about yourself, customers, or your product into the model. I’m not implying there are specific security concerns … but you can never be too careful!
Conclusion
With that out of the way, do yourself a favor and start messing around with an AI Language model if you haven’t already. I have a hunch that in a short span of time, you’ll be wondering how you ever made it through the work day without it!
If you liked this article, you might enjoy one of our other blog posts on breaking into the world of data analytics. And if you have more specific questions, or need help with analytics on a project of your own shoot me a note at james@southshore.LLC .
About Us
South Shore Analytics (SSA) is an Analytics Consulting Firm, co-founded by James Burke and Nick Lisauskas, both highly skilled professionals with more than a decade of invaluable experience in the field of Analytics. Their shared passion for data-driven insights and business optimization led them to establish SSA, aiming to provide top-notch services to various businesses, irrespective of their size or stage of development.