GPT-4 and the future of Data Analysts
The release of Chat GPT-4
In mid-March, OpenAI released GPT-4, the latest milestone in OpenAI’s effort to scale up deep learning. People were in awe of what it could do, from drafting lawsuits to passing standardized exams, to building a working website from a hand-drawn sketch. Chat GPT-4 is not the evolution of ChatGPT 3.5. It is an exponentially more sophisticated generative AI. To create this new version, the engineers at @OpenAI completely rebuilt their deep learning stack and, together with @Microsoft Azure, co-designed a supercomputer from the ground up to handle the workload!
Working with Generative AI as part of @MobyAnalytics’ efforts to bring together the business and data analysis, I also decided to “rebuild” my views on generative AI, based on what I learned diving into the advanced analytics plug-in available for GPT-4. Here are some of my thoughts on it so far:
OpenAI, the creator of GPT-4, admits that, while GPT-4 is their best generative AI, it is still “far from perfect,” as the company puts it. One thing, however, that caught my attention is how you can intertwine GPT-4 with programming languages : while the technology seems young, the fact that this version can easily output graphical analysis, as shown in the video below, makes it evident, to me at least, that there is an imminent shift of paradigm on the way we interact with data. Both data collection, treatment, and visualization will be vastly different tomorrow than today. I predict that these data analytics subfields will be blurred, if not blended.
In this quick test run alone, I could create a graph from raw data and then have GPT-4 explain it to me as if it were a financial analyst. What would normally take a team of people with different skills to do took me seconds, sitting at home. No meetings, email exchanges, misunderstandings, and, above all, no need to segment and treat the information from one step to the next. A seamless creation of a graph from raw data, instantly.
Interacting with GPT-4
Another point that really gave me pause is that, throughout my interaction with GPT-4, my commands were pretty much a conversation. The most technical part of my instructions was to ask GPT-4 to change the color to ‘#CEE2F’ - although, I am also pretty sure GPT-4 would understand “hint of blue” –if I had told it that. The accessibility of a natural language tool cannot be overstated: anyone who can write in English (or in other 50 languages, including Spanish, French, German, Chinese, Japanese, and Arabic) can create with GPT-4.
Does that mean, however, that the age of technical language is over? I don’t think so. For the task, I asked GPT-4 to develop colloquial, daily English, which was more than sufficient. However, a more complex task would require more precise instructions. I think this can only be done with, at the very least, a blend of natural and technical language.
So, if GPT-4 is so good, and if it can do the job of 4 different people in seconds, using only input in natural language, why do I even bother having a data analytics company in this day and age?
Am I not worried that GPT-4, or 5, or 6, or whichever ones come after that will also replace my company?
Not at all.
My view is that this is just a paradigm change. The role of analysts will change, but the demand will only grow. GPT-4 is a tool, and, just like the invention of the electronic spreadsheet by Allen Sneider, this only means that people working in our field will have to learn and adapt. Once this, no doubt challenging, adaptation period passes, a much broader range of talents can appear, as technology makes it easier to “technically” interact with data.
Programming language is still crucial
The same is still valid for learning programming data, in my opinion. I recently saw a video of someone who claims to have re-created the game Flappy Bird only by prompting ChatGPT to write code. And although this is really amazing and quite fun, the truth is that this person knew what to ask from ChatGPT to get these results. This means that learning a programming language is still crucial, if not for the syntax, to equip ourselves with a structured way to think about data and the underlying problem. Many people can drive. Do they all know how cars work?
At Moby, we are obsessed with bridging the gap between business and data analytics. Moby was born based on our certainty that, although data analytics has so much to contribute to businesses, its potential is still absolutely undermined because most business leaders don’t know how to use data when making decisions. Many don’t even know how to use their own data to their benefit, and we want to change that.
GPT-4, and this generation of generative AI, is a great opportunity to bring a better understanding of data to the business world and the reverse. As professionals in the data analytics industry, our job is to guide businesses to this place where they can understand such a powerful tool and shape it in a way that helps our clients.
Our clients know what they need, generative AI can deliver it, but who knows how to ask?