I’m not sure if this AI trend will last or not, but vibe coding is coding (writing a computer program) by simply prompting a large language model (an LLM, e.g., ChatGPT) to write the code. Software developers are concerned that LLMs are coming for their jobs, and vibe coding is understandably a serious concern if anyone can code with no experience.
But I’m not sure they should be worried just yet.
In statistics, the vibe coding equivalent is using a GUI (graphical user interface) to perform a statistical analysis. That is, a person can enter numbers into a program, click some buttons — and presto — they’ve “run the stats.” Statistics GUIs have been available for decades (+/- ?), and they allow someone with little-to-no stats knowledge to perform statistical analysis. Below is an example of a GUI.
Even with easy to use stats software run by GUIs, I still have a job.
What I find is that while some people with minimal statistics experience will perform statistical analysis, these folks are unable to perform non-standard statistical analyses or are not confident if the analysis is important (e.g., research for a high impact journal).
So I get to focus on the interesting stats problems and the less experienced can to the basic stuff. This is what I think will happen with vibe coding if it sticks around.