One of my favorite books is Statistical Rethinking by Richard McElreath. Why it’s my favorite? Because I start it repeatedly, typically once every six months or something.
In the book McElreath explains that teaching statistics is like teaching engineering physics backward: first teaching bridge building and ending up with basic physics. For many people who use statistics in their work, this is what they need since they don’t need very deep knowledge about the math behind statistical methods. McElreath compares this to plumbers: they are critical professionals who don’t know and don’t have to know much about fluid dynamics. For researchers using statistics in their work, this might be a problem since statistical methods are unique to the context where they are used, and it’s important to understand all underlying assumptions about the model to make good science.
Development of machine learning and artificial intelligence tools since 2015, when Google announced their Tensorflow library, has made ML/AI tools available to huge masses. These easy-to-use libraries and relatively cheap computing power offer us lovely and powerful AI tools that are useful in our everyday lives.
If AI or ML model development required a Ph.D. fifteen years ago, now you can build those quite quickly with tools that don’t need any (or not too much) mathematical background. For many of us, taking advantage of AI development requires nothing besides Google or Microsoft account, and we are ready to use powerful models. Sometimes you need a credit card.
If we consider this situation with the same metaphors from the Statistical Rethinking -book, many of us might feel they are plumbers in AI plumbing. But, unfortunately, I have bad news for you: if you are only using ChatGPT, Midjourney, or other tools and not writing any code yourself, you are not even a plumber in this new world. You are only the one who takes water from the tap or flushes the toilet.
Most of us are happy to flush the toilet after the toilet business and use AI-based tools in daily life. It’s OK. But if we want radical innovations for specific use cases, we need more plumbers and engineers who understand fluid dynamics.
Originally posted in LinkedIn early 2023, but still relevant
