I feel your pain. I’ve experienced this frustration, too. Over time, though, I’ve come to understand perhaps why this is happening. I’ll share my thoughts with you so that you may have some sympathy for the next person that uses these terms incorrectly. For brevity, I’ll skip some historic details and cut to the chase.
It used to be so simple. “Artificial Intelligence”, AI. What does that mean? A computer that could think like a person. Papers were published, programs written, startups funded, and public excitement created.
Then, those papers stopped coming with more advancements. Programs reached their limits, and startups failed. The public wasn’t updated with these failures. We kept thinking that progress was continuing on the AI front. But, those people who handed money to those companies to build “computers that could think like a person” remembered and stopped funding. Cue the first AI winter.
So, now funding has dried up and “Artificial Intelligence” has earned a bad reputation. What do people who pursue that work do next? Well, they re-label things. It was no longer “Artificial Intelligence”. It became called “Expert Systems”, or “Rule-based Production Systems”. Eventually, these name changes include “Machine Learning” as researches focus deeper on the parts that they are able to get to work, rather than the parts they have not been able to make work, after yet another AI winter.
Simultaneously, the same thing was happening with “Artificial Neural Networks”, ANNs. They went through a labeling-evolution from “Perceptrons”, to “ANNs”, to “CNNs” (Convolutional Neural Networks), to “Deep Learning” popular today.
Each of these labels highlight a more specialized area of work that shows some success, enough to entice a new generation of investors to fund a new generation of startups. “Machine Learning” which is an umbrella term that covers statistical techniques traditionally used in the sciences and is quite hum-drum, to complex neural networks that work on a specific use case. These become easier problems to solve then creating, “A computer that could think like a person”. Which makes it easier to convince others that the work can be turned into a product which can be sold. Suddenly, that becomes a buzzword.
To the uninitiated person, even those that I’ve met who are extremely intelligent and highly technical experts in many areas of science and technology, all of these terms are extremely confusing. They are still working off the belief that work is being done on Artificial Intelligence. They’re neither interested nor informed about these sub-fields. After all, the common layperson only cares about the idea of getting computers to walk-and-talk like people.
To me, these are all distinctions without an important difference. True machine intelligence won’t come about through the silo’d work on separate sub-fields, but through a holistic solution. I think most laypeople have the same impression. So, even trying to navigate the reason for these new terms becomes cumbersome. When that happens, then the terms start being used incorrectly.
So, let’s cut them some slack. Precision of language is important, but sometimes it is more important to relax some constraints to get work done.