Where I revisit my stance towards “AI” and “ML”

The pervasive use of the term “Artificial Intelligence” to describe the surprising and hyped technology du jour unnerves me, as I think it muddies the discourse. Laypeople understand it to mean that machines are becoming human and technologists draw parallels to the human mind and cognition that are just not warranted, especially if you know how the technology works.

Similarly, I've always tried to call “Machine Learning” applied statistics, since (in my amateurish understanding) it is mostly about doing statistics with computers, at scale. It seems more respectful, more precise to me, less prone to be confused with “learning” as done by humans (which I think we don't know nearly enough about).

However, I've come to be a bit less despondent towards these terms. I think they make perfect sense when talking about a field. The field of Artificial Intelligence is indeed trying to make machine artificially do things that require human intelligence to do. The field of Machine Learning is trying to make machine learn concepts, remember facts and generalize its “knowledge” to new problems.

The fact that statistics work especially well in both cases doesn't mean that Artificial Intelligence is just statistics. It's very far from that. The field has been the source of many discoveries and innovations that we take for granted (depth first search, graph unification, Lisp, etc...). “Artificial Intelligence” is a mission statement (and an inspiring one at that, if you like machines). It is a North star never to be reached.

As such, I find it an apt description of a goal and a poor and confusing term to apply to actual solutions and technologies (which have much more precise definitions).