From AI to ML to AI: On Swirling Nomenclature and Slurried Thought (2018)

I feel this switch was largely pushed by Google marketing.

I wonder how much of the name change was to acknowledge the fact that a lot of Google smartness is actually not ML (eg Google search is mostly string matching and giant lookup tables) or at least not Deep Learning.

At the start of my career in ML, I was shivering when someone was labelling my work as AI. At first because I found it too old school, then because I found it over-hyped. But I now use this world myself because that's what people read in the newspaper.

Also, it's often hard to guess that eg the Google assistant is mostly based on regexes and small grammars, while the photo classification is actual ML.

What I'm very cautious though, is to only AI as field name. I don't speak of Google assistant as "an AI" but as an "AI powered" software.

(Ex-Googler)

It does annoy me that one small part (classifiers) has colonized the broader term that encompasses planners, reasoners, semantics, etc.

I find it interesting that this discussion takes any interest in the underlying technology, to me that seems utterly irrelevant. Machine Learning largely seemed to me to become a term used by academics and engineers to avoid having that discussion about AI, because 5 years ago if you told anyone that you worked in AI they'd think you were building Wall-e. As the ML research paid off there were suddenly some hype around it and that's where marketing came in. You actually have to sell the product. No longer was it enough to tell people you were recognizing cats on the internet - you were developing a DL ML AI CNN for feline categorization. It's like your 4K UHD HDR AMOLED HDTV.

When marketing gets involved it's actually quite positive for them to have those associations with the bicentennial man.