I enjoy taking ML work and pondering what it might look like in the context of human education.
Imagine human physics education with meta-learning characteristics. Having to classify problems. No more "here are numbers for solid Argon... apply the ideal gas law" clueless plug-and-chug. Developing a sense for reasonable values. Encountering unfamiliar problems. Thus developing skills of rough quantitative reasoning, and of system decomposition and characterization. Encountering descriptions of unfamiliar problem domains. And having to extract understanding from them.
With human science education at present, even correct labeling and baseline models, let alone transference, are distant, distant dreams. But even thus shackled and buried, visualizing dance might have value, if it supports improved recognition of opportunities and preparation for escape.
I was disappointed the article was about ML and not about human learning.
When I was young, I got by on picking basics up quickly. Now I'm in my forties and I'm having to actually learn how to learn because there are no simple basics (in my areas) left (or I'm just old and the basics no longer seem so basic). Everything I read/listen/watch is a collection of boring rehash...and then I'm suddenly behind because I missed something while skimming the material I thought I knew.
Me too. Mid 40's and trying to figure out how to make something stick, especially with stuff I had learned years ago (linear algebra especially).
I have recently come to the revelation that the notion of "rote" that I had always eschewed in favor of "a true and deep understanding" should no longer be so eschewed.
I've recently adopted Anki (that other people love so much on this forum) as a way of forcing me to catalog forevermore the inherent vocabulary of things that cannot stick without internalizing. I now have an Anki board for Rust, Linear Algebra and APL. Fingers crossed it will work.
Another meta-learning technique I have found extremely useful is creating an ontology map. The act of copying down a term and relating it to others spatially gives me so much more context. Here's a nasty one I did seven years ago: https://i.imgur.com/MYCsl7F.png
I've given presentations on "Human Learning" (specifically titled so as not to be confused with ML!) that might have some interesting tidbits. Recordings of two of the presentations are here: https://ted.dev/learntalks. Slides from the talks are also available on my web site: https://ted.dev.
The main sell is that it allows the robot to quickly generalize to new domains (surroundings and objects being manipulated), albeit not tasks.