Bullshit AI: the praise of the "Recommender Algorithms with Machine Learning"

 Just a short thought after reading for yet another time the bullshit praising of how amazing and "machine learning" are the recommender algorithms in social media etc. from the abstract of a recent Pedro Domingos' work: https://scholar.google.com/citations?view_op=view_citation&hl=en&user=KOrhfVMAAAAJ&sortby=pubdate&citation_for_view=KOrhfVMAAAAJ:T_ojBgVMvoEC

"The Master Algorithm Revolution" ... Let's go to Amazon. With buying a book or Netflix to watch a video, and the automatic learning system recommends us titles we may like. Facebook uses automatic learning to decide which updates to show, and Twitter does the same with tweets. Whenever we use a computer, it is more likely that some automatic learning is involved in the process."  

The system recommends you movies with the same leading actors, directors, genre; or the same or a set of fuzzy matched or similar to what other users have clicked after this one etc., the movie data is listed and available, annotated etc. This could work experimentally in the 1950s and very well on the 1960s computers (as much as fits in the memory):  string matching to keywords (lexer, simple parsing) and fuzzy matching, and exact or fuzzy sequence matching (allowing partial mismatch for selection). 

The tokenizers with transformers and "decoders" of the transformers output are actually doing something like that, LOL, however with a gazillion of examples and FLOPS.

One consequence is how trivial what counts as intelligence actually is, which is true either, as suggested in the "antiquity" and more recently by this media as the Sacred Computer & Artificial Mind: Intelligence is simpler than it seems: (e.g.), the literal "AGI is way simpler than it seems.", Todor in AGI list at 27.4.2012 in 

"Chairs, Caricatures and Optical Illusions..." ( or AGI Digest, or Chairs, Buildings, Caricatures ... (cited with various titles)

https://artificial-mind.blogspot.com/2017/12/capsnet-capsules-and-CogAlg-3D-reconstruction.html

https://research.twenkid.com/agi/2012/AGI_2012_Chairs_Caricatures_and_Object_Recognition_as_3D_Reconstruction.pdf 

Also:
https://artificial-mind.blogspot.com/2013/08/issues-on-agiri-agi-email-list-and-agi.html

https://artificial-mind.blogspot.com/2021/07/todors-comments-on-article-ai-is-harder.html

PS. There were such "automatic recommender systems" in the local videotheques and libraries: the items were spread in sections: "comedy, action, p*rn, educational, animation, ...". When you take a movie from the "comedy" section, e.g. "Dumb and Dumber", it's likely that this "recommender system" will notice that you liked it and will 'recommend' you the title next to it, e.g. "The Mask" (with the same leading actor). If you "click" on another section (touch the case of a VHS videocassette in the next section, titled "Action"), the "recommender system" will recognize that you have changed your preferences and will now "recommend" you "Rambo 3", "American Ninja", "Commando" etc.

In another occasion, when you return a movie (you didn't forget "to rewind it"), another subsystem of the "automatic recommender" will activate - you're in line after another "user" who you know and like, who's returning a movie. You ask her "Did you like it? Would you ... oops recommend it?", she says "Oh, yes, amazing, wonderful, I am sure you will like it, we used to have a similar taste for movies..." - yes, you've watched movies together, especially when you were "sweethearts" for a month in 7-th grade. That "review" counts as "5 stars", a "sentiment analysis", it appears in the list of recommendations - and you "click" it. "Maybe we should go to cinema together again?" - you dare to ask. - "Oh, sorry, I have a boyfriend" - she cuts your request. Oops... Her "weight" for the calculation of the "recommendation machine learning system" is reduced and the next time the "engine" won't use her "reviews". However you're 9-th grade, you'll meet other "users" in the weekend!