💾 Archived View for gemini.ctrl-c.club › ~lettuce › recommendations.gmi captured on 2024-08-18 at 17:32:17. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2021-11-30)
-=-=-=-=-=-=-
2021-06-04
Growing up, in high school in the late 90s, I would take a train downtown a few times a week to go to South Street in Philadelphia. There were 5 or so record shops (well, mostly CDs and some tapes) that I would regularly visit. In particular, I loved Repo Records for international music and ocassional noise and jazz albums; Spaceboy for the latest electronic music; Relapse for noise; and there are others whose names escape me. I spent $10 to $25 a week (a lot of money for me) on albums, and would listen to tons of previews. I took trips up to New York to visit Tower Records and Other Music near Union Square, a 5 minute walk from each other, and conveniently near The Strand. There would always be listening stations at Tower and Other Music, and every store had staff selections and recommendations out. A significant part of my life was learning about new music from alternative papers (The Wire Magazine, Bull Tongue in Arthur Mag) as well as these listening stations, staff selections, and the stickers with staff notes on the used and new albums. More importantly, at these places I could talk to someone who worked at the shop. I liked Merzbow or Sun Ra or Red Snapper or No Neck Blues Band. Who else would I like? I had a really strong sense of my interests and 'bands', but it was nice to learn about new music. Listening and searching for new music took up much of my life at that point. I didn't buy music online, though I did use napster. But what if, unlike me, you weren't a teen with dozens of free hours every week to gratify your desire for new music? How could you get it?
Let's take an aside to talk about the birth of 'recommendation engines.'
Recently I listened to a Planet Money podcast episode where they interview Doug Terry, "Distinguised Scientist at Amazon," inventor of the recommendation system. Back in the early 90s he was working at Xerox-PARC and was already getting too much email a day, which he said took him about an hour to get through (the irony!). He created a system where you could mark certain senders as low importance, high importance. He also added in automatic detection: group emails -> low importance. Emails from one person just to you: higher. Etc.
Runaway Recommendation Engines episode on Planet Money podcast
With this system, meant to sort large inboxes automatically, over time they added in that my choices for higher priority emails would be used by other people on the same network as well. In the 90s, the first online place I purchased things through was eBay. Occasionally I'd click through to look at other items sold by a seller. They could make store pages or shops, with info about them and a list of their inventory. As I started to use Amazon to buy books (perhaps back in 1999 or so?), I remember over time their list of suggested books. I think it was somewhat useful to get a sense of books on similar subject matter, but I don't remember purchasing much this way.
In the 2000s Netflix ran a million dollar contest to improve their recommendation engine. They published anonymized data of tens of thousands of users so that researchers could use the data to improve on their recommended movies. This kickstarted a worldwide interest by companies in using recommendation engines. But this doesn't trigger any positive association for me.
I was never a big Amazon user but stopped altogether by 2015. I haven't had a Netflix account since then and perhaps longer. Ocassionaly I'll be at a friends house to watch a movie and open it. I'm always disappointed by the suggested movies. When I had a subscription, the suggested movies were always frustrating: mass market schlock that held no appeal. At best they'd suggest a generic category like 'foreign film' or 'documentaries' or 'science fiction'. That's their precious recommendation engine? Please. Today there's probably some form of machine learning on the backend. These algorithms absolutely do influence selection of movies (or products, depending on the website). But if you're offered barely meaningful options from piles of (garbage) selections is that recommendation meaningful? To many people (who would certainly resent my language previously), I suppose it is. It's certainly 'convenient' in a certain sense. But it's a far, far cry from the chalkboard recommendation boards at video stores and bookstores, and certainly much less meaninful than the helpful employees at these shops.
As I try to think of alternatives, for films I've found that I've reduced my intake. I do occasionally read reviews that help me make choices, or get recommendations from friends, but overall my watching of shows and things is way down. For music, I've gravitated to Bandcamp for finding lots of new music. They publish blog posts with embedded listening, and descriptions and recommendations. It's a digital version of the record shop recommended listening. And it works well for me. These are posts written by people, and the recommendations are meaningful to me. They're full of weird music selections, not the digital equivalent of Disney. Beyond that, I don't shop much online and can avoid the un-useful recommendation engines for the most part. Online book sites exist for people to recommend reading selections. I like to read blog posts occasionally of suggested books, and I read book reviews. And I read e-books from the library or visit small bookshops here in NYC now that they are open again in the pandemic. Don't get me wrong: I'm not trying to argue that recomendation can't be useful to someone for certain tasks or needs; I just haven't found them useful for me. And so 'blog posts', word of mouth and recommendations from people are my favored way of finding new things I'll like.
---
Comments can be emailed to lettuce@ctrl-c.club