4 ms·
I use ublock origin and privacy badger not because I am worried about privacy but because the internet is basically unusable without it. Because of this, I don
by jccalhoun 9y ago
I use ublock origin and privacy badger not because I am worried about privacy but because the internet is basically unusable without it.
Because of this, I don't see many ads. But I have been an amazon customer since 1999 (according to what they say on their website when I'm logged in.) Looking at what they recommend for me, this personalization stuff is crap.
In music, Amazon recommends bands I never listen to like Montrose, Metallica, and the Doors (and to be fair, some people I've never heard of so I guess it is possible that I would be interested in them. Greta Van Fleet? William Patrick Corgan?)
In books, I do like scifi but they recommend a bunch of books with spaceships shooting each other on the cover - not what I have ever been interested in.
In the "humor and entertainment" section of books they do list some books that I would be interested in but, strangely, none of them are "humor" but are all academic books about videogames (which I am interested in). Even here the recommendation engine is very unsophisticated because in between academic books on videogames there are books on the art of Zelda and other coffee table books that I am not interested in.
And the first book in their recommended children's book section is 1984. (and I don't have any kids any way).
If this is the best they can do with 18 years of tracking my purchases then I am not worried.
- bambax 9y agoIt's really amazing how bad those personnalisation engines are; AI seems pretty stupid for now. But as I was making the exact same point a few days ago here on HN, someone responded to say that maybe it was on purpose, that if recommendations were too good they would creep us out. Don't know what to think of it but I found the objection interesting...
- black_puppydog 9y agoanother point is that you (the recommender, and, incidentally, also the user) are happy if there is at least one novel and interesting item in the recommendations. This might apply less to music (list) recommendations, but in general that's how I (as a customer) also approach lists of recommendations.
- komali2 9y agoTo be fair, you sound like one of the harder people to track accurately. However, I as well have been "disappointed" by the ability of websites to judge my interests. After reading about cases such as Target, who have "spooky" ability to gauge interest, I was expecting better. So I searched for a very specific motorcycle jacket with very specific features and now my page is inundated with every clothing item that has the tag "motorcycle_jacket?" That's.... it? The same result I'd get for hitting google with "site:amazon.com 'motorcycle jacket'" ?
- Chaebixi 9y agoThe impression that I got was that Target thing wasn't some spooky self-learning AI product recommender, but rather they had the data to write some special-purpose analytics to detect a specific but common purchasing pattern. Your motorcycle jacket example isn't the kind of thing that would get that kind of special attention.
- komali2 9y agoReally? Because I would expect Amazon to be like "oh fuck, this dude bought a motorcycle! Inundaaaaate!" and start seeing a bunch of ads for Senas, helmets, gloves, etc.
- bluGill 9y agoUnless you buy your motorcycle from amazon as well they don't know to send you that. Target was able to get the spooky results because there is in fact a real correlation: girls who switch from scented products (soaps) to unscented are very likely to buy a maternity clothing in a couple months, and a few months latter baby products. Note that this all starts because then girl was buying scented products at target to begin with, and so the change in habit was the important part.
- edgarvaldes 9y ago>To be fair, you sound like one of the harder people to track accurately. But Amazon, in this case, doesn't need to track other surfing habits, just his purchase history. That history is not affected by privacy addons.
- wutbrodo 9y ago> In books, I do like scifi but they recommend a bunch of books with spaceships shooting each other on the cover - not what I have ever been interested in. You're making a huge mistake by judging a book by its cover in this case. The copy of the Foundation series that I had as a kid was very space opera looking too, and I've seen gaudy covers on everything from Dune to Kim Stanley Robinson. For Sci fi in particular, publishers have an incentive to trick a large chunk of the audience into thinking that its Star Wars-y, and there's not much incentive in signaling the things that you or I would get out these books.
- gknoy 9y agoI second this. Heck, the Ancillary series has paintings of spaceships as their covers, and they're some of my favorite scifi books in recent memory.
- tomjen3 9y agoI block far less than you do, but all recommendation engines appear to be crab: I brought a few books on Amazon that could be purchased by middle-aged women and now most of the books I get recommended are romance novels with a vampire, time travel or magic theme - meanwhile what I actually buy is mainly SciFi and a bunch of books about NK. Come to think of it the recommended movies on Netflix is also crazy, but they may be screwing it towards their own selection. But lets be honest even when I tell Facebook what my interests are, it can't give useful ads - and even mighty google assumed I was interested in Palaeontology at one point.
- leeoniya 9y ago> William Patrick Corgan? Billy Corgan was the lead singer of Smashing Pumpkins. also has a lot of solo stuff. seeing the full names rather than stage names of artists can be weird.
- jccalhoun 9y agoapparently he has just released a solo album under his full name. weird. Smashing Pumpkins was never a band I was into but at least this recommendation is somewhat understandable
- koyote 9y agoI have pretty much the same amount of history with amazon (2000) and it's even worse here. Most of the recommendations are items that I actually already purchased through them ("How about a second electric razor or some more batteries?") or items that I literally was just looking at while browsing ("We recommend these jeans that you just looked at"). Utterly useless.
- pqh 9y agoIt's hard to be worse than YouTube. "Here are some videos you've already watched, mostly that you've already liked, but also a few from the same channel, and even a helping of some you've disliked but which other people have uploaded duplicates of." And that's Google, the AI experts, with a crapload of data on my viewing and others' viewing habits, including in depth learning features such as length of time spent on videos, liked videos, playlists, etc.