SoylentNews is people
SoylentNews
from the now-let's-work-on-human-intelligence dept.
A recent Wired article tells us about the progression of the Amazon product recommendation algorithm.
Amazon helped show the world how machines can learn. As far back as the late ’90s, the company’s online retail site would track every book, CD, and movie you purchased. As time went on, it would develop a pretty good sense of what you liked, serving up product recommendations its code predicted would catch your eye.
It wasn't rocket science. It was an algorithm. But it worked. And in the years since, the field of so-called machine learning has evolved in enormous ways, with the likes of Google, Facebook, and Microsoft training enormous networks of machines to identify faces in photos, recognize the spoken word, and instantly translate conversations from one language to another.
On Thursday, Amazon unveiled a similar machine learning service, pitching it as a way for any business to use the AI tech the company has spent years developing inside its own operation. Known as the Amazon Machine Learning Service, it’s designed for software developers “with no experience in machine learning,” AWS head Andy Jassy said on stage at a mini-conference in San Francisco.
(Score: 3, Interesting) by Aichon on Saturday April 11 2015, @06:55PM
Don't get me wrong, I'm glad Amazon is making something like this available, since it'll spur others to do similar things, but of all the major sites I use that have recommendation engines, Amazon's engine is easily the least useful, despite my having provided it with ample data.
Buy a run-of-the-mill boxed set for a film trilogy, and you'll see the individual films, the ultimate collector's edition boxed sets, coffee table art books, video game spin-offs, video game guides, vinyl figurines, action figures, Legos, and soundtracks recommended to you. Never mind that you've marked "Not Interested" on every single art book, game guide, figurine, action figure, or toy that they have ever recommended to you, simply because those aren't things that interest you or that you've ever been interested in purchasing. And never mind that they know from your browsing history that you chose to purchase the cheap edition instead of the high-end one, so they're wasting your time with the recommendation of the collector's edition boxed set. Or that the individual films sold separately add nothing of value to what they know you already own. Oh, and that soundtrack? You actually purchased it in the same order with the boxed set, but because it's gotten a reprint with a new bar code in the last year, they think it's something new that you don't have, so they're recommending it to you again. Even worse, if you rate the boxed set as 2/5 to indicate that you didn't like the films, you'd expect that they'd stop recommending stuff in that vein. Instead, Amazon will fill your recommendations list with similar films of similarly crappy quality, and it will keep doing so, despite your marking each and every one of them as "Not Interested", until you tell it to stop using the original item for recommendations.
Not that I see any of that on a regular basis, of course. Though I suppose I bring it on myself, since I check my Amazon recommendations regularly and prune out things I'm not interested in, with the hope that it will get better. It hasn't. For every 1000 items I mark Not Interested in, I'd say there's five I already own and can rate, and just one that catches my eye.
Contrast that with Netflix, which I'd peg at something more like 1 in 3. Netflix is hardly perfect, but it manages to dig up a lot of stuff that I'm interested in, and on those occasions where I use their Max virtual assistant [techcrunch.com] on the PS3 version of the app, which leads to my watching films I've generally never even heard of before, I find that I'm pleasantly surprised far more often than not.