Just as I’m wondering why I always get these abandoned cart emails asking if I “need help” every time I’ve added something to check shipping cost and then immediately removed the item, I saw something a bit more unnerving.
The other day in a post I mentioned the Klon Centaur pedal, today in a link I’m shown an advertisement for the same pedal which I’ve never even shopped for.
I guess it’s only natural, some webcrawler is reading my posts, knows who I am and is delivering that data to advertisers, but it’s a bit distasteful to see it play out that way.
I thought I had somewhat dodged this kind of thing by choosing things like “necessary cookies only”, is this just the way it is? AI reads our posts, nothing you can do about it?
The fact that it understood the name of a rare boutique item is what’s odd to me. if I said Doritos or something I figure it would make more sense, I guess I’m just naive. I’m assuming it’s Reverb specifically that’s trying to sell me the Klon.
It could also be coincidental. There are a lot of “my phone is listening to me, I got an ad based on a convo I had in person” posts, but we focus in on the seeming correlations and ignore the non-correlations (did you shop for any of the other three items shown?).
yep. I bought an sp-202, I looked at potentiometer knob adapter rings this morning. And not whatever that guitar is, but I’ve looked at various guitar related things pretty recently.
Klon centaur is pretty specific. I don’t know the mathematical equation to define how random that ad would have to be, but it’s a pretty unlikely coincidence.
You could also have seen a mention of the Klon Centaur, perhaps last time you were on Reverb, and that put the idea of mentioning it into your head when you were reaching for an example for “transparent”.
I have adblockers where I can install them, and do most of my browsing in save-no-cookies private mode, but sometimes I see ads, and they’re generally hopelessly wrong for me. When was the last time you cleared your caches?
The Klon isn’t all that unknown. I know next to nothing about pedals beyond reverbs and delays, but I know the Klon because JHS did a bunch of videos on them. It seems there are a ton of Klones, so it’s not terribly surprising that a recommendation algorithm would suggest a Klon to someone who is interested in music gear.
Cookies are on the way out. I don’t fully understand how they work so I won’t try to explain it here, but there are algorithms that can identify you very quickly with relatively little data, and those algorithms are the basis for some of the biggest mobile ad networks.
Probably no “AI” going on here, unless your definition of “AI” is simply “cool algorithms.”
I haven’t cleared my cache in quite a while but there has been no klon on my radar, it goes for between like $4000-10,000USD and upwards. It was produced in a quantity over its lifetime of maybe 8000 and not since 2008, although it’s highly copied. super famous pedal.
No offense to whoever the famous artists are that swear by them, but I just don’t have a budget or a delusion towards a $10,000 overdrive pedal to set on “one” because of the transparency I probably have so much hearing loss I can’t tell it’s there.
is my post here from 3 days ago enough? because I haven’t been shopping for pedals on reverb. it would make sense then.
However, I will say that if AI is involved and knows that I looked at a boutique guitar, then marketed a boutique pedal to me, that would make more sense. But that kind of marketing requires more than a cool algorithm, I assume? unless it’s like “x is above value y then produce accessory value z”.
The benevolent godes of computation are trying to steer you towards a Klon, for reasons known only to them. Maybe consider picking one up? (I’ve been meaning to get one to see if it sounds good when driven by a 303-clone)
Could be. Like I said, I don’t understand the algorithm well enough to explain it.
Pedantic mode: there is nothing but cool algorithms in computer science. (there are also boring algorithms, but that’s really in the eye of the beholder). The class of things called “AI” is an ever-changing subset of the general field of algorithms.
The funny thing about all this “AI” stuff is that most of it is just building a point cloud in a high-dimensional space (picture a spreadsheet with tens of thousands of rows) and then finding an optimal surface that divides the points into two sets such that they satisfy the training criteria. It’s really stupidly simple, but giving it massive amounts of data yields interesting results.
In many cases, a more traditional algorithm yields similar results more quickly. Neural network methods are a kind of brute force approach to problem solving that is interesting because it is so widely successful. Massive optimizations are possible, and frequently demonstrated.
TL;DR: actually, yeah, you can probably get the results you are seeing without AI
(looking forward to @plragde’s takedown, since he is much, much closer to the theory than I am. Hopefully I haven’t embarrassed myself too much here)
That is less than a cool algorithm, but it’s probably what is happening. Not highly targetted, not spying on you, and not what I would call AI (I am old enough to remember Eliza in the '60’s and '70’s, a few hundred lines of pattern-matching code that fooled a lot of people at the time, and probably still does).
Superpedantic mode: I was an “algorithms and complexity” researcher for the first part of my career. Most useful algorithms are boring, and the cool ones aren’t that useful. But there I use the word “algorithm” in its original sense, rather than the new sense of “something I don’t understand and the implementors didn’t either, but they put it into production anyway”.
Joseph Weizenbaum, who programmed the original ELIZA in the mid-60s, once found his secretary chatting with it. He said, in effect, “You know I wrote this and there’s nothing to it, right?” and she said something like “I know, but I find it comforting.” Sometimes we really want to be fooled.
When I see marketing offering me things like a monomachine or other items I’ve actively looked at, I always assume this is part of the tracking that is left behind by cookies. The scope of the coincidence is odd, but if I’m to understand it as “this is a commonly sought out boutique item by those who looked at another boutique item in a similar category” the window of coincidence doesn’t seem quite as narrow as before.
It feels like if they were firing blindly in the dark, there wouldn’t be anything to speak about or to question in the first place, someone knew how to make the best of their advertising and implemented a strategy based marketing that could roughly be labeled as effective in generating content.
But it certainly wouldn’t have seemed so pertinent had I not made mention of this random pedal I wouldn’t bother talking about ever, just 3 days before. The other marketing in that pop up was of no interest because I knew why it was there.
Essentially, the “algorithm” is smart enough to throw pitches roughly in your strike zone.
If you sat atop a synth shop in a heavily foot-trafficked area with a good set of binoculars, you might notice a pattern. As some people walk by, they do a double take when they recognize the synth shop for what it is. 80% of the time, they eventually turn towards the shop and walk in.
If you yell into a megaphone “hey, you! in the red shirt! I know you want to check out the synth shop, just go right in!” the dude might reasonably surmise that you have been stalking him for months and know of his secret synth addiction.
If you are into the nuts and bolts of retail, this is probably a very interesting time to be in e-commerce. The last 20 years were mostly about building a minimally reliable and functional storefront, but no one really replicates the quality sales person experience online. Search is getting radically better, and that will have profound effects on media and music.
