One of the important lessons of the internet is that unless there's very deliberate suppression, noise will drown out signal if there's the slightest incentive to spam.
Well put. It's definitely a signal vs noise type issue, all other aspects of the problem are secondary.
They can make a dent in the problem short term with more manual review but as the spammers evolve better automation, if all Amazon has in the race are humans, they'll keep losing.
As more and more of these types of stories come out, over the course of years now, I keep expecting to hear about a big push from Amazon to clear out the noise. And it keeps failing to happen.
The knee jerk interpretation is that Amazon doesn't care about fake products because they still get paid. But, historically at least, Amazon is smarter than that. Part of their dominance comes from understanding that, long term, there is more profit in happy customers even if it that means accepting otherwise avoidable costs.
I wonder if this is the result of a changing culture or of it's an indication they've become too large and inefficient to respond to problems and iterate at modern tech company speed.
It's in the nature of the technology. If you can figure out how to do it once, doing it 1000x or 1000000x isn't going to cost you that much more. Quantity is easy, quality is hard. So we've created a selection force for "business" models that benefit from the nearly free ability to scale in quantity while ignoring quality. Ie, all sorts of spamming.
One of the reasons ML worries me is because it can sprinkle just enough of something that vaguely resembles quality into the pot to take the quantity (spam) scaling to a whole new level.
Anything that can be faked will be faked in bulk.