This problem exposes an aspect of the AI difficulty chasm between Amazon’s dream of assigning a unique barcode (ASIN) to every object under the sun that has bee

This problem exposes an aspect of the AI difficulty chasm between Amazon’s dream of assigning a unique barcode (ASIN) to every object under the sun that has been or ever will be sold, and Amazon’s dream of doing so without spending a single dollar on expensive human curators. Amazon is trying to solve a well-known library science problem using AI, and it’s been a miserable failure. Their AI systems can’t detect when sellers bait and switch because there’s no way to detect this that doesn’t have a massive false positive rate, and they aren’t willing to spend a single dollar on ‘app store’ reviewing product changes by third parties. Another demonstration of this chasm can be found by searching for ‘claritin’. There are pages of duplicates, terrible and missing product photos, and a wide array of metadata errors due to human inconsistency by the sellers listing the product on the marketplace. Fixing this would require human curation, since their AI curation is incapable of doing it, and so Amazon simply allows the rot to spread unchecked.