4 ms·
What I find most interesting about this is that a lot of supposed "smart" algorithms of Big Tech are in fact a patchwork of "dumb" rules rules and human-picked
by pembrook 2y ago
What I find most interesting about this is that a lot of supposed "smart" algorithms of Big Tech are in fact a patchwork of "dumb" rules rules and human-picked winners. This would explain why the quality of search results is failing to keep up with developments in LLMs.
This also explains why it's impossible for incumbents to unseat the winners in many search categories -- because they've literally been picked as the winners by humans at Google.
Looking at my Twitter/X feed, I also see an oddly similar dynamic. Certain accounts appear to have been manually boosted, showing up all the time -- whereas others posting even the same exact content will never appear.
Silicon valley will loudly tell you all about how wonderful they are at "democratizing," however, if you look under the surface it appears they're just hand picking the winners.
- trogdor 2y ago> because they've literally been picked as the winners by humans at Google Is there evidence of that in the leaked documents?
- pembrook 2y agoYes, it’s in the linked article.
- trogdor 2y agoI read the linked article. It doesn’t say that.
- corentin88 2y ago> #4: Employing Quality Rater Feedback
- gundmc 2y agoYou mean the Search Quality Raters that Google has written about extensively in public[1] including the 170 page quality rater guidelines doc? This isn't a secret, nor is it the smoking gun you seem to think it is. [1] - https://www.google.com/search/howsearchworks/how-search-works/rigorous-testing/ https://www.google.com/search/howsearchworks/how-search-work...