On AI Becoming 'The Feed'
One of the emergent issues I'm starting to see commented with AI is the averaging out of data and the smoothing out of people's content and plowing under of anything slightly outside the approved "normal".
The process is something like:
1) Company does surveys of which websites or content have good feedback or engagement.
1/x
2) AI analysis of the survey establishes common vectors of traits that correlate with site/content being more popular. But in doing so, they either:
2a) Overlook some traits that are difficult to get clean data about, and/or:
2b) Overlook sites/data that were popular but don't fit the more typical pattern (successful outliers) aren't reflected clearly in the AI model.
2/x
3) A search engine or similar content ranking system is changed not to reflect *actual* site/content popularity, nor to reflect actual engagement or business success. Instead ranking becomes based on how closely the site adheres to the AI determined pattern of "successful" traits.
So if you were previously successful but a little atypical, searches will have a hard time finding you now, unless you redo your content to be more like the "normal" pattern AI considers successful.
3/x
This guy found it out the hard way, and his previously successful business tangibly suffered because search engines changed their criteria for what was considered an appealing business:
Parf 1:
https://youtu.be/II2QF9JwtLc
Part 2:
https://youtu.be/6uKZ84zwJI0
In the last 5 minutes of this video, the woman interviewed mentions something similar that seems to have happened with a different guy's blog post:
https://youtu.be/G7-gNp8GAHU
4/x
Just kind of reminds me, yet again, of M.T. Anderson's book "Feed" and how the hyper-consumerist oriented algorithms of "The Feed" don't know how to market to you if you don't fit their preestablised patterns well. On the surface it sounds great to live outside mechanized analysis. Until you desperately need some service or business interaction and the system to provide it can’t recognize you.
5/5
Peter Kisner ≈