4 ms·
I don't think people outsource their thinking to machines. I think the algorithms just work well at finding similar things to what people choose to consume or
by Noos 9y ago
I don't think people outsource their thinking to machines.
I think the algorithms just work well at finding similar things to what people choose to consume or like. They aren't perfect, but there's this need to explain things by this great conspiracy of fake news and invisible influencers in a paranoid style.
I think this has gained traction due to a crisis of the knowledge class that happened due to the rise of populism via trump and brexit. They can't understand that those sentiments were chosen because people agreed with them; they had to have been influenced and their free will taken by "algorithms." This to me is like saying my free will is taken by the suggestions Amazon shows me on their site, or that my taste in movies is shaped by Netflix recommendations.
This gets annoying because it's so obviously a reaction in that particular style of contradicted elites. We see it in many forms, like the handwaving over internet comments and "civility." The knowledge elite hates to have its power or dominion challenged, and will try to explain the doing so via these systemic solutions.
- basch 9y agoit confuses what i liked in the past with what i might enjoy in the future. or what might feel familiar or what someone like me liked. how am i ever supposed to have new foreign experiences if recommendations are based on past experiences or grouping me into demographics to make me more like others? sure everyones a mimicry or pastiche of others, but its losing its randomness and becoming more uniform and standardized.
- Houshalter 9y agoA perfect recommendation algorithm would be optimal in the exploration/exploitation tradeoff. Suggesting new things with some probability to see if you might like them or not. Most recommendation algorithms today are way too far on the exploitation side of the tradeoff, but it's not a requirement.