Totally get what you're saying and agree to an extent, but can you supply examples where algo-driven feeds did not eventually devolve into clickbait/inflammatory/etc recommendations?
I am not saying algos are bad, so I should clarify that. What I should have directed my ire at is the people that implement them since, historically speaking, they can not be trusted to do so without bias or other motives. Chronological feeds cannot be abused in this way, to my knowledge, and represent a state of purity in my mind.
Appreciate you taking the time to make salient points, though.
The alignment problem is an extremely difficult one, and one that we've basically punted on for all AI research. Making an algorithm is easy, scoring content by external factors like engagement or time on screen are also easy. Designing a scoring algorithm that is aligned to a person's preferences, desires, opinions, etc are extremely hard. At best we attempt to find proxies and correlations, the algorithm tracks engagement and we assume that means they want to see more similar content.
IMO this is the fundamental reason for chronological feeds. Aligning the algorithm is an unsolved problem, I'd rather just not use it at all.
I am not saying algos are bad, so I should clarify that. What I should have directed my ire at is the people that implement them since, historically speaking, they can not be trusted to do so without bias or other motives. Chronological feeds cannot be abused in this way, to my knowledge, and represent a state of purity in my mind.
Appreciate you taking the time to make salient points, though.