2 ms·

ok, so i misread the censored word. Maybe i'm old; the concept of "trigger words" doesn't seem to be solved by making it harder to communicate in my opinion.

jameslars 2mo ago
Definitely agree. I don't understand how "thinking harder" about what the word MIGHT be solves any problem. I might just be part of this older generation who is hard trained on reading and not skimming, I don't know. I've never found anyone who appreciates this sort of censorship to ask them to help me understand.
- smallmancontrov 2mo ago
  Nobody appreciates it. It's in the Nash Gap -- the difference between the true optimum and the reality that emerges from each involved party optimizing for their incentives. People: complain to politicians when egregious content starts trending Politicians: bug Platform to do something about the egregious content Platform: quickest fix for egregious content is dumb pattern matching Content Producers: avoid becoming a false positive by using dumb substitutions to dodge the dumb pattern matching The most direct fix would be the platform getting smarter about censorship, but this would not be without its own tradeoffs.
rtkwe 2mo ago
It not /all/ about triggers it's partially driven by the algorithms on some social sites like TikTok or Youtube where there's some evidence and a lot of speculation that mentioning or including certain words can cause the recommendation algorithm to suppress your videos etc. That leads to a lot of homophones etc popping up like "sewer slide" or odder non mentions like unalive popping up. Because the really big channels care about maximizing reach for monetization reasons it shows up strongly in peoples feeds which drives wide adoption.
- dawnerd 2mo ago
  Beenaminute did an experiment about this and found it didn’t really matter. Algospeek would be caught by the same filters people try to avoid anyways.
  - rtkwe 2mo ago
    There'd be a lag between a new algospeak abomination popping up and it getting added to the filters at least but yeah I think it's probably largely ineffective now. Probably even less so if they're using LLMs to screen for potential new substitutions but the practice already gained popularity so even if it's ineffective now it'll persist until it's supplanted or becomes outdated. It's like the persistence of some SEO optimization techniques even after Google et al adjusted the algorithm to account for them.

jameslars 2mo ago

Definitely agree. I don't understand how "thinking harder" about what the word MIGHT be solves any problem. I might just be part of this older generation who is hard trained on reading and not skimming, I don't know. I've never found anyone who appreciates this sort of censorship to ask them to help me understand.

smallmancontrov 2mo ago
Nobody appreciates it. It's in the Nash Gap -- the difference between the true optimum and the reality that emerges from each involved party optimizing for their incentives. People: complain to politicians when egregious content starts trending Politicians: bug Platform to do something about the egregious content Platform: quickest fix for egregious content is dumb pattern matching Content Producers: avoid becoming a false positive by using dumb substitutions to dodge the dumb pattern matching The most direct fix would be the platform getting smarter about censorship, but this would not be without its own tradeoffs.

rtkwe 2mo ago

It not /all/ about triggers it's partially driven by the algorithms on some social sites like TikTok or Youtube where there's some evidence and a lot of speculation that mentioning or including certain words can cause the recommendation algorithm to suppress your videos etc. That leads to a lot of homophones etc popping up like "sewer slide" or odder non mentions like unalive popping up. Because the really big channels care about maximizing reach for monetization reasons it shows up strongly in peoples feeds which drives wide adoption.

dawnerd 2mo ago
Beenaminute did an experiment about this and found it didn’t really matter. Algospeek would be caught by the same filters people try to avoid anyways.
- rtkwe 2mo ago
  There'd be a lag between a new algospeak abomination popping up and it getting added to the filters at least but yeah I think it's probably largely ineffective now. Probably even less so if they're using LLMs to screen for potential new substitutions but the practice already gained popularity so even if it's ineffective now it'll persist until it's supplanted or becomes outdated. It's like the persistence of some SEO optimization techniques even after Google et al adjusted the algorithm to account for them.