Conversation

Notices

1. Cayhr (cayhr@varishangout.net)'s status on Tuesday, 25-Jul-2023 05:38:09 JST Cayhr
   I look forward to the irreversible tainting of an AI data model's synapses by giving it false positives, sarcasm, and misinformation.
   
   https://invidious.varishangout.net/watch?v=JXdmqZMX3mA
   • kamehamic (kamehamic@poa.st)'s status on Tuesday, 25-Jul-2023 05:50:49 JST kamehamic

     in reply to
     @Cayhr That's why I think sarcasm or any kind of "forced irony" should be reserved for comedy or in a context where people are VERY aware of it, otherwise it shouldn't be used.
     Like If I see an anon doing it and I can't discern if he is trying to be funny or he means the opposite I can't be bothered to continue the exchange.
   • Cayhr (cayhr@varishangout.net)'s status on Tuesday, 25-Jul-2023 05:50:49 JST Cayhr

     in reply to

     • kamehamic
     @kamehamic It'll be interesting because text is the easiest to generate and parse. However, meaning can get lost in it, as evidenced by humans being unable to detect sarcasm sometimes through text, hence why we had the convention of using '/s' to indicate a sarcastic comment that was posted out of context. Sometimes it's obvious, other times it isn't.
     
     We don't write comments with a voice recording, so these AI are going to struggle pretty hard at scraping authentic newsworthy content. The input is REALLY garbage instead of just a little bit garbage that can be easily filtered.