Conversation

Notices

1. Embed this notice
   Profane tmesis (ptmesis@writing.exchange)'s status on Sunday, 22-Dec-2024 23:26:32 JST Profane tmesis

   • Charlie Stross

   @nini @cstross It's a mixture of everything. It's definitely being hyped, but it's also definitely being put to use (good or bad).

   Let's say you have a ton of business records and you want to check if any of them include a specific shady practice. You can't just ask ChatGPT, but you can use it to annotate the records based on a natural language description of the practice.

   It won't catch everything, but it will serve as an extra check.

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Sunday, 22-Dec-2024 23:26:31 JST Rich Felker

     in reply to

     • Charlie Stross

     @ptmesis @nini @cstross Meanwhile you just leaked all those documents and the shady practice you were concerned you might have documented to Microsoft Blackmail.

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Sunday, 22-Dec-2024 23:31:46 JST Rich Felker

     in reply to

     • Charlie Stross

     @ptmesis @nini @cstross It doesn't actually work there because it amplifies the risk they wanted to mitigate.

   • Embed this notice
     Profane tmesis (ptmesis@writing.exchange)'s status on Sunday, 22-Dec-2024 23:31:48 JST Profane tmesis

     in reply to

     • Charlie Stross
     • Rich Felker

     @dalias @nini @cstross Yes, entirely agree.

     I didn't say it was a good idea to do it. I just said that it was a use case where the technology works (well enough). That is why "it doesn't actually work" and "there are no use cases" are such dangerous arguments.

     People with the legit use cases are going to stop listening to the anti-AI crowd, and do stupid shit like upload sensitive information to an irresponsible company that will train their next model on it.