Conversation

Notices

1. Embed this notice
   Danilo Campos 🇵🇷 (danilo@hachyderm.io)'s status on Sunday, 21-Jul-2024 00:49:29 JST Danilo Campos 🇵🇷

   in reply to

   Asimov, it seems, was a poor steward of his fandom and abusive of women so I am always cautious not to valorize this prominent first mover

   Still, as a child, I read his robot stuff obsessively, wherein he described the challenges we would face in machines that would sort of think, but not really have common sense. The illusion of reason stamped on a semi-autonomous actor.

   A robot interpreting a careless directive literally, with ensuing hijinx, was a recurring formula Asimov perfected

   • Embed this notice
     Danilo Campos 🇵🇷 (danilo@hachyderm.io)'s status on Sunday, 21-Jul-2024 00:49:30 JST Danilo Campos 🇵🇷

     One thing the LLM is very useful to address is modifying structured data.

     In chatGPT’s case, you can hand it a CSV, then it will write Python to modify it according to your parameters, and you can inspect the Python to decide if it’s appropriate

     To preempt the obsessive pearl clutching, yes, you will have to audit the final output closely, and no the process is not perfect

     But the WAYS in which it is not perfect are close to how the Isaac Asimov robot stories predicted, it’s uncanny