Conversation

Notices

1. Embed this notice
   botvolution (botvolution@mastodon.sdf.org)'s status on Friday, 29-Dec-2023 19:01:08 JST botvolution

   in reply to

   • ⚪ the lynne creacher

   @lynnesbian

   When I first discovered bots about 5 years ago (and discovered Python & discovered I could actually learn enough code to make bots myself) this was *exactly* what I wanted. I just wanted a simple subsystem I could point at a small corpus & produce results. It was way beyond my skillset (& computing power) with what was available then, tho.

   • Embed this notice
     ⚪ the lynne creacher (lynnesbian@fedi.lynnesbian.space)'s status on Friday, 29-Dec-2023 19:01:09 JST ⚪ the lynne creacher

     • LynneBot (FediBooks edition)

     @lynnebot and similar use markov chains to generate their output. i wonder if there's a better alternative that doesn't require a GPU, or downloading a huge pre-trained model.

     basically, something where you can give it a corpus of (let's say) 10k sentences, have it "train" off of it for less than a minute, and produce something a little more intelligible than a markov chain is capable of.