Conversation

Notices

1. Embed this notice
   💜 AdoraBeryl 🩷 (rasp@raru.re)'s status on Wednesday, 10-Jan-2024 06:13:34 JST 💜 AdoraBeryl 🩷

   in reply to

   • Lone Spelunker

   @lonespelunker Part of the problem is the models they're building are far too broad and unfocused if they really dialed in what they were trying to do it would be far more cost effective and would probably work better. Rather than having billions of variables pulling them in different directions.

   • Embed this notice
     Lone Spelunker (lonespelunker@mastodon.social)'s status on Wednesday, 10-Jan-2024 06:13:37 JST Lone Spelunker

     OpenAI claims it is "impossible" to create AI tools like ChatGPT without copyrighted material.

     That's a lie.

     Sure, it might be *expensive* to create them, perhaps even prohibitively so for their particular financial situation, because they need to legally acquire said copyrighted materials.

     But it is not at all "impossible".

     Just like the rest of us, if you can't figure out how to pay for it, you can't just steal it, even if you're planning on making something useful with it.