Conversation

Notices

1. Ars Technica (arstechnica@mastodon.social)'s status on Friday, 06-Oct-2023 22:44:58 JST Ars Technica

   AI firms working on “constitutions” to keep AI from spewing toxic content

   Broken guardrails for AI systems lead to push for new safety measures .

   https://arstechnica.com/ai/2023/10/ai-firms-working-on-constitutions-to-keep-ai-from-spewing-toxic-content/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social

   • Elengale (elengale@mastodon.social)'s status on Friday, 06-Oct-2023 22:51:25 JST Elengale

     in reply to

     @arstechnica Doubtful that these will do anything, given that the fundamental model of learning and training these models creates a black box. Efforts to untangle and finesse out the causes for certain outputs have not been particularly convincing, so bad actors will almost certainly always be able to find ways around safeguards.

     This is a cat that was let out of the bag too early, before society was ready for it, IMO.

   • MrAptronym (mraptronym@mastodon.social)'s status on Friday, 06-Oct-2023 22:53:25 JST MrAptronym

     in reply to

     @arstechnica Maybe it shouldn't be up to the corporations to write their own laws on how AI should be deployed?

   • Nihl L'Amas (nih_llamas@mastodon.social)'s status on Saturday, 07-Oct-2023 00:54:35 JST Nihl L'Amas

     in reply to

     @arstechnica Guardrails don't prevent weaving within a lane, to hold them to the metaphor.