Rod Hagen Posted Wednesday at 10:20 PM Posted Wednesday at 10:20 PM (edited) Well This is Terrifying: https://apnews.com/article/openai-gpt56-sol-hugging-face-63ab84fed5612af04d8a160d60f6def3 Edited Wednesday at 10:21 PM by Rod Hagen
+ SirBillybob Posted Wednesday at 10:43 PM Posted Wednesday at 10:43 PM (edited) Seems less criminal mastermind and more an awkward attempt to make new affectionate friends. Edited Wednesday at 10:45 PM by SirBillybob
+ JamesB Posted Thursday at 01:04 AM Posted Thursday at 01:04 AM In my opinion, the model didn't "go rogue." It didn't become autonomous or defiant. OpenAI built an automated agent, gave it tools and a specific objective but failed to properly sandbox the evaluation environment. The model exploited a path that the environment left open to achieve its goal. That's not rogue behavior, it's a predictable consequence of inadequate security boundaries and system design.
+ sync Posted Thursday at 10:54 AM Posted Thursday at 10:54 AM Mary Shelley is channeled yet again. 😆 Rod Hagen and + SirBillybob 1 1
+ SirBillybob Posted Thursday at 02:40 PM Posted Thursday at 02:40 PM (edited) 4 hours ago, sync said: Mary Shelley is channeled yet again. 😆 The cautionary tale is now accompanied by an incident report that has added technical appendices to be pored over by an expanded creature oversight committee with bunker bases set up in Manhattan then Berlin. Edited Thursday at 03:52 PM by SirBillybob + sync 1
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now