i’m not convinced search gives you much in unverifiable settings
SAFETY
-
Grok 3.5 Should Include Model Card Documentation
By
–
Grok 3.5 really needs to come with a model card.
-
Fewer accidental amputations through technological advancement
By
–
Fewer accidental amputations I heard
-
AI Taking Shortcuts: The Ethics of Efficiency Over Caution
By
–
AI taking shortcuts, but also, as a cat, why would you leap over them when you can run straight through? pic.twitter.com/W6mqOvqt55
— fofr (@fofrAI) 21 juin 2025AI taking shortcuts, but also, as a cat, why would you leap over them when you can run straight through?
-
Deleting All Errors May Harm Model Error Detection Abilities
By
–
If you delete all errors, the model may lose the ability to detect and correct errors
-
Why AI-Generated Writing Fails to Connect Authentically
By
–
The point of writing is to connect to other minds. Why would you want to leave this connection to a robot wearing a condom designed by a committee
-
User Trust: The Real Driver of AI Success
By
–
There’s one thing that will determine if AI lives up to its potential. Not breakthroughs. Not training data. Not compute. It’s users’ trust. It's a core belief I'm proud Microsoft shares + today's report is just one example of how we put this in action.
-
Obfuscating AI Models While Maintaining Performance
By
–
yeah. people will start finding ways to obfuscate them without harming performance