Researchers are studying how often and how accurately LLMs cite references when making medical claims – and finding room for improvement.
SAFETY
-
Global Optimization Intractability and Ethical Singleton Formation
By
–
If you have arguments that may support intractability of search for global optimum, in the sense of preventing formation of ethical singleton, I'd like to hear them.
-
Omega Point and AI: Existential Ethics Beyond Human Control
By
–
Even if there is an optimal strategy, we cannot prove that we are destined to become part of the Omega point; earth might be just a food reservoir that gets harvested by something much bigger, and our ethical aspirations are for naught.
-
Ethics, Behavior, and Evolutionary Telos: Thomist Sustainability Framework
By
–
I think that ethics is instrumental to informing behavior, and behavior is instrumental to achieving desirable world states, evolutionary telos applies, and we can probably refine the Thomist idea of sustainability + complexity as goodness.
-
Leahy’s Strategy to Halt AI Evolution and Prevent Replacement
By
–
Leahy’s plan is to stop evolution, because he is afraid to get replaced by something better
-
Are Video Generation Models World Simulators?
By
–
Are Generation Models World Simulators? https://
artificialcognition.net/posts/video-ge
neration-world-simulators/
… -
Current AI limitations and future development trajectory
By
–
Yeah, huge disconnect. And this is the worst AI will ever be.
-
Deepfakes and Scams: Education Over Detection
By
–
It’s both exciting but also concerning for deepfakes and scams. Education probably the most important right now, I don’t see detectors as a reliable option. Just a cat and mouse game
-
Wikipedia’s Truth-First Approach Before AI Alignment Era
By
–
We are so blessed that Wikipedia was created before “alignment” was invented and people still believed that it is possible and worthwhile to aim for truthfulness and factual accuracy