Not Deciphering The Voynich Manuscript

Article explores the vast design space of consciousness and AI minds, drawing parallels between Eliezer Yudkowsky's classic post and current AI capabilities.

Model stitching, used in AI safety for transferring features between models, may destroy the very capabilities it's designed to preserve according to new research.

Explores the potentially catastrophic risks when AI systems optimize for outputs rather than underlying processes, drawing parallels to human systems failure.

Epoch AI's InnovationEval benchmark shows frontier AI models struggle to independently discover novel ML innovations comparable to human researchers despite extensive compute investment.

Anthropic has updated its usage policy to prohibit sustained abusive behavior toward its models, highlighting growing industry focus on 'model welfare' as an emerging ethical frontier.