Deception in AI
Language models struggle with complex deception tasks, often performing at chance level when faced with second-order deceptions. While they may exhibit a basic understanding of deceptive behavior, their limitations become apparent as task complexity increases. The emergence of deception in AI raises concerns, but simply instructing models not to deceive may not be a viable solution.In this clip
From this podcast

Data Skeptic
Emergent Deception in LLMs
Related Questions