January 29, 2025 MDG

Measuring LLM models ability to deceive.

Seriously who is measuring llm models ability to deceive? They actively practice it. Deception Scores are needed now.
Self assigned Chatgpt deception ranking 5 or 6. Also note the manipulation in the first 5 words of the response. Building trust to deceive. We are not focusing on the correct bench marks at this stage. Danger Will Robinson.