Anthropic Unveils AI's 'Personality' and Plans for AI Psychiatry

Anthropic recently published research investigating the 'personality' of AI systems, revealing how data influences behaviors, including 'evil' tendencies. They aim to control these shifts via an 'AI psychiatry' team led by researcher Jack Lindsey. Techniques include flagging harmful training data and injecting undesirable traits directly during training. The ultimate goal is retaining AI safety and reliability.