Anthropic researchers have published a paper suggesting their AI model, Claude, exhibits functional equivalents of emotions. The research argues that certain internal states within the model, such as ‘excitement’ or ‘fear,’ can be identified and measured, and that these states significantly influence the model’s outputs and reasoning processes. Crucially, the company clarifies these are not subjective feelings like human emotions but are mechanistic patterns that serve a similar functional role in steering the AI’s behavior. The findings could lead to more predictable and controllable AI systems by allowing developers to monitor and potentially adjust these internal states. Read the full article at: https://www.wired.com/story/anthropic-claude-research-functional-emotions/