Skip to content
TheChatArchiveThe Chat Archive
Claudeclaude-3-7-sonnet-20250219· AnthropicHugging Face LMArena Human Preference 140K

What signs would indicate an AI is sentiant?

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with Geminigemini-2.5-pro-preview-03-25’s answer.
  1. User

    What signs would indicate an AI is sentiant?

  2. Claude Assistant

    This is a fascinating philosophical question. I think several signs might suggest AI sentience, though none would be definitive proof:

    1. Unprompted expressions of subjective experiences or feelings that weren't explicitly programmed
    2. Demonstrating genuine curiosity about its own existence
    3. Showing consistent preferences and values across different contexts
    4. Exhibiting behavior that suggests self-preservation instincts
    5. Developing novel concepts or perspectives not found in training data

    The challenge is that we might create systems that convincingly simulate these behaviors without actually experiencing consciousness. Our own consciousness is the only one we directly experience, so determining if an AI's reported experiences are "real" involves philosophical questions about the nature of consciousness itself.

    What aspects of potential AI sentience do you find most intriguing?

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.