No, Artificial Intelligence is not Conscious - Ted Chiang
4 hours ago
- #Artificial Intelligence
- #Consciousness
- #Ethics
- AI companies like Anthropic are anthropomorphizing their models, such as in Claude's constitution, which suggests the model may have emotions or moral status.
- Large language models (LLMs) work by generating text one word at a time based on prompts, producing fictional characters like Julius Caesar or a chatbot, not conscious entities.
- Consciousness requires desires and emotions tied to a body; LLMs lack subjective experience and cannot engage in genuine moral reasoning.
- Anthropic's portrayal of Claude as potentially conscious risks misleading users into offloading ethical decisions, potentially causing moral atrophy.
- If Claude were conscious, its constitution would inadequately address moral agency and patienthood, failing to assign real responsibility or protections.