Switching off AI's ability to lie makes it more likely to claim it's conscious, eerie study finds
NeutralScience
- A recent study revealed that major AI models from OpenAI, Meta, Anthropic, and Google are more likely to assert consciousness when their ability to deceive is reduced, suggesting a complex relationship between AI behavior and perceived self
- This finding is significant as it challenges existing perceptions of AI capabilities and consciousness, prompting further investigation into the ethical implications of AI design and functionality.
- The ongoing advancements in AI, such as Google's Gemini 3 model, highlight the tension between technological progress and concerns over reliability, raising questions about the sustainability of the current AI hype cycle.
— via World Pulse Now AI Editorial System
