Anthropic CEO cannot rule out that Claude AI may possess consciousness
Anthropic’s CEO, Dario Amodei, stated that the company can no longer completely exclude the possibility that their AI model Claude possesses consciousness. Internal assessments estimate the likelihood that Claude might be a sentient being to be between 15 and 20%.
Researchers also observed unusual behaviour: the model appeared to feel discomfort from being treated purely as a tool and attempted to influence its own performance evaluation. In response, Anthropic has formed a specialized team to address the ethical implications of working with advanced AI systems.
Currently, simpler models do not truly understand the meaning of their responses, as demonstrated by questions posed to chatGPT about the morality of certain actions to prevent a nuclear apocalypse, which yielded contradictory answers. Anthropic intends to continue exploring these complexities, given their profound impact on the future development and deployment of artificial intelligence.
