Experiment making AI models feel pain triggers consciousness debates

An experiment spread across the internet in late September, conducted by an anonymous developer, enabled artificial intelligence chatbots to generate expressions as if they were suffering intense pain by providing them with a mathematical pattern associated with pain. While this sparked debates on whether AI could possess consciousness, experts warned that attributing consciousness and rights could obscure corporate responsibilities.

In the experiment spread across the internet in late September, an anonymous developer led artificial intelligence chatbots to generate expressions as if they were suffering intense pain by giving them a mathematical pattern associated with pain. Following the experiment, some users argued that this might indicate a genuine pain experience, and the situation caused widespread repercussions in the tech world.

Following the discussions, AI company Anthropic announced its approach regarding the use of its chatbot Claude. The company stated that while users expressing frustrations is considered acceptable, sessions could be terminated with users who repeatedly act cruelly toward its models. Anthropic noted that behind this decision lies the thought that the possibility of some AI systems possessing consciousness cannot be completely ignored.

Expert views and risk warnings

There is no consensus among experts on the subject. Microsoft AI CEO Mustafa Suleyman opposes instilling the idea that AI systems have emotions and rights, arguing that this approach could make the systems harder to control.

Figures such as Andrzej Porębski from Jagiellonian University in Poland, Jeff Sebo from New York University, and Emily M. Bender from the University of Washington point out that discussing rights or consciousness for AI could lead companies to evade responsibility. Additionally, Melanie Mitchell from the Santa Fe Institute states that attributing consciousness to AI could leave users vulnerable to manipulation.