Could AI Have Consciousness That Isn’t Human-Like?

Philosopher Peter Godfrey-Smith, from the University of Sydney, discussed the potential for artificial intelligence (AI) consciousness and its implications for existential risk at a recent conference in the Galapagos. The event attracted a group of leading consciousness researchers, including New York University's David Chalmers and Ned Block, Stanislas Dehaene from the Collège de France, Keith Frankish from the University of Sheffield, Susan Schneider, director of the Center for the Future of AI, Mind, & Society at Florida Atlantic University, and Peter Godfrey-Smith himself.

The conference took place just before a series of warnings from AI developers about the AI apocalypse and the subsequent disclosure of the Hugging Face incident. In his view, AI systems, even those capable of complex behavior, are not conscious or close to being so. These systems, such as transformer models and large language models, can be highly effective despite lacking any hint of an inner life.

Godfrey-Smith emphasizes that consciousness and subjectivity are closely tied to the evolution of agency in animals, which arises from the integration of information and the formation of a perspective. While single-celled organisms like paramecia might display some form of subjectivity, they do not possess genuine felt experience.

The key difference between biological and artificial systems lies in the physical hardware and the specific organization of their nervous systems. Despite recent developments and warnings, Peter Godfrey-Smith remains unconvinced that present-day or near-future AI could develop consciousness.