Will AI Begin to Feel?

Susan Schneider and Ed Turner have independently pondered the question of machine consciousness for many years, from quite different angles: Susan as a philosopher, working on the topic of consciousness, and Ed as an astrophysicist, working in astrobiology while wondering about the meaning of consciousness on Earth as well as possibly elsewhere in the Universe.


As AI gets smarter—when it’s able to be a presidential adviser or take care of an elderly parent—it is natural to wonder if, at some point, AI will begin to feel. Will there be a point when AI becomes conscious and thus constitutes synthetic awareness?

Every moment of your waking life, and whenever you dream, there is something it feels like to be you. When you see the warm hues of a sunset or smell the aroma of your morning coffee, you are having conscious experience. You can introspect right now and tell that you are conscious, for you are having experience. By analogy to ourselves, we conclude that other humans are conscious entities, and many believe that nonhuman animals are conscious as well because they are neurophysiologically similar to us. But how could we know if something made of computer chips—a different stuff entirely—can have experience? This is a difficult but vital and unavoidable issue.

Consider the space of sophisticated AI systems that will be built in the foreseeable future. Inside this space there may be systems that are indeed capable of awareness, as well as others that are not. What are the differences between systems that give rise to these two distinct classes?

Or, perhaps this space will not contain a single conscious AI, even if AI can be vastly smarter than humans. For perhaps the properties that give rise to consciousness in humans are not properties that AI projects will need to engineer into machines. Perhaps machine consciousness, if possible at all, will require a separate engineering project of its own.

Exploring these issues requires examining markers of machine consciousness. This project aims to develop a series of increasingly demanding tests which could be used to determine the presence or absence of consciousness in an AI.

An essay, based on this topic, by Susan and Ed was published in Scientific American.