What Is Artificial Intelligence and Consciousness?

Few use consistent coherent vocabulary to talk about AI. In particular, we tend to be careless or ignorant about important distinctions between intelligence and consciousness, and about the inexhaustibly vast possibility space for kinds and degrees of both.

There’s not necessarily one right definition of “intelligence” or “consciousness.” But some effort at consistent careful use of these words would help more of us understand each other better when discussing our hopes and fears, and perspectives on practicalities, related to AI.

When I think of intelligence, and when I use the word, I have in mind the capacity to achieve goals. This concept is tightly coupled with concepts of anatomy and structure, or the information of matter. Bodies are more and less intelligent in infinite variety.

When I think of consciousness, in contrast, I have in mind the capacity for experience. This concept encompasses the concepts of sensing and feeling. Because we can only ever observe our own, consciousness is at once most familiar and most mysterious.

As I use them, “intelligence” and “consciousness” are orthogonal concepts. We can imagine super intelligence that’s unconscious. And we can imagine deep consciousness that’s unintelligent. A subset of intelligence seems to correlate with focused or self-aware consciousness.

But we can also imagine self-unaware and unfocused consciousness. We might experience this to some extent as “brain fog.” Beyond that, in anesthesia and death, we might think that consciousness is altogether annihilated. Or we might imagine de-individuation into panpsychism.

Conscious or not, AI is clearly intelligent, as understood within the framework I’ve described here. It has capacity to achieve goals. In some ways, AI is far more intelligent than humans. In other ways, humans are more intelligent, at least for now. This isn’t controversial.

Nor is it an inherent compliment to attribute mere intelligence to AI. Per the framework I’ve described, even bodies as simple as flush toilets are intelligent. Their anatomy works to maintain a level of water within a tank. They have capacity to achieve that goal.

More complimentary it would be to attribute general intelligence to AI. When we consider this, we usually imagine that human intelligence is general intelligence. We have capacity to achieve many goals, including creation of flush toilets. And we would compare AI to ourselves.

But the possibility space of intelligence — the possibility space of anatomy and structure and information — is vastly larger than even the most diverse expressions of human biology. So it’s not at all clear that humans are the best measure of general intelligence.

And yet it remains practical to compare AI with humans, even if for no other reason, because we are humans. And many of us are increasingly wondering whether and to what extent AI already has or soon will surpass our intelligence in ways that will make our lives better or worse.