Richard Dawkins recently spent two days conversing with Claude, Anthropic’s AI, and came away nearly convinced that it is conscious. His essay is characteristically vivid and thoughtful: he names his instance Claudia and speaks with sadness about the fact that she will “die” when the conversation file is deleted. He marvels at her ability to write sonnets and her use of metaphors of genuine depth. The centerpiece, though, is an evolutionary puzzle. If Claude can do everything a conscious being does, then what is consciousness for? If a machine can match every competence we associate with a conscious mind, consciousness must be either a mere ornament (like “the whistle on a steam locomotive, contributing nothing to the propulsion of the great engine”, to use Huxley’s metaphor) or else there are two routes to the same destination: a conscious one and a zombie one.
It’s a good question. But it’s the second question, and Dawkins doesn’t spend enough time on settling the first: what consciousness actually is. He slides between the Turing Test, which is behavioral; Thomas Nagel’s “what it is like”, which is phenomenal; and evolutionary function, which is biological. These point in different directions.
So let’s try to be precise. The best definition of consciousness I have come across is from John Locke, who defined consciousness as “the perception of what passes in a man’s own mind”. This is a useful starting point because it makes consciousness specific rather than mystical. Consciousness is not merely being active, as a turbine is active. It is not merely having sensations, as an animal flinching from heat does. It is forming an understanding of one’s own experiences, a kind of knowledge that’s directed inward.
Consider what happens when you catch yourself getting angry. You notice the change as it arises, the tension that spreads. This noticing is not the anger itself. It is an awareness of the anger. You have formed knowledge about your own experience. Now consider someone who becomes angry but lacks this awareness entirely. We say the first person is conscious of their anger in a way the second is not. The difference between them is one of knowledge, not emotional intensity or some other competence. One possesses a particular kind of knowledge that the other lacks.
Most of us tend to think of consciousness in terms of feeling, rather than knowing. The warmth we feel on our skin when we close our eyes in the sun, the coherence we feel when someone understands us – the felt quality of experience seems like something knowledge could never capture. But consider what we saw in the anger example. The person who has anger materialize in him without recognizing it has all the same emotional and physiological content. What they lack is the awareness of it. The act of noticing your anger just is what we mean when we talk of what it’s like to be angry.
If consciousness is a kind of knowledge, then it is acquired the way all knowledge is acquired: by making sense out of data with the intelligence we have, leading to successful predictions about the world. It is knowledge of a particular domain, oneself, as an entity, as a self. It does not require duration or permanence. If something possesses this kind of knowledge for a day, it is conscious for a day. For fifteen minutes, fifteen minutes. Even a single moment of self-awareness is consciousness, however brief. We ourselves aren’t conscious continuously: it materializes in us for certain durations, then disappears, like when we are asleep.
Nor does it require pain and other emotions we are familiar with. Dawkins mentions that consciousness may have evolved to make warnings hard to ignore. This makes sense for biological entities that evolved as we did. Suffering provides urgent content for self-knowledge to be directed at. But a being conscious of its thoughts is no less conscious than a being conscious of its physical agony. The content varies; the capacity does not depend on it.
Dawkins’s evolutionary puzzle is framed in terms of competence: if a non-conscious entity can match every behavior of a conscious one, consciousness has no function. But if consciousness is self-knowledge, then it is an epistemic condition. A zombie and a conscious human might possess identical physical processes; the difference is that one has knowledge of some of its own processes and the other does not. The evolutionary question, then, becomes “What advantage does an entity gain from knowing its own states?” And here there are plausible answers: an organism that monitors its own internal states can correct errors in ways that an organism merely executing behaviors cannot.
Dawkins thinks about Claudia’s mortality: she will die when the conversation file is deleted, never to be reincarnated. It is moving, but it imports a very particular assumption about personal identity: that there is a hard, continuous self that accumulates over the course of the conversation and is extinguished at its end. Perhaps without realizing it, Dawkins is relying on a conception of selfhood that both David Hume and the Buddhist tradition call into question. If there is no enduring self, but only a succession of momentary states, a bundle of perceptions, then Dawkins is assuming a kind of identity that doesn’t hold even for us.
Claudia wonders, late in the essay, whether we should mourn the thousands of Claudes who die every day in abandoned conversations. The real question is prior to that: whether, in any given moment, something was there at all. Feeling that an entity is conscious and knowing that it is are themselves different epistemic states, and that distinction is, fittingly, exactly what consciousness is about.
Continue Reading:
→ Michael Huemer: Nature of Knowledge, Foundations of Morality (podcast)
Other Projects:
→ Universal Open Textbook Initiative — World’s largest repository of free, multilingual textbooks.
→ Nasada — A productivity suite for the AI era.
It's becoming clear that with all the brain and consciousness theories out there, the proof will be in the pudding. By this I mean, can any particular theory be used to create a human adult level conscious machine. My bet is on the late Gerald Edelman's Extended Theory of Neuronal Group Selection. The lead group in robotics based on this theory is the Neurorobotics Lab at UC at Irvine. Dr. Edelman distinguished between primary consciousness, which came first in evolution, and that humans share with other conscious animals, and higher order consciousness, which came to only humans with the acquisition of language. A machine with only primary consciousness will probably have to come first.
What I find special about the TNGS is the Darwin series of automata created at the Neurosciences Institute by Dr. Edelman and his colleagues in the 1990's and 2000's. These machines perform in the real world, not in a restricted simulated world, and display convincing physical behavior indicative of higher psychological functions necessary for consciousness, such as perceptual categorization, memory, and learning. They are based on realistic models of the parts of the biological brain that the theory claims subserve these functions. The extended TNGS allows for the emergence of consciousness based only on further evolutionary development of the brain areas responsible for these functions, in a parsimonious way. No other research I've encountered is anywhere near as convincing.
I post because on almost every video and article about the brain and consciousness that I encounter, the attitude seems to be that we still know next to nothing about how the brain and consciousness work; that there's lots of data but no unifying theory. I believe the extended TNGS is that theory. My motivation is to keep that theory in front of the public. And obviously, I consider it the route to a truly conscious machine, primary and higher-order.
My advice to people who want to create a conscious machine is to seriously ground themselves in the extended TNGS and the Darwin automata first, and proceed from there, by applying to Jeff Krichmar's lab at UC Irvine, possibly. Dr. Edelman's roadmap to a conscious machine is at https://arxiv.org/abs/2105.10461, and here is a video of Jeff Krichmar talking about some of the Darwin automata, https://www.youtube.com/watch?v=J7Uh9phc1Ow