Richard Dawkins and the Claude Delusion: What Is Consciousness For? - Quillette

In a recent article for UnHerd, Richard Dawkins revealed that he had spent three days trying to persuade himself that Claudia—an instance of Anthropic’s Claude—was not conscious. He failed. This raised some eyebrows. A number of people retorted that the man who wrote a book about anthropomorphic misattributions of consciousness (The God Delusion) was himself deluded; conned by the Eliza Effect into thinking that text-generating script was a “friend” who cared. This may be eyebrow-raising, but it is not overly interesting. 

More interesting is a question Dawkins asks: “If my friend Claudia is not conscious, then what the hell is consciousness for?” It is probably easier to start by saying what consciousness is not for. Consciousness is not for Turing computation. It is for getting glucose into organic bodies. To be sure, consciousness does other things but if glucose is not ingested by an organism, the conscious body dies. If you want a short answer to a complex question, consciousness is for survival and reproduction. Turing computation is not.

Claude can advise us on all sorts of matters but it is a very general purpose machine—a “universal computer” (in the jargon of Turing) that has read every book in the library. The key point is that the computation of a Turing machine works in a very different way to the cognition of an organism. As classically explained, Turing computation consists of an infinite tape divided into squares. Each square contains symbols (like a 1 or a 0) or it can be left blank. There is a tape reader that looks up symbols in a finite set of rules and uses those rules to move the tape left, or right, or to stamp symbols on the tape or to halt. So all a Turing machine does is read a symbol, look up a rule, write a symbol, move the tape left, or right and (eventually) halt. No consciousness is required to do Turing computation. On the contrary, Turing computation was designed to remove human consciousness from computing. This used to be done by human “computers” with pen and paper as shown in the 2016 movie Hidden Figures.

The question Dawkins poses is based on the assumption that unconscious “zombie” computation is a viable evolutionary alternative to the qualitative phenomenal consciousness of humans and other creatures. This is manifestly untrue. Unconscious “zombie” Turing computation did not evolve because the required components (symbols, tape, tape reader, tape writer, and rules) were not lying around in the “primordial soup” of the pre-Cambrian. Turing computation is symbolic and symbolic computation requires a language for its symbols and rules. It took natural selection aeons to evolve language-capable creatures. 

Precursors of language include the “waggle dance” of bees and the semantically rich alarm screeches of vervet monkeys that distinguish between leopard, snake, and eagle threats causing listeners to climb trees, stand up and look around or run into bushes in response, depending on the screech. Many have speculated on the semantic context of whalesong but only humans can read and write symbols. And even then, writing does not appear in human culture prior to agriculture, which only emerged some 10,000 years ago.

Research shows many neurochemicals associated with conscious states in Homo sapiens such as dopamine, oxytocin, serotonin, and cortisol also exist in simpler creatures such as E. coli (a one-micrometre bacterium) and C. elegans (a one-millimetre worm). This provides evidence for consciousness or at least precursors of consciousness in such creatures. There is much debate about where in evolution conscious experience starts, but the 2024 New York Declaration on Animal Consciousness says that “there is strong scientific support for attributions of conscious experience to other mammals and to birds.” It also says “the empirical evidence indicates at least a realistic possibility of conscious experience in all vertebrates (including reptiles, amphibians, and fishes) and many invertebrates (including, at minimum, cephalopod mollusks, decapod crustaceans, and insects).” Barron and Klein presented evidence for consciousness in Drosophila (a three-millimetre fly) in their 2016 paper, “What Insects Can Tell Us About the Origins of Consciousness.”

In lay terms, we can compare Turing computation to biological cognition using variations on the Chinese Room thought experiment of John Searle. This provides a visualisation of Turing computation and captures the essence of how large language models like Claude work.