Rafi Letzler has written a fascinating article titled, Is a simulated brain conscious? in which he poses several interesting thought experiments. For instance, would a group of people standing together in a giant field, with each person obeying the rules that one of the 86 billion neurons in a typical human brain obeys, possess a collective group consciousness? What about a supercomputer that can simulate the entire brain of a human being? Or what about a Boltzmann brain, formed by atoms suddenly coalescing together at random in an exact replica of the atoms in a human brain? Dr. Scott Aaronson, a theoretical computer scientist at MIT, offers his own thought-provoking answers to these questions, but in this post, I’d like to make a few observations of my own.
1. In a previous post, I argued that even if Moore’s law (that the number of transistors an integrated circuit can hold doubles every two years) continues to hold, there’s no way that the Internet could match the complexity of a human brain before 2115. In any case, Moore’s law will be dead by 2022, according to a leading expert in the field, Robert Colwell, who is director of the microsystems group at the Defense Advanced Research Projects Agency (DARPA). (See also this article.) I also cited an article by Dr. James Rose (who has researched the neural requirements of consciousness at considerable length) arguing that anything with a radically different structure from our neocortex would have a different function: whatever it would be capable of, it certainly wouldn’t be consciousness.
2. The Boltzmann brain problem is much-overhyped, in my opinion: there are excellent chemical reasons why it would simply be impossible for a bunch of atoms to suddenly coalesce into an arrangement which exactly duplicates the atoms in somebody’s brain. Neurons don’t just magically sit in place; they need to be held together by glial cells. In a nutshell: a Boltzmann brain would fall apart long before it had completed assembling itself.
3. The experiment involving people standing in a big group, with each one obeying the rules that a neuron obeys, is vulnerable to one major criticism: neurons don’t always behave deterministically. If non-deterministic behavior of neurons plays a vital role in consciousness, then attempts to recreate consciousness in this fashion are doomed to failure. The thought experiment also assumes that causality is always bottom-up and never top-down. Again, if irreducible top-down causality is a real fact of Nature then there are serious grounds for doubting the possibility of artificial consciousness.
4. In my post, Zombies, duplicates, human beasts and consciousness, I argued that conscious reflection is an activity that cannot, even in principle, be explained in materialistic terms. If my argument is correct, then the likelihood of anyone constructing a self-conscious entity out of physical parts is precisely zero. But that doesn’t rule out the theoretical possibility of someone creating an artificial structure with a capacity for animal consciousness but without a capacity for self-reflection.
And that’s all for today. Any thoughts from readers?