What is consciousness?

We can divide theories about consciousness into three categories:

  1. Consciousness is a special non-physical property (dualism).
  2. Consciousness is the result of the physical structures of the brain (identity theory).
  3. Conscious mental states are the result of their functional role within a process (functionalism).

In particular, I want to talk about Turing machine functionalism, a specific form of functionalism which states that consciousness is computation on a Turing machine. I want to talk about Turing machine functionalism in particular because it is probably correct.

I know that I am conscious. I strongly believe that other humans are conscious, and I strongly believe (although perhaps less strongly) that many other animals are conscious. They do things that make me believe they’re conscious: they react to stimuli in complex ways, they seek pleasure and avoid pain, and (in the case of humans) they can explain that they are conscious. All of these behaviors could be replicated by a sufficiently advanced computer. If a computer can present all the same evidence of consciousness that an animal can present, and if that evidence persuades me that an animal is conscious, it should also persuade me that a computer is conscious.

(I’m confused as to why this isn’t consensus opinion among philosophers; I’ve read a good bit of philosophy of mind and don’t often hear philosophers touch on this, but it seems to me that it’s clearly the best explanation of consciousness. I don’t think all the academic philosophers who disagree with me are stupid, so I’m probably missing something.)

What type of computation is consciousness?

So Turing machine functionalism is probably true. But that still leaves a few details to work out. There are three main things that consciousness could be:

  1. Consciousness is a Turing machine itself (i.e. a series of instructions).
  2. Consciousness is the output of a Turing machine.
  3. Consciousness is a series of steps taken by a Turing machine.

(1) is implausible. Turing machines are platonic mathematical objects. (1) implies that it does not matter whether a Turing machine is ever instantiated in the real world, which defies intuition.

(2) is implausible as well. For any possible output (represented as a series of zeros and ones), we can write a trivial Turing machine to produce it. For example, to produce “01110”, we can use this machine:

output 0;
move tape;
output 1;
move tape;
output 1;
move tape;
output 1;
move tape;
output 0;
halt;

It is unlikely that such a simple mechanistic program could be conscious.

Therefore (3) must be correct: consciousness is a series of steps taken by a Turing machine.

Is every individual computation conscious, or is consciousness an emergent property of many computations? It seems more likely to be the latter. Each operation in a Turing machine acts on a single bit of information. But I can consciously perceive a lot of information at the same time, which seems to indicate that I am having the subjective experience of running lots of computations at once.

Additionally, a Turing machine may operate over bits, but you could create a different kind of Turing-equivalent computing machine that operates over some other unit. Presumably any Turing-equivalent machine could be conscious, but they can operate in lots of different ways and “single unit of computation” means something different for each of them. In a single step on a Game of Life computer, lots of state changes occur. Consciousness must be a property that can arise in either a Turing machine or in the Game of Life, so the specific state changes can’t be quanta of consciousness–consciousness must be something that all computing machines can have in common.