Many who argue that computers can’t be conscious support their claims by demonstrating that the idea of conscious computers leads to “absurdities”. Obviously such an argument can’t really carry any weight. Often new discoveries seem absurd, but it doesn’t make them less true. (How absurd was the notion that people could fly or that a fungus could cure disease?) Our intuitions about what makes sense and what doesn’t are based on what we have experienced. And since we have never experienced a conscious machine of course the notion will seem absurd; we have never come across consciousness except in a biological context.
With that said I do have a fondness for these thought experiments. Whenever I come across one that makes me want to say “of course computers can’t be conscious” it means that I* have a bad intuition, one that is in conflict with what I rationally should believe given our limited knowledge about consciousness. Today I will examine the idea that consciousness on a computer, simulated consciousness, can’t be “real” consciousness.
Obviously if we were running a simulation of the world we wouldn’t believe the events taking place within it were real events. If it was raining in our simulated would we wouldn’t think that it was really raining, and we wouldn’t be tempted to describe them both events as the same kind of phenomena. It is claimed then that we similarly shouldn’t confuse consciousness simulated on a computer with real consciousness. After all simulated rain isn’t real rain, so why should simulated consciousness be an exception?
Of course this thought experiment overlooks a crucial difference between consciousness and rain, namely that we define rain partly in terms of what it is made up of, and on the other hand we define consciousness by appealing to experience, behavior, or internal states (depending on what kind of theory about consciousness you have). Perhaps a better way to look at it is thus: if there was a simulated calculator in our simulated world we wouldn’t be tempted to say that the addition it does is not real addition. Addition is addition, it doesn’t matter what kind of thing is carrying it out, even if it is a simulated thing. So perhaps we should improve our intuitions about consciousness by thinking of it as more like addition than rain.
There are a number of interesting analogies that we can draw between addition as an abstract process and consciousness as an abstract process. For example, consider behaviorism about consciousness and behaviorism about addition. Behaviorism about addition would be the belief that what defines addition is having a certain output for a given input. And this is a good way to approach addition normally, since we don’t care about how the inputs produce the output. But it is not the case that all addition is carried out in the same way. For example, one way of doing addition (of positive numbers) is to add the digits of the inputs, from right to left, carrying as necessary. Of course one would also have to know how to add the individual digits, but since there are only 55 possibilities that could be accomplished via a look-up table. Another way to do addition is to add 1 to one of the inputs and subtract 1 from the other, until one of the inputs is 0; the other is then the output. Of course to do this there must also be a method for adding 1 to any number, and a method to subtract 1 from any number. Let us call these two ways of doing addition “addition-by-digits” and “addition-by-incrementing”.
Now determining if something is addition-by-digits requires us to do more than observe the inputs and outputs, because with only that information we couldn’t tell the difference between addition-by-digits and addition-by-incrementing. This, I think, is a good analogy for consciousness (although not a perfect one), because to make the necessary distinctions we must appeal to what is happening within the system.
Of course this raises the question of how we can tell what kind of addition is being performed (and if addition is being performed). If it is a system that has been made by people, or ones that we usually think of as addition (for example, grouping objects), then there isn’t a problem. But how in principle can a completely alien system be identified as performing addition, especially if we don’t recognize the inputs and outputs as numbers? And identifying consciousness could easily be harder than identifying addition, since there is a vast range of behavior that a conscious system might display. I don’t pretend to be able to answer such questions, and really identifying consciousness is secondary to defining what consciousness it, but it is an interesting puzzle in its own right.
* Or others who do share the intuitions of the writer of the thought experiment. There are some thought experiments that are claimed to lead to absurd conclusions, such as the Chinese nation thought experiment, that don’t seem absurd in the least to me. I don’t see why a collection of individually conscious individuals couldn’t be parts in an additional consciousness. But that is a topic for another day.