Can a Computer Be Conscious?

As with so many other posts, I’m indebted to David Yerle for getting me to think about this (thanks to an article on consciousness he pointed me to). 

Can a computer be conscious?  There are some problems with this question.

First, we don’t understand consciousness.  We know it correlates with activity in a part of the brain, that we lose it when sleeping, regain it when waking, and can even lose it while awake (going on autopilot), but don’t know why it would ever arise in the first place.  To even ask how the material could give rise to the mental is a confusing one, and some go so far as to claim the very question itself is incoherent.

What’s more, the question may be unanswerable until we learn what causes consciousness.  To answer whether something can exhibit some property, it helps to know the causes of this property.  So the question about a computer being conscious can be rephrased as:

We don’t know what causes consciousness, but want to know if a particular thing can  satisfy these unknown causes.

That seems unanswerable.  Regardless, speculating on this question can teach us some things.

First, that we’d wonder if a computer can be conscious reveals an assumption that consciousness does not require a (fleshy) brain, that it’s an activity that happens to occur there, but could easily occur elsewhere.

Second, why would we even think to ask if a computer can have consciousness?  I rarely see similar questions about rocks or trees (and certainly, not asked with the same passion), so why computers?

Perhaps the answer lies in why I think others are conscious.  It’s true I can only experience my consciousness, yet I believe others are conscious because they look and act similarly to me, so I assume they are inwardly similar as well (including having an inwardness to begin with).

Very well, but what do I have in common with a computer?  It has no legs, no face, it doesn’t talk,  and it’s cold metal instead of warm flesh.  Shouldn’t I more easily identify with a dog or a mannequin?  Yet, I identify with a computer to the point of allowing it the potential for consciousness.  Why?

Could it be because of attachment to my mind?  A computer seems to think.  It does things that I can do mentally (and better than me), like calculate, search and play chess. Since the mind is often valued (in the West) as who we “really” are, it could be that I ignore all the ways a computer differs and focus on the thing it may share with me, a thing which I happen to identify as my essence:  A mind.

Yet, what is this mind?  My brain is structured differently than a computer — it uses neural networks, while the computer uses logic gates.  The brain relies on more pattern matching, weighted inputs and associative maps, while the computer is more reliant on binary operations (and can contain multiple computational models).  Yet, I ascribe a commonality that transcends this.  Have I postulated a new entity?  Is the concept of mind/thought/computation a chimera?  Is this just a label for an interesting series of interactions, and I’m now making the mistake of giving it an ontological status?

Now since consciousness is near and dear to me, it’s easy to conflate those two concepts and assume that consciousness must arise as the result of thought (which itself may not exist in the disembodied way I imagine).

But there are plenty of reasons to doubt this. First, there are people with diminished mental functioning (to the point that they can’t think as many do) who are conscious.  Second, many complex processes — like driving, doing dishes and multi-tasking — are done unconsciously.  Think of a time you arrived at home, with absolutely no memory of how you got there . A great deal of our functioning is unconscious. Third, we program computers to do many things that seem to require thought, yet do not program them with consciousness (as if we could!).   A robot could be programmed to walk, talk, run, react, even scream in pain when damaged in the same way it can be programmed to add a list of numbers.  Indeed, a Philosophical Zombie is not the least bit contradictory.

So….

  1. Could a computer be conscious?
  2. What causes consciousness?
  3. What is thought?
  4. What is computation?
Advertisements

20 thoughts on “Can a Computer Be Conscious?

  1. It’s a nice point about rocks and trees. A tree seems a far more likely candidate for consciousness than a piece of software. For me it would seem more likely to have a mind.

    There is a fifth question I’d like to ask. How could we ever know that a computer is conscious, even if it is?

    Btw. Para 7 has ‘and’ rather than ‘an’, which I’ll mention because it was confusing

  2. Thanks for the mention! I pretty much said all I had to say by e-mail. I’ve always tended to think that computers can be conscious because of this:
    Computers are universal Turing machines.
    Universal Turing machines can simulate any other machine (i. e. our universe or us). Even if the machine is analogue, it can be translated to any arbitrary precision into a Turing-machine digital form.
    Therefore, computers are capable of simulating us. Since we are conscious, computers are capable of consciousness. Bear in mind, though, that we’re speaking about computers with an arbitrary amount of processing power. So which computers will be conscious or how to program them is a completely different ball game.
    That said, maybe “computers can be conscious” is bad phrasing. Maybe what we really mean is “certain software can be conscious” or, put more abstractly, “consciousness can be expressed mathematically” which is, now that I think of it, quite a strong statement. But heck, if the laws of physics can be expressed mathematically, anything that emerges from them should too.

    1. I think you could be more optimistic DY. This is like saying that computers can be conscious therefore computers can be conscious. It assumes you are no more than a pocket calculator and then derives the fat that you are. What is to stop us from making a different assumption? is there any evidence that it is a correct assumption?

      1. I don’t see what optimism has to do with anything. Computers being conscious or not have nothing to do with desired/undesired outcomes.
        The sentence “you are no more than a pocket calculator” carries a value judgement and it is extremely misleading. It somehow implies it is a bad thing to be a pocket calculator.
        Moreover, a Turing machine is a universal, abstract machine which is able to perform any computation. It has unlimited memory and processing power. It therefore has very little to do with a pocket calculator.
        That said, we are not speaking about humans and pocket calculators, but asking ourselves this question: can the laws of physics be expressed mathematically?
        There is a theorem that says that any mathematical function can be evaluated by a Turing machine. Computers are Turing machines. Therefore, powerful enough computers can evaluate any mathematical function. That is not open to debate.
        What is open to debate is whether the universe follows mathematical laws. However, there is a huge weight of evidence to support this view: the fact that the laws of physics work to a mind-boggling degree of precision. On the other hand, there is next to no evidence to support the opposite view.
        If the universe follows mathematical laws, then by definition the universe can be simulated by a powerful enough computer. If the universe can be simulated by a powerful enough computer, then anything in it will also be simulated, including conscious beings.
        This has nothing to do with making value judgements about people and calculators, nor imagining a brave new world. I am not making predictions about the future nor predicting a timeline for when computers take over the world. I am making a statement about the mathematical properties of the universe, a statements that turns out to be supported by an absurd amount of evidence.

      2. Sorry if that was a bit inflammatory, but what is a computer except a calculator? And even quite complex ones can be pocket sized. Making it bigger or putting arms and legs on it isn’t going to help. It’s the idea that consciousness is merely computation that I find impossible to take seriously. We would have to assume that there is no ‘hard’ problem, and clearly there is. There is no argument for machine consciousness other than our inability to demonstrate that it is impossible.

        Or, not unless we say that consciousness is computation, and then use another name for awareness, feeling, desire and so forth, the properties that distinguish us from computers. This is not a new idea and it seems quite a good one. .

    2. Yes, the article “Can a Computer Be Conscious?” might have better been phrased with regards to software, but that’s its usual phrasing, so I thought I’d stick with it. But maybe it’s not badly phrased at all? When we talk of simulation and organization, what are the things that are simulated or organized? Is it abstract concepts, or patterns of electrons? That is, would software give rise to computer consciousness by invoking the right patterns of electrons?

      Put another way, imagine a computational device built out of a different medium (like bowling balls or dominoes). Assuming it’s large enough (yes, this is theoretical) would it be conscious? If so, it seems that we’re either talking about a pure relation where the physical parts are irrelevant…

      1. Exactly! That is my point: that the physical make-up is completely irrelevant and that what determines consciousness is organization. Any computer, made of anything (silicon, blue cheese, bananas, you name it) will be conscious if it has the right relationships. Heck, you could make a computer with people waving their hands.

      2. So here’s the other question: what is the nature of this organization?

        For instance, is it a “physical” pattern formed by the components? Is it that the components “compute”? But if it’s the latter, what qualifies as computation — especially since how we interpret something can turn a nonsensical device into a computational one, and seems to define computation (and hence consciousness) in anthropomorphic terms. Something computes when it produces something we recognize as a result, but we recognize these results based on our agendas, so a physical principle seems to have been transmuted to a human-centric one…

      3. Here I’m going to go out on a limb and give Wolfram’s answer: everything is computation. Everything that involves some rule affecting some elements is a computation. There is no non-computation going on. In this sense, humans are just able to recognize some of these computations. The key of course is that just a very limited subset of those computations give rise to consciousness. Which subset that is, I have no idea.
        What I mean is that it really doesn’t matter how we call it or how we recognize it. Computation is just some “rule” being implemented. A mechanical (though “mechanical” is probably not the right word because of its connotations) process. Of course one can look at one same process and see different computations going on in it, depending on how we define what the elements are. But that does not matter: anything that is isomorphic to some computation of a Turing machine will qualify. Both what we see and what we don’t.
        Then, you could of course distinguish between micro-scale (“fundamental”) computation and macro-scale (“emergent”) computation. Maybe only micro-scale computation really takes place, but since macro-scale is isomorphic to some possible micro-scale universe, it looks as if there is some other, different computation going on that has nothing to do with where it comes from (the laws of physics). That would be consciousness.

      4. Thanks.

        I have a problem with the “everything is a computation” definition because it means nothing. Why not just drop “computation” because of its connotations? At the end of the day, aren’t we just talking about interaction? By this definition, an avalanche, pot of water and human brain all compute.

        It seems that instead of clarifying or postulating anything new, we just eliminated a term by depriving it of all meaning. With this view of computation, is anything new being said about the universe or laws of physics?

        In the end, aren’t we just saying that some patterns of interaction give rise to consciousness? Why postulate any order or intelligence which is implied by computation?

        Put another way, how is computation different from the run-of-the mill interactions we have always been talking about?

      5. I wouldn’t say it means nothing. There is a very clear definition of what a computation is: any operation that can be carried out by a Turing machine. That is, any process where I can associate certain states to states of a Turing machine and then model its evolution by certain operations performed on those states is a computation. The statement “everything is a computation” is not a definition but an observation: for any state in the universe A, I can associate a state A’ in a Turing machine so that the evolution of A can be mirrored by A’. This is not generalising the notion of computation: it is making a statement about the universe. That is, the universe is computable. This didn’t have to be the case.
        The question here is not whether something is a computation, but which kind of computation it is. For example, most cellular automata are completely uninteresting (see Wolfram’s book “A New Kind of Science”) whereas some of them give unpredictable results with emergent behavior. “Everything is a computation” gives us a framework to study the universe, not in terms of differential equations but in terms of algorithms. As such, it is extremely meaningful. In fact, the fruitfulness of this approach was shown by Wolfram, who wrote a whole book based on this premise and managed to make meaningful predictions concerning biology and physics, to give a few examples.

      6. One more thing regarding micro vs. macro computation. The idea is that some systems can be “summarized” by taking bigger elements that behave according to some macroscopic rules emergent from the microscopic ones. These systems behave in a special way: in most cases, these summaries will be impossible and the only possible description will be in terms of the micro-laws. That is what makes organized systems (machines, living things) special: they can be understood in terms of bigger parts that interact in their own unique way. This means we can give a computational description of a living thing (or a human being) in terms that make no reference to their microscopic make-up. Hence, background independence: any background that gives rise to the same set of computations that define a human will contain humans in it, regardless of how similar those background are.

    3. But isn’t this notion of computation simply rephrasing the belief that the universe follows orderly laws? The ability to simulate the universe follows from its orderly nature.

      I’m familiar with “A New Kind of Science” and Cellular Automata. Cellular Automata is good stuff — especially the 2D variety. They’re really good for demonstrating emergent order and as such might be a good candidate for gaining insight into consciousness.

      I’ve thought about whether CA would be a good model for physics. Perhaps macro physics, as the locality requirement seems like a problem for micro (unless one postulates a different type of locality, but that’s opening up a can of worms).

      1. Yes, in a way. But not exactly. For example, in “The Emperor’s New Mind” (I think, though it could have been the other book) Penrose shows examples of noncomputable laws of nature which are nonetheless “orderly.” That is: the set of all possible orderly laws of nature is smaller than the set of all computable laws of nature.
        I also think that, if that were not the case, the computational approach is still fruitful. For example, Euclidean geometry and Riemannian geometry can be shown to be equivalent, but Riemannian geometry is much better suited to describe curved spaces. In the same sense, having laws based on differential equations may be less well suited to reality than the computation paradigm, even if they are both equivalent descriptions.
        That is: any programming language can be shown to be equivalent to any other. That doesn’t mean you always want to write in assembly language. A different set of tools usually allows you to tackle problems you were not able to solve before.

  3. Thanks for another interesting post.
    Seems to me there are two basic responses to your question: either you’re a physicalist, and believe that there’s nothing to consciousness that can’t be instantiated in physical terms. This would make the notion of a zombie logically contradictory and therefore impossible in practice. (As Dennett, for example would hold). Or you believe that consciusness is real, but can’t entirely be explained by physics. In that case, since we have no idea as yet how consciousness is generated, the only possible answer to “Can a Computer be Conscious?” is “Don’t know”. That’s my answer, anyway.

    1. Thanks for the good points. I’m inclined to give the zombie view more weight because there have been many times in my life (arguably, the vast majority) in which I’ve been a philosophical zombie. In a way, one could rephrase the zombie argument as a 24/7 extrapolation of this scenario.

      Still, you make an interesting point. If consciousness is not necessary, but arises as a byproduct, then one can logically show its lack of necessity, but that would not show its “actual lack of necessity”.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s