Decoding consciousness: What computers can and cannot tell us about our own minds

An engineering professor believes consciousness is not computable, as in it works fundamentally different from our current mathematics, but is that really the case?  Computers, in fact, have a lot to teach us about our own minds, just not in the way most people think.  Read on for a few musings on what makes us tick and how we can unravel the mystery of ourselves…

According to technology companies and the media, Artificial Intelligence is everywhere, that is computers are now capable of making human-like decisions including the ability to learn.  There’s no doubt that modern computers are incredible devices.  They perform tasks that were inconceivable just a few years ago, recognizing people in photos, translating text to speech, recognizing speech itself, and more.  This capacity has prompted comparisons to human intelligence and human consciousness.  These comparisons usually take a similar form:  Computers are performing human-like tasks, but machines still show no signs of consciousness, therefore there is something magical about consciousness.

For example, Subhash Kak, a professor of electrical and computer engineering at Oklahoma State University, claims computers can never be conscious in a recent article for The Conversation.  “Some researchers continue to insist that simulating neuroscience with computers is the way to go. Others, like me, view these efforts as doomed to failure because we do not believe consciousness is computable. Our basic argument is that brains integrate and compress multiple components of an experience, including sight and smell – which simply can’t be handled in the way today’s computers sense, process and store data.”

While Mr. Kak is certainly correct that human and animal brains process and store data very differently than computers, that tells us next to nothing about whether or not consciousness can be simulated on some future machine or whether it’s really true that consciousness isn’t computable under any circumstances.  Personally, I think he is mistaken in that regard.  For starters, calling computers intelligent doesn’t really make them so, as even the most advanced Artificial Intelligence can’t reproduce the complexity of behavior present in an insect colony, much less a human being.  There are limits to comparisons between machine and man, some of which we will touch on later in this post.

In the meantime, there is no doubt that many of the tasks our minds perform are computable.  We have mapped out some of our mental processes reasonably well, including how our senses work, how data flows from an eye or an ear into the brain for example, even how our brains process reflex reactions like flinching.  No one doubts these processes are computable, nor does anyone believe the individual neurons that make up our nervous systems are capable of non-computable feats.  They are cells like any other.  It is only when we consider higher order brain functions that claims of non-computable or the idea that consciousness could not have naturally evolved come in, but if these claims are to be correct, we have to believe that readily computable parts add up to an non-computable whole.  The mechanism to achieve that defies explanation, leading to the assumption that there remains some kind of ghost in the machine.

In my opinion, Mr. Kak and others reach these and similar incorrect conclusions by embracing some sort of two track belief system inherited from the French philosopher Rene Descartes, where consciousness is a necessarily privileged state that doesn’t fully conform to traditional physics and biology including evolution.  As Mr. Kak puts it, “Consciousness appears to be a privileged state of the mind.”

He sees evidence for this in a mathematical scenario known as the “halting problem.”  Simply put, some mathematical operations generate infinite loops, that is once you start them, you never end them.  A computer could calculate the next step forever and no conclusion would ever be reached.  This is normally because some step in the program sends the computer back to the beginning, for example, simply X equals X plus one, repeat.  Obviously, it’s easy to tell that such a basic loop will never end, but more complicated loops can be much more difficult to determine in advance.  In fact, there is no mathematical test that can be performed to differentiate between programs that reach an endpoint and those that do not.  The only way to know for sure is to run the program and see what happens.

Mr. Kak believes that consciousness can square this circle and halt our internal processes at any time.  “Let us define ‘consciousness’ as some privileged state of the mind that makes its processes halt (we don’t bother to specify it beyond this description) and its contents registered (which is what we imply by awareness). Humans can get into the state of ‘awareness’ at any time, which means that the earlier computation has halted, and this is irrespective of the initial state of the immediately preceding process. The exceptions to this are if a person is sleeping or unconscious as in coma. But such halting to arbitrary input is impossible from a computability point of view. Therefore, it follows that consciousness is not computable.”

In my opinion, this is an unsatisfying interpretation for several reasons.  First, human consciousness does not actually violate the halting problem in the sense of solving it when computers cannot:  Shown a computer program we have no special means to discern whether or not it halts in violation of the mathematical principle.  Instead, we have to follow the steps that a computer would to identify a loop, meaning our brains have no access to privileged knowledge machines do not.  While this isn’t precisely Mr. Kak assertion, it’s important because if we could magically solve problems computers simply can’t, something non-computable would have to be occurring and he would certainly be correct.  There remains no evidence of that, either in the halting problem or any other mathematical construct, only that we do things differently and in many cases better, at least for now.

Second, the definition of consciousness Mr. Kak uses is self-referential and relies on unproven assumptions.  He starts by declaring it privileged and claiming it does _____ without knowing what the _____ really is, but assuming it violates the halting problem regardless, primarily because of appearances alone.  There is no biological reason to conclude that we achieve a state of awareness because we halt another process in our brains.  Indeed, experience alone tells us that consciousness and the processes that fill our mind all operate continuously, without halting, more on that in a moment.

Ultimately, Mr. Kak is starting with consciousness as a “privileged” state of mind and then declaring that, because of its privilege, it must have special properties.  The real underlying question is whether or not consciousness is truly privileged or is that simply an unwarranted assumption.  Of course, it certainly seems privileged given that our experience of the world is unique and, though it shares certain similarities with other animals, there appears to be nothing else like human consciousness in the known universe.

Appearances can be deceiving, however, leading to the first and most common mistake:  Consciousness is generally regarded as some process that runs outside of the underlying physical reality, meaning our senses process incoming information, prepare it in some manner, and then our conscious selves watch the show like a we would a movie screen.  There is a line between the inside and outside world.  As Mr. Kak describes it, “We think of ourselves as being outside of the physical world. Even our conceptions of the universe are as if we are not a part of it, and in the words of Schrödinger: ‘We do not belong to this material world that science constructs for us. We are not in it; we are outside. We are only spectators. The reason why we believe that we are in it, that we belong to the picture, is that our bodies are in the picture. Our bodies belong to it.’ If this sense of being outside of the physical world is true, it would be impossible to emulate it by hardware and processing that is within the world.”

This, I think, is very deceiving:  When it comes to consciousness, there is no outside world at all.  I don’t mean that in any deep philosophical sense as in how do we know the word really exists, but rather in a purely practical sense:  The only information we have about the outside world is generated on the inside.  From the perspective of the self, we know nothing of the outside world, only our internal representation of it, which can certainly be wrong if you are mentally ill or hallucinating.  The same way it is difficult to know you are dreaming while dreaming, or science fiction authors imagine we are actually in a simulation, we only experience the outside world from inside our brains and nervous systems.  Our brains themselves set up an inside and outside dichotomy for (largely) obvious reasons, but it is only an illusion.

Therefore consciousness isn’t “acting” on anything.  We do not look at a table and then consciousness, like a gnome in the back of our mind, sees the table and labels it as a table. Instead, input comes in from our senses and is assembled into a unified representation that includes our thoughts, emotions, and physical state.  This representation might sometimes seem like we are watching our lives on a movie screen and the “extras” of consciousness are separate, added later, but when you probe deeper our thoughts and feelings and everything else we consider conscious are just as embedded in the output of our senses.

For example, consider what happens when you see a table.  You instantly recognize it as such, even if you can’t recall the word.  Your emotional state is just as relevant as your sensory perceptions.  If you are expecting guests and excited, you might imagine the table set for dinner.  If you are fearing a conversation, your thoughts will turn darker.  The recognition of what you are seeing and your ability to reflect on it is not a privileged state; it’s the only state of experience and it is fundamentally inseparable.  

If you don’t believe me, try looking at that table and then separating everything else that comes along with it.  You just want to see the table.  You don’t want to know it as a table.  You don’t want to think of the last time you sat at the table.  You just want to observe the object without thought or feeling.  Can you?  Now, try to imagine what the world would look like if you didn’t have language at all and things didn’t have names, or you were incapable of considering abstract versions of things, imagining scenarios around the table or anything else.

You simply cannot; it’s impossible to separate out elements of our unified experience.  This doesn’t mean we cannot imagine what it must be like, considering what it would be like to separate rational thought from emotional impulse, but no matter what we do we cannot suppress the totality of our conscious experience.   It’s there, whether we want it to be or not, with the sole exception of mind-altering states like deep meditation or drugs.  Otherwise, the word and object are one and the same in our minds, existing separately only as abstractions in our inner monologue; likewise, if you think of the word “table” without seeing a table, your mind will conjure the image of one.  Our inner monologue isn’t like Facebook adding tags to an image, it’s inseparable from the image.  This is because consciousness is everything we know about the world, wrapped up in our experience of it.  As I said earlier, it is the only state.

I didn’t pick the Facebook comparison out of thin air, either.  This leads to a valuable distinction between information processing and information synthesis, two topics that are usually blended together or used alternatively when they are not synonymous.  Generally speaking, Artificial Intelligence is good at information processing.  Powerful computers and sophisticated algorithms can process incredibly large amounts of data in complex ways and extract certain information from it, like Facebook tagging a face in a library of images.  Our brains do this as well:  All of our senses operate without consciousness and reflex actions can react to things before conscious awareness dawns.  Furthermore, insects and other lower animals can see and process the data generated, yet no one thinks a fly is conscious.

In my opinion, this leads to two conclusions:  First, everyone seems to agree that unconscious information processing is computable.  It’s taken for granted that everything short of our inner monologue, the feeling of what happens, and imagination (perhaps a few others to be fair) are readily explicable, but because we have a few traits that lack satisfactory explanations we need to resort to some non-computable, non-evolutionary process.  It’s interesting to note that our inner monologue is likely generated from unconscious language processing systems in our brains, meaning we don’t pick and choose each individual word most of the time, rather the stream is generated just like sight and sound.  Second, we shouldn’t be surprised that computer systems that mimic basic unconscious processes are themselves not conscious, meaning, at this point, our most intelligent computers are performing tasks that are unconscious in the animal world. Why would we think computers should require consciousness to perform them and why would we draw any conclusion based on it?

On the other hand, consciousness is all about information synthesis, taking the disparate data produced by our senses and other unconscious processes, including language, and assembling it into a coherent, manipulatable “image.”  First, coherence:  Our experience of the world is seamless, both in the present and over time.  Though our eyes are composed of rods and cones, our ears take in sound via the discrete beats of a bone, our nose analyzes specific chemicals based on different sensor molecules, and our sense of touch is based on individual nerve endings throughout the body, our experience of the world isn’t segregated or bifurcated in any way.  It is whole, entire, complete, enveloping us.

For example, we do not separately perceive the shape of an object and its motion; we perceive objects in motion, and we can easily pick them out and label them whether or not they are moving or still.  The disparate senses themselves are even blurred and intertwined, and all of it is coherent with our inner monologue and emotional state, both inseparable from it.  This is true over time as well:  Our bodies change, but our sense of self remains stable, even though we aren’t precisely the people we were or are going to be.  Second, manipulatable:  We can do more than simply extract information from the coherent image assembled by consciousness, we can rearrange and reimagine it.  We can consider what that table from earlier would like in a different color or with a different set of chairs.  We can compare it to other tables we’ve seen.  We can imagine sitting with family and friends around it for dinner.  We can dream about the table and not even know it, entering a state where it exists only in our own minds yet still possesses solid form.

This facility should be considered distinct from merely processing the information to identify a table:  We can literally build a whole new world around it in our minds, using the exact same “space” we would just when watching the table.  There is no Artificial Intelligence software on the planet or even in development that can do anything remotely like this, not even close.  The best they can do is interpret an image and tag it with a table, similar to how some of our own unconscious processes can recognize objects outside our primary field of vision before we become aware of them.  Consciousness, however, somehow extracts the “essence” of a table and all that comes with it as part of the simple experience of seeing a table.  Moreover, it does this seamlessly, all of the time, with everything we see and do.  It never stops while we are awake.

When you consider that no other organism in the known universe can do this, the real question becomes:  How do we know that consciousness is not the only means with which to accomplish this level of coherence and manipulation?  We don’t, nor do we need to directly answer that question.  Instead, we can ask what would computing this seamless nature of consciousness look like?  Or more specifically, what processes could conceivably bind disparate, unconscious data streams into a coherent, manipulatable whole?

Alas, I am not a mathematician or computer scientist, but I can try to provide a few pointers as to what such a thing might look like.

First, the seamless coherence is an illusion.  The mind itself is neither seamless, nor coherent, but for some reason that is likely of incredible importance, consciousness renders it that way.  The question should be: Is there any other way to do it or is consciousness the only way to bind all this information together? Ultimately, as I argued recently:  Brains are expensive, if we could simply be unconscious zombies, we would be.  Therefore, it seems obvious to me that this coherent nature of consciousness is critical to understanding both how it functions and why it evolved in the first place.

Second, the illusion of coherence from incoherent data leads to another suggestion:  Forget how computers work, where everything is broken down into rigid binary code.  We know the mind doesn’t work that way, but there are mathematical structures that are far more elegant than a granular stream of ones and zeroes.  Einstein’s General Relativity for example uses a structure called a tensor that simultaneously encodes both the gravitational force and its position in space.  These tensors do not occur on a fixed Cartesian plane like familiar Euclidean geometry.  The plane itself is part of the tensor.  They are whole, in and of themselves, and their output can be infinitely smooth.  While I am not suggesting consciousness uses tensors, it strikes me that there are similarities to the structure: Experience is encoded coherently, binding the senses, thoughts, and emotions into a single expression that cannot simply be subdivided.  Waveforms are another mathematical and physical entity that cannot simply be cut in half or pulled apart.  I am sure there are others that possess this kind of unity.

Third, consciousness is ever-present, always-on.  You can’t escape it or turn it off, except while sleeping, drugged, or seriously injured.  In my opinion, this is another potential clue.  Why would this be so?  Why does stopping consciousness require effectively shutting down your brain or at least significant parts of it?  Surely, we can imagine an equivalent human race that can turn their minds on and off, experiencing only when they want to.  We, however, have to live it every day, every moment.  When your brain does shut down for sleep, it’s almost like it de-coheres, not so much stopping, but being pulled apart, meaning if you remove one part of the edifice, the whole thing won’t run.  If you really think about it, it’s almost the opposite of Mr. Kak’s halting problem.  The problem isn’t that consciousness violates the halting problem; it’s that consciousness never halts.  I suspect there must be a reason for this as well and that the computability of consciousness will rely on a structure that must be whole or doesn’t truly work at all.

Fourth, when I said humans are the only organism in the known universe that can do this, I was intentionally overselling it.  Human consciousness with our inner monologue and ability for abstraction appears to be unique, but the coherent nature of consciousness is likely shared across at least all mammals if not other complex organisms like birds.  It also seems likely that they cannot turn off their minds either.  They’re just as stuck in the world as we are.  For our purposes, shared aspects of consciousness with different animals would mean that consciousness is not a unitary phenomenon, much like the coherence is an illusion, so is the concept that consciousness is a single thing.  In my opinion, one of the keys to understanding consciousness and making it computable is to identify it’s independent parts, things like experience itself, memory recall, decision making, inner monologue, etc., and map them out separately prior to assembling the whole.

Think of it this way:  Language is a critical part of our modern consciousness, but full fledged language likely didn’t arise until somewhere around 100,000 years ago.  There were proto-languages before then, but the infinitely fluid languages of today that can impart abstract thoughts are an incredibly recent development in evolutionary terms.  This is critically important because modern humans evolved long before this fully fledged language, likely at least a hundred thousand years.  This implies two options:  Either these early humans that shared our exact DNA and cognitive ability weren’t conscious, or they were conscious, but in some different form without a fully realized inner monologue.  I find it impossible to believe they were unconscious zombies.  Therefore, consciousness must be able to exist without language and is more evidence that consciousness itself has constituent parts.

Ultimately, if everything feeding into these constituent parts is computable, there is no reason to believe that consciousness itself isn’t and we can finally get rid of the ghost in the machine, a dualism that has haunted us since Descartes.  Comparisons to computers will only get us so far, at least for the foreseeable future.  In the meantime, we would do well to consider how many other mysteries of the human body we have solved that seemed impenetrable at the time.  Consciousness, in my opinion, will ultimately prove no different.