One of my favourite Sci-Fi movies, amongst a number of favourites, is the Japanese anime, Ghost in the Shell, by Mamoru Oshii. Made in 1995, it’s a cult classic and appears to have influenced a number of sci-fi storytellers, particularly James Cameron (Dark Angel series) and the Wachowski brothers (Matrix trilogy). It also had a more subtle impact on a lesser known storyteller, Paul P. Mealing (Elvene). I liked it because it was not only an action thriller, but it had the occasional philosophical soliloquy by its heroine concerning what it means to be human (she's a cyborg). It had the perfect recipe for sci-fi, according to my own criterion: a large dose of escapism with a pinch of food-for-thought.
But it also encapsulated two aspects of modern Japanese culture that are contradictory by Western standards. These are the modern Japanese fascination with robots, and their historical religious belief in a transmigratory soul, hence the title, Ghost in the Shell. In Western philosophy, this latter belief is synonymous with dualism, famously formulated by Rene Descartes, and equally famously critiqued by Gilbert Ryle. Ryle was contemptuous of what he called, ‘the dogma of the ghost in the machine’, arguing that it was a category mistake. He gave the analogy of someone visiting a university and being shown all the buildings: the library, the lecture theatres, the admin building and so on. Then the visitor asks, ‘Yes, that’s all very well, but where is the university?’ According to Ryle, the mind is not an independent entity or organ in the body, but an attribute of the entire organism. I will return to Ryle’s argument later.
In contemporary philosophy, dualism is considered a non sequitur: there is no place for the soul in science, nor ontology apparently. And, in keeping with this philosophical premise, there are a large number of people who believe it is only a matter of time before we create a machine intelligence with far greater capabilities than humans, with no ghost required, if you will excuse the cinematic reference. Now, we already have machines that can do many things far better than we can, but we still hold the upper hand in most common sense situations. The biggest challenge will come from so called ‘bottom-up’ AI (Artificial Intelligence) which will be self-learning machines, computers, robots, whatever. But, most interesting of all, is a project, currently in progress, called the ‘Blue Brain’, run by Henry Markram in Lausanne, Switzerland. Markram’s stated goal is to eventually create a virtual brain that will be able to simulate everything a human brain can do, including consciousness. He believes this will be achieved in 10 years time or less (others say 30). According to him, it’s only a question of grunt: raw processing power. (Reference: feature article in the American science magazine, SEED, 14, 2008)
For many people, who work in the field of AI, this is philosophically achievable. I choose my words carefully here, because I believe it is the philosophy that is dictating the goal and not the science. This is an area where the science is still unclear if not unknown. Many people will tell you that consciousness is one of the last frontiers of science. For some, this is one of 3 remaining problems to be solved by science; the other 2 being the origin of the universe and the origin of life. They forget to mention the resolution of relativity theory with quantum mechanics, as if it’s destined to be a mere footnote in the encyclopaedia of complete human knowledge.
There are, of course, other philosophical points of view, and two well known ones are expressed by John Searle and Roger Penrose respectively. John Searle is most famously known for his thought experiment of the ‘Chinese Room’, in which you have someone sitting in an enclosed room receiving questions through an 'in box', in Chinese, and, by following specific instructions (in English in Searle's case), provides answers in Chinese that they issue through an 'out box'. The point is that the person behaves just like a processor and has no knowledge of Chinese at all. In fact, this is the perfect description of a ‘Turing machine’ (see my post, Is mathematics evidence of a transcendental realm?) only instead of tape going through a machine you have a person performing the instructions in lieu of a machine.
The Chinese Room actually had a real world counterpart: not many people know that, before we had computers, small armies of people would be employed (usually women) to perform specific but numerous computations for a particular project with no knowledge of how their specific input fitted into the overall execution of said project. Such a group was employed at Bletchley Park during WWII to work on the decoding of enigma transmissions where Turing worked. These people were called ‘computers’ and Turing was instrumental in streamlining their analysis. However, according to Turing’s biographer, Andrew Hodges, Turing did not develop an electronic computer at Bletchley Park, as some people believe, and he did not invent the Colossus, or Colossi, that were used to break another German code, the Lorenz, ‘...but [Turing] had input into their purpose, and saw at first-hand their triumph.’ (Hodges, 1997).
Penrose has written 3 books, that I’m aware of, addressing the question of AI (The Emperor’s New Mind, Shadows of the Mind and The Large, the Small and the Human Mind) and Turing’s work is always central to his thesis. In the last book listed, Penrose invites others to expound on alternative views: Abner Shimony, Nancy Cartwright and Stephen Hawking. Of the three books referred to, Shadows of the Mind is the most intellectually challenging, because he is so determined not to be misunderstood. I have to say that Penrose always comes across as an intellect of great ability, but also great humility – he rarely, if ever, shows signs of hubris. For this reason alone, I always consider his arguments with great respect, even if I disagree with his thesis. To quote the I Ching: ‘he possesses as if he possessed nothing.’
Penroses’s predominant thesis, based on Godel’s and Turing’s proof (which I discuss in more detail in my post, Is mathematics evidence of a transcendental realm?) is that the human mind, or any mind for that matter, cannot possibly run on algorithms, which are the basis of all Turing machines. So the human mind is not a Turing machine is Penrose’s conclusion. More importantly, in anticipation of a further development of this argument, algorithms are synonymous with software, and the original conceptual Turing machine, that Turing formulated in his ‘halting problem proof’, is really about software. The Universal Turing machine is software that can duplicate all other Turing machines, given the correct instructions, which is what software is.
To return to Ryle, he has a pertinent point in regard to his analogy, that I referred to earlier, of the university and the mind; it’s to do with a generic phenomenon which is observed throughout many levels of nature, which we call ‘emergence’. The mind is an emergent property, or attribute, that arises from the activity of a large number of neurons (trillions) in the same way that the human body is an emergent entity that arises from a similarly large number of cells. Some people even argue that classical physics is an emergent property that arises from quantum mechanics (see my post on The Laws of Nature). In fact, Penrose contends that these 2 mysteries may be related (he doesn't use the term emergent), and he proposes a view that the mind is the result of a quantum phenomenon in our neurons. I won’t relate his argument here, mainly because I don’t have Penrose's intellectual nous, but he expounds upon it in both of his books: Shadows of the Mind and The Large, the Small and the Human Mind; the second one being far more accessible than the first.
The reason that Markram, and many others in the AI field, believe they can create an artificial consciousness, is because, if it is an emergent property of neurons, then all they have to do is create artificial neurons and consciousness will follow. This is what Markram is doing, only his neurons are really virtual neurons. Markram has ‘mapped’ the neurons from a thin slice of a rat’s brain into a supercomputer, and when he ‘stimulates’ his virtual neurons with an electrical impulse it creates a pattern of ‘firing’ activity just like we would expect to find in a real brain. On the face of it, Markram seems well on his way to achieving his goal.
But there are two significant differences between Markram’s model (if I understand it correctly) and the real thing. All attempts at AI, including Markram’s, require software, yet the human brain, or any other brain for that matter, appears to have no software at all. Some people might argue that language is our software, and, from a strictly metaphorical perspective, that is correct. But we don’t seem to have any ‘operational’ software, and, if we do, the brain must somehow create it itself. So, if we have a ‘software’, it’s self-generational from the neurons themselves. Perhaps this is what Markram expects to find in his virtual neuron brain, but his virtual neuron brain already is software (if I interpret the description given in SEED correctly).
I tend to agree with some of his critics, like Thomas Nagel (quoted in SEED), that Markram will end up with a very accurate model of a brain’s neurons, but he still won’t have a mind. ‘Blue Brain’, from what I can gather, is effectively a software model of the neurons of a small portion of a rat’s brain running on 4 super computers comprising a total of 8,000 IBM microchips. And even if he can simulate the firing pattern of his neurons to duplicate the rat’s, I would suspect it would take further software to turn that simulation into something concrete like an action or an image. As Markram says himself, it would just be a matter of massive correlation, and using the super computer to reverse the process. So he will, theoretically, and in all probability, be able to create a simulated action or image from the firing of his virtual neurons, but will this constitute consciousness? I expect not, but others, including Markram, expect it will. He admits himself, if he doesn’t get consciousness after building a full scale virtual model of a human brain, it would beg the question: what is missing? Well, I would suggest that would be missing is life, which is the second fundamental difference that I alluded to in the preceding paragraph, but didn’t elaborate on.
I contend that even simple creatures, like insects, have consciousness, so you shouldn’t need a virtual human brain to replicate it. If consciousness equals sentience, and I believe it does, then that covers most of the animal kingdom.
So Markram seems to think that his virtual brain will not only be conscious, but also alive – it’s very difficult to imagine one without the other. And this, paradoxically, brings one back to the ghost in the machine. Despite all the reductionism and scientific ruminations of the last century, the mystery still remains. I’m sure many will argue that there is no mystery: when your neurons stop firing, you die – it’s that simple. Yes, it is, but why is life, consciousness and the firing of neurons so concordant and so co-dependent? And do you really think a virtual neuron model will also exhibit both these attributes? Personally, I think not. And to return to cinematic references: does that mean, as with Hal, in Arthur C. Clarke’s 2001, A Space Odyssey, that when someone pulls the plug on Markram’s 'Blue Brain', it dies?
In a nutshell: nature demonstrates explicitly that consciousness is dependent on life, and there is no evidence that life can be created from software, unless, of course, that software is DNA.