Transumanar

Singularity Fallacies: An essay by Extropia DaSilva (part 1)

Another thoughtful essay by Extropia DaSilva, on the Technological Singularity and the evolution of the Internet towards the Omninet.

“In this extended essay, I look at some of the fallacies that crop up in discussions of the Technological Singularity, and I don’t just mean the fallacies made by the people who think it is all a load of nonsense. In fact, before proceeding, it’s worth noting that, just because the arguments put forward by most critics contain inaccuracies and a general lack of understanding, that does not mean to say that their conclusion (that there will be no Singularity) is wrong. Indeed, at the end of this essay I argue that a proper understanding of what the Singularity represents does show that its physical existence is an illusion… “

Also, the ‘Omninet’: “Sci-fi visions of becoming immersed in cyberspace imagined this would ocurr via us ‘jacking in’ by plugging a cable into our brains. Cyberspace might indeed enter our brains, albeit via a network of nanoscale transponders communicating with neurons and each other on a local area wireless network. But, ultimately, if this idea of an omninet is valid, immersion will happen because the Internet spreads out into ubiquitous sensors that pervade the environment. The sheer quantity of data and diversity of knowledge that will exist in this age would overwhelm us, absolutely requiring advanced machine intelligence to help organize and make sense of it”.

SINGULARITY FALLACIES: AN ESSAY BY EXTROPIA DaSILVA.

“A beginning is the time for taking the most delicate care that the balances are correct. This every sister of the Bene Gesserit knows”- Frank Herbert.

In this extended essay, I look at some of the fallacies that crop up in discussions of the Technological Singularity, and I don’t just mean the fallacies made by the people who think it is all a load of nonsense. In fact, before proceeding, it’s worth noting that, just because the arguments put forward by most critics contain inaccuracies and a general lack of understanding, that does not mean to say that their conclusion (that there will be no Singularity) is wrong. Indeed, at the end of this essay I argue that a proper understanding of what the Singularity represents does show that its physical existence is an illusion.

But first, the fallacies. Before you can say whether a theory is right, you have to be right about what a theory says. A particularly common error is to improperly define the Singularity. According to the Singularity Institute, ‘the arrival of greater-than-human intelligence is the original and only correct definition of the Technological Singularity. (‘Greater-than-human’ does not equate solely to AI, but to any technological amplification of natural human intelligence). Other definitions, focusing on the acceleration of technological change, the greater global co-operation of human beings, and so on, are contortions of the original definition’.

A lot of critics fall at the first hurdle, by arguing that the Singularity is defined as the accelerating growth of innovation. Having established this (mis)representation, they then argue that greater innovation occurred in the past compared to that which we see today. This is contrary to what (their incorrectly defined) Technological Singularity predicts, so it must be wrong. To be fair, there is a tiny grain of truth with regards to the Singularity being about increasing innovation, in so much as an increase in smartness is likely to result in the ability to see things in a new light, and reveal hitherto unrealised potential in the relevant technologies. However, by and large the Singularity has nothing to do with innovation as most people understand it. Indeed, a propper appreciation of its enabling factors shows that the aparrant decrease of innovation, supposedly falsifying the Singularity hypothesis, is in fact an expected outcome of intelligence amplification.

Fundamentally, the Technological Singularity is a consequence of the growth in the power and performance of information technologies. Our ability to inherit, refine, add to and pass on a growing body of knowledge is the underlying reason why machines, tools etc tend to improve over time. For practical or economic reasons, machines, tools etc will eventually adopt a form that does not appreciably change thereafter. For instance, the basic shape of airplanes is unlikely to undergo any significant alterations, so we can expect aircraft of the future to resemble aircraft of today. But we cannot reason (as some critics do) ‘we don’t see very much innovation in aircraft design so the Singularity must be wrong’, because the Technological Singularity does NOT require (or entail) a constant and eternal stream of new inventions in every area of technology. ALL it requires is a certain level of improvement in the ability to store and process information. Any one technology will inevitably run up against limits, but a more generalized capability like computation, storage or bandwidth tends to follow a pure exponential, bridging across a variety of technologies. One consequence of enhancing our ability to store and process information is that we are likely to arrive at optimal designs for any one technology sooner now (and even sooner in the future) than we would have in the past. So, the ‘S’ curve (a line drawn on a graph depicting a rising curve of improving capability, followed by a levelling-off as economic or physical limits are reached) should be expected to form more quickly as smartness increases. This is one reason why an apparent decrease in innovation is to expected as we progress toward Singularity.

But the main reason why innovation decreases is that most people equate the word with something truly novel. Introducing a new capability, inventing a machine that hitherto did not exist, these achievements represent innovation as most people understand it. But improvements to a pre-existing capability, or refinements to an established invention, seem less innovative since they do not introduce anything particularly new. Kurzweil identified seven distinct stages in the life cycle of a technology. First, there is the ‘precursor’ stage, in which the prerequisites of a technology exist and dreamers explore conceptual designs, written explanations etc, of an invention they think could be realised if elements are brought together correctly. However, as Kurzweil pointed out, ‘we do not regard dreaming to be the same as inventing…Da Vinci drew convincing pictures of airplanes and automobiles but he isn’t considered to have invented either’. The next stage in the cycle is ‘invention’, in which a new technology is brought to life (IE someone or some group actually builds it, rather than merely explores conceptual designs of it). This is followed by ‘development’. It’s extraordinarily unlikely that a brand new invention will represent the optimal solution to any given problem or need. To pick on aviation again, the Wright Brothers demonstrated the feasibility of sustained, powered flight but their craft was hardly the best that could theoretically be built. ‘Development’ leads to ‘maturity’ in which the invention becomes an established part of the community. This sets the stage for ‘pretenders’, or alternate and better ways to achieve an aim already demonstrated by an existing invention. Typically, these alternate methods fail to overthrow the established way of doing things, until stage six is reached via the introduction of something that retains the benefits of the old technology, while also introducing new capabilities. Think DVD versus videotape. The old technology is thereafter moved to stage seven: Antiquity.

The thing you should notice about this life-cycle, is that what it mostly produces is failure. Most conceptual designs are destined never to become actual inventions (perhaps because the design is wildly impractical). Most inventions fail to corner a lucrative market. Most ‘pretenders’ fail to overthrow mature inventions. This tendency towards failure is grist to the mill for Singularity naysayers, who are often to be found reciting the failed predictions of past futurists, thereby implying that contemporary forecasts will prove wrong as well. “They said we would have hotels on the moon and flying cars by the year 2000”. Yes, well, it is difficult to imagine the actual benefit of holidaying in a place so inhospitable it makes the North Pole in winter seem like a luxury stay in the Costa Del Sol. As for flying cars, they sound great in theory but in practice vertical take-off and landing require immense skill, generating sufficient thrust produces deafening noise and consumes enormous quantities of fuel, and the co-ordination of three-dimensional traffic lanes in Rush Hour would be a logistical nightmare. How such examples bare any relation to something like improvements in our ability to store and process information, the knock-off effect of which brings great benefit to so many areas, is hard to fathom.

The other thing to notice is that the life-cycle of technology rarely results in innovation as most people understand it. We see it at stage two, when a brand new technology is invented. We also see it to a much lesser extent at stage six, when a rival technology succeeds in improving a functionality enough to make it a worthwhile investment. Call these ‘visible’ innovations. But the other stages are mostly concerned with improving a pre-existing capability. I would argue that such an endeavour seems less innovative to most people, since all it entails is ‘more of the same’. Call this ‘invisible innovation‘.

Before going on, I must emphasise that I am NOT saying that innovation objectively diminishes as we progress towards a future of greater information storage and processing. Rather, I am saying that certain factors conspire to make this the subjective viewpoint of most people. The key to understanding why the rate of innovation appears to slow down lies in appreciating that there are ‘visible’ and ‘invisible’ innovations. Visible innovations are those that introduce a change that is without precedent whereas invisible innovations are those that incrementally improve an established invention.

An excellent example of a visible innovation occurred in July 1969. In an event that was truly without precedent, a human being set foot on the surface of another world. Now that the goal of travelling to another world has been achieved, all other demonstrations of this ability must necessarily have less of an impact on our psyche because they will just be variations on a proven concept. That does not mean to say no progress will have been made regarding improvements in shielding or rocket booster technology or whatever. No doubt, a great many innovations will have been accomplished in fields like material science to ensure the vehicle that will take us to Mars will be comprised of, and equipped with, a great many components that did not exist during the Apollo mission. But they are all designed to perform the same basic job: Achieve escape velocity and journey to another orbiting body. The public has seen it all before in countless documentaries. It will have nowhere near the impact of those grainy images of Armstrong stepping onto the lunar landscape.

It may well be the case that, in order to witness an increasingly large impact of visible innovations, our gaze must be directed into the past as opposed to the future. Marvin Minsky noted this trend: ‘In spite of all we hear about modern technological revolutions, they really haven’t made such large differences in our lives over the past half century. Did television really change the world? Surely less than radio did and even less than the telephone did. What about airplanes? They merely reduced travel time from days to hours- whereas the railroad and automobile had already made a much larger change by shortening those travel times from weeks to days’. Now that we have an established travel industry, can incremental improvements to this established innovation have anywhere near the same impact? Of course, much remains to be done in terms of safety and pollution but even a robot car with a hydrogen fuel cell engine is only going to drive us from A to B, much like we have been used to for generations.

As the number of revolutionary changes builds up, the impact of revolutionary change diminishes. This may sound like a paradox but it is easily resolved. The more technology a civilization has accumulated, the less likely it will be to encounter a change that is truly remarkable, that really opens up a whole new opportunity. The Internet was made possible thanks to a staggering amount of technology in the form of computers, satellites and telecommunication. Moreover, each of these incorporate yet more technology in the form of modems, integrated circuits, solar panels, hard drives…An overwhelming amount of technology in fact, but what is it FOR? What purpose does all that tech SERVE? It is transmitting, storing and processing information in the form of data. In ‘Future Hype’, Bob Seidensticker argued, ‘much has happened to improve communication in the past half century. But is this even close to the amount of change brought about by the introduction of the telegraph? Imagine when information travelled only as fast as a horse or ship could carry it and consider the enormous change brought about when the telegraph made communication not just faster, but instantaneous. Information, no longer cargo, could be transmitted as data rather than carried’.

The data processed on the Web manifests itself in ways that could not be achieved via the telegraph. We can listen to music, watch films, perform calculations, run simulations, play games. But this is all so familiar. We have long been used to radios and hi-fi, cinema and television, computers that perform calculations and run simulations, and games consoles that provide interactive entertainment. A digital camera and a computer installed with ‘Photoshop’ allow a person to capture and preserve images in a way that is much more convenient than dealing with various chemical solutions in a dark room, but the end result will just be another photograph. It cannot have the impact that the invention of photography ITSELF had. How many members of the public know what technology is packed into a digital camera or even NEED to know? It is just a camera! Point and shoot.

However, none of this has any negative effect on the Singularity hypothesis. That is because its eventual realization (the establishment-by technology- of greater-than-human intelligence) is far more dependent on the ‘invisible’ innovations bringing about cumulative growth in the power and performance of IT. Computers do arithmetic using as few gates and switches as possible. In contrast, human calculation is a laboriously learned, ponderous and awkward behaviour in which tens of billions of neurons strain to process a digit a second. Computers can be made to recognize patterns that humans cannot recognise, learn behaviour that humans cannot learn, explore data too vast for human exploration, assemble logical reasoning too complex for humans.

But equally, humans recognise patterns machines cannot perceive, have learned behaviour machines struggle to perform, have evolved abilities that are executed with ease, while machines labour to produce dramatically inferior results. We have our means of processing information and acting on the results, machines have theirs. Whenever one encroaches on the other’s territory, the result is a change from ‘effortless’ to ‘laboured’ information processing. Search software can ‘read’ the data representing the tags we apply to images uploaded to sites like flickr and within mere seconds locate one image amongst millions. Contrast that with the sheer amount of time a person sifting through files and files of images would take to locate a particular photograph. But ask a person to classify objects within an image and the tables are turned. Computer vision still chokes on the ability to recognise objects, whereas a child can perform this task with no effort.

Of course, both examples of ’effortless’ information processing are in outward appearance only. Even now, while you sit quietly understanding the meaning of the words you see before you, immense work is being performed by the brain. And when you enter keywords into Google and it locates the data you requested, immense amounts of data mining had to occurr. Yet outwardly, scant effort could be discerned.

The Singularity is a result of technological advances that progressively reduce laboured information processing. Step by step, advances are making it progressively easier for people to work in cyberspace, while others are equipping computers with the skills to operate intuitively in realspace. Should this trend continue, a new form of intelligence will arise- one that finds both manifestations of ’thought’ (human thinking and machine thinking) effortlessly easy. That would be a dramatic result and there is a tendency to expect dramatic results only from dramatic causes. But Drexler showed that it is sometimes the dullest of facts that lead to the most dramatic of consequences. The fact that certain electrical switches can be made very small and have the ability to turn one another on and off does not seem particularly important. But when properly connected, such switches form computers, the engines of the information revolution.

Kurzweil commented that ’the kinds of scenarios I am talking about are not being developed because there’s one laboratory sitting there creating a human-level intelligence in a machine. They’re happening because of tens of thousands of little steps. Each little step is conservative, not radical…just the next generation of some company’s product.’ How small, how conservative can each step be? Maybe so undramatic that the majority of people live their daily life unaware that the advance has occurred. And what, really, is the difference between advances that produce scant noticeable impact, and no change whatsoever? The only real difference manifests itself as we project into the future. No progress is to be expected if nothing is advancing. But if advances are continually accumulating, building on the knowledge of the past, the result must be a future of dramatic change.

But will the change seem so dramatic if and when we arrive at the future? Maybe not. Think about it, how amazed are you to know that your computer’s hard drive can store gigabytes of data; that its CPU is clocked at gigahertz speed; that its graphics card performs realtime shading, hardware colour compression, dynamic gamma correction and adaptive texture filtering? Is it not incredible to think that the number of floating point operations per second performed by a single NVIDIA pixel shader is five times the projected world population in 2050? Actually, no. That was what graphics cards of 2003 were capable of but this is 2007 and we have achieved a lot more since then. A modern top-spec PC is significantly more powerful than the best PC of 1997, and compared to the computers of 1967 it is truly the stuff of science fiction. But the reason we don’t find this technology in any way overwhelming is because we judge it from the perspective of the previous generation- computers slightly less powerful. Those in turn were preceeded by similarly less powerful platforms and so it goes on, step by progressive step all the way down to the computers of 1967. In contrast, when forecasts project out to 2067 we leap to tremendously powerful systems, speculate that they will perform tasks with names as baffling to us as ’dynamic gamma correction’ might have been to the people of years past, and that these tasks will enable methods of work and play that stretch the imagination to breaking point. But we will arrive at this future in the manner we arrived at the present. Via a gentle climb up the technological slope.
Each step will seem conservative, not radical. The technology will not be incomprehensible, merely sensible.

Does this mean the Technological Singularity isn’t the world-changing event the hype suggests it will be? It is all a matter of perspective. Honestly, I believe that from OUR perspective we are being MUCH too conservative in our speculations regarding what the Singularity is. We should not merely be speculating on machines achieving ’consciousness’, that emergent property of hundreds of billions of brain cells. Nor should we search for it in ’computation’, that emergent property of zeros and ones. When both methods of information processing converge on an entity that finds both equally effortless, the result will be an entirely new form of cognition. No words exist to describe it, no thought can be expressed to encapsulate it, no computation can calculate it. When it comes to describing the capabilities of post-humanity, NO futurist ever succeeded. At best, we must content ourselves with crude comparisons. ’Like a chimpanzee trying to understand quantum loop theory’. ‘Like a sightless person trying to comprehend the colour blue’.

Personally, I find it impossible to visualize a ten dimensional geometric figure rotating left while simultaneously rotating right and therefore may be tempted to label the ability to perform such a feat ‘post-human’. But that would be a mistake because the imagery is not sufficiently incomprehensible to qualify. I may not be able to visualize geometric figures with more than three dimensions but I do know what a geometric figure is. I cannot imagine how an object might rotate in two opposing directions at the same time, but I do know what it means to ‘rotate’. To what extent should it be possible to break down post-human capability into understandable fragments? Vernor Vinge’s ‘fish’ analogy suggests an utter inability to grasp post-singularity concepts. ‘If you have a time machine, you could bring forward Mark Twain into 1997. In a day or two, you could explain everything to him about how the world works. (But) try to explain to a goldfish whats going on in 1997 and it would remain permanently clueless…progress after the Singularity will be fundamentally and qualitatively different to progress in the past’.

This sounds like the Singularity marks a dramatic division between graspable ‘human’ concepts over here, and incomprehensible post-human concepts over there. Indeed, there is some speculation that perhaps a routine twiddling of the parameters of a hitherto catatonic system will chance upon the right emergent pattern, resulting in a spontaneous leap to self-awareness. But maybe we should remember that the path to Singularity is measured in tens of thousands of tiny steps, and consider the illusion that there is a sharp discontinuity between a human being and a fish. It is an illusion because, in a historical context, no such discontinuity exists. Evolution tells us that there are lines of gradual continuity linking every species with every other. It is only because the intermediates are nearly always extinct that we can get away with catagorizing life forms into this species (fish) or that (human).

What reason is there to suppose that the intermediates of biologically inspired technology and technology-infused biology will represent anything like as patchy a record? Certainly today we cannot divide the human race into those with computers and those without. Instead we have a smooth continuum linking people with top-spec platforms, those with slightly less powerful systems, those with only partial access to pretty basic computers and those with no access to a computer of any kind. We really aught to expect as much with cyborgization: Some will have 0% technological enhancement, others will have 1%….5%…and so on all the way up to completely synthetic. How then, can we get away with dividing the world into, say, ‘robots’ and ‘humans’? And what about ‘biologically-inspired technology’? Numerous crude examples exist today, such as worms, cellular automata, genetic algorithms and artificial life. Nobody considers such code to be alive in a litteral sense, the names are purely metaphorical. But how are we to decide when the line has been crossed and metaphore has become literal? Scientists investigating the origin of life have given up the notion that life can be defined in absolute terms, since there is a progressive hierarchy of steps leading from a pre-biotic Earth enriched in organic molecules to cellular life. It is decidedly arbitrary to pin down the exact moment when any system of gradually increasing complexity becomes ‘alive’.

But what about the precise moment when our increasingly intimate relationship with increasingly capable computer networks converges on this hypothesized new form of thinking representing Singularity? It must surely be overly simplifying things to divide the world into those who have access to the technology and those that do not. Rather, there will be degrees of accessability of and integration with the myriad technologies that will comprise the metaverse, and that suggests varying ability to understand post-human thought.

Is that really any different to how things are today? Is it not the case that the world is already populated by people with varying degrees of comprehension? Some people do not understand maths at all. Others have a basic grasp, yet others are quite proficient and an elite few are gifted enough to see profound truths in mathematical equations, opaque to the rest of us. Or, rather, they used to. Some modern theorems rely on hundreds of specialised tools, each one of which will be understood by a member of the maths community, but who cannot understand the other parts. It is becoming increasingly necessary to rely on computer networks to check such proofs. One such example is the ‘Four Colour Question’, which is a deceptively simple conundrum. ‘What is the minimum number of colours required to colour any conceivable map, such that no two regions having a common border have the same colour?’ It was quickly determined that some maps require at least four colours and by the 1970s it was shown that any map with 39 regions or less needed no more than four colours. But still nobody could prove that no map amongst the infinite variation of maps might need five colours or more.

The challenge was eventually met by Wolfgang Haken and Kenneth Appel. Their method relied on a discovery of Heinrich Hesch, that a finite number of building-block maps could be used to construct the infinity of infinitely variable maps. This reduced the Four Colour problem to 1482 building block configurations and the possible colouring combinations within each map. If it could be proven that these building-block maps required no more than four colours, ALL maps would be four-colourable.

The duo employed a computer, but it required something more than the brute-force approach of simply checking the permutations. Even with a computer, that would take a century. The program would have to employ shortcuts and strategies to accelerate the map-checking procedure. After working on this program for 5 years, the two men noticed an interesting development:

‘At this point the program began to surprise us…It started to act like a chess-playing machine. It would work out compound strategies based on all the tricks it had been ‘taught’, and often these approaches were far more clever than those we would have tried. Thus it began to teach us things about how to proceed that we never expected. In a sense, it had surpassed its creators in some aspects of the ‘intellectual’ as well as the mechanical parts of the test’.

In 1976, after 1200 hours of computer time, the Four-Colour Map conjecture was finally proved. This was a milestone event in the history of mathematical theorems, because it marked the first time in which a computer did more than simply speed up the calculation. It had contributed so much to the result that the proof would have been impossible without it.

Since then, even more powerful methods of machine learning have found proofs of other longstanding conjectures, and computer scientist Ed Friedkin believes that one day a computer will discover an important proof independent of mathematicians. Another question it raises is perhaps this: Did the teamwork of those mathematicians and their computer represent the kind of transcendent thinking we might associate with the Singularity? One reason to believe that was not the case is the fact that neither human nor computer actually understand the theorem in the fullest sense. Traditionally, a mathematical theorem begins with a series of axioms, or statements taken to be self-evidently true, and then a logical argument is constructed that, step-by-step, arrives at a conclusion. Provided the axioms are correct and the logic is flawless, the conclusion is undeniable. This is the theorem.

Conventionally, every axiom and line of logical reasoning are scrutinized for mistakes, false assumptions etc in a peer-review process. But the Haken-Appel proof could not be proof-read by teams of human mathematicians because it is simply beyond the ability of humans to check a logical argument that relies so heavily on the unique capability of computers. Rather, the program was fed into an independent computer in order to verify that it produced the same result. We accept that the theorem is true, but we do not understand why, in the way we understand something like Pythagoras’ theorem.

So, humans cannot follow the chain of logic, but what about the computer? Surely it understands as it executes the program? But the comparison to a chess-playing machine is revealing, because an eminent philosopher has argued that, even though computers can now defeat the grandest of grandmasters, they do not ‘know’ chess. Berkely Professor of Philosophy John Searle’s ‘Chess Room Argument’ asks us to imagine a person (who does not know how to play chess) is locked in a room with a set of what-to him- are meaningless symbols. In actual fact, they represent chess pieces and positions on a board. The room also contains a rulebook that contains instructions for manipulating these ‘meaningless’ symbols. People outside of the room pass written instructions to the man, who dutifully follows the rulebook’s instructions, passing the results on to the people outside. Provided the instructions for handling the symbols are expertly written, one might imagine that the man’s answers represent competent chess moves. The people outside the room might well conclude that whoever is inside understands the game itself, but actually he does not. He does not know that the symbols represent chess pieces and board positions, and he does not know that the instructions he is following represent strategy. He does not ‘know’ chess at all!

If the man does not know chess on the basis of running a program (which is essentially what the symbols and rulebook are) neither does a computer solely on this basis. So, Deep Blue had a bunch of symbols that represented chess pieces, but the machine itself had no understanding of their meaning. It also had a bunch of equally meaningless symbols that the programmers used to represent possible moves. One could apply this argument to the computer that resolved the Four-Color conjecture. It was given a set of symbols that represented building-block maps and possible colouring combinations. The computer could not have understood that the program represented a mathematical theorem because (in Searle’s words) ‘the computer does not know anything’.

Searle’s opinion that computers do not know anything is based on his Chinese Room argument. It can be described in various ways, but all share these things in common. There is a room, inside which are one or more people who understand no Chinese. There are also a set of instructions, perhaps stored on a computer, perhaps written down in a book, or something else. Questions written in Chinese are given to the people, and by following the rules of the ‘program’ an answer to that question is produced, and passed on to the people outside. So, the people outside ‘input’ a question written in Chinese and receive an answer, also written in Chinese. They can only conclude that the people in the room speak the language but of course they don’t. They are merely following instructions for shuffling meaningless symbols. They do not understand the questions, they do not comprehend the answers. Therefore, argues Searle, any other computer ‘solely on the basis of carrying out the program’ doesn’t understand Chinese (or anything else) either.

However, this argument has a flaw. It looks for understanding in the individual parts of the system. Do the people in the room understand Chinese? We know they do not. Do the slips of paper containing questions or answers written in the language know Chinese? They know nothing, all they do is store a finite amount of information. Does the rulebook or computer know Chinese? Again, no, it merely helps store and process data. No single part of the setup ‘knows’ Chinese, then, but you cannot conclude from this fact that the setup in its entireity does not understand the language. ‘Understanding’ is spread across the entire system, and indeed this is what we would find with the brains of native Chinese speakers. No single part of their brain ‘knows’ anything a person knows, but clearly understanding arises as an emergent pattern of the entire system.

Searle’s thought experiment is not meant to show that a computer can never be programmed to converse in Chinese, but rather to show that computers cannot be capable of consciousness. His argument is that the people in the room have no conscious awareness of what the symbols mean. In fact, since no part of the room can possibly be aware of what the Chinese symbols mean, the system cannot be conscious. But can we really reach such a conclusion? One could, of course, argue that the people in the room are conscious because, after all, they are people and we generally accept that fellow humans are conscious beings. But they are not really necessary. The room is irrelevant, the people are irrelevant. The only necessary elements are A) the information and B) the means to process it. We tend to believe that simple systems are not conscious and Searle’s sleight of hand lies in the fact that he presents his Chinese Room as though it were a simple setup, just a bunch of symbols and a limited number of rules to be carried out. In fact, it would be a fantastically-massive set of instructions specifying in detail the operations carried out by the brain as we comprehend language. Think of a human brain, and its one hundred trillion interneural connections, each potentially processing information simultaneously. A machine designed to process information in such a way as to achieve the premise of Searle’s argument (answer any question posed to it) would have to be as complex as the brain itself.

Assuming we ever build such a device, can we conclude that it is NOT conscious? No doubt, some will claim it is, while others will insist it instinctively follows instructions but has no inner awareness. Such arguments are heard today with regards to other animals. Unfortunately it is the case that arguments are all we have, because there can be no scientific test that conclusively determines if consciousness is present or not.

The so-called ‘hard problem’ of consciousness need not concern us. We need to look at the Chinese Room’s distributed understanding of language and ask if there is a similarly distributed understanding of the Four Colour Problem in the brains of the mathematicians and the computers running their program. Obviously, the answer is ‘yes’ but is this transcendent knowledge? To transcend means ‘to go beyond the range of ‘(experience, belief, etc). You might be able to argue that the compound strategies discovered by the computer transcended Haken and Appel’s expectations of what such a machine was capable of, but a human’s understanding of concepts like ‘map’ and ‘colour’, coupled with a computer’s ability to run through more details in an hour than a human can hope to do in a lifetime cannot, in this instance, add up to properly transcendent knowledge. It would be quantitively, but not qualitatively, different.

Here’s what I mean. The basic explanation for why this conjecture is true can be understood by anyone. There are 1482 building-block configurations, from which all possible maps are constructed, sort of like the way the infinite variations of written English are derived from a finite set of 26 letters. Because these building blocks never require more than four-colours, all maps are necessarily four-colourable. As a human, you could sit down and work through a few examples. With a computer’s speed and ingenious shortcuts you could work through ALL combinations. But all you will have done is taken the knowledge ‘the maps I have checked are four-colourable’ and expanded it to ‘all maps are four colourable’. Why this should be so has not changed.

Another example, again from the field of mathematics. Fermat’s Theorem (that there are no whole number solutions to X^n + Y^n = Z^n where n is any number higher than 2) is now known to be true because such a solution would contradict the Taniyama-Shimura Conjecture. That is, ‘every ecliptic curve must be related to a modular form’. The mathematician Gerhard Frey proved that, if a whole number solution to the equation did exist, it would have to be in the form of an ecliptic curve that could never be related to a modular form. After an heroic effort, another mathematician called Andrew Wiles proved that the Taniyama-Shimura conjecture is true, meaning Frey’s hypothetical equation does not exist, and that therefore there IS no whole number solution to Fermat’s equation.

I had never even heard of ‘modular forms’ until I read about the 358 year-long effort to prove Fermat’s Last Theorem, and yet apparently they are one of the five fundamental operations (the other four being addition, subtraction, multiplication and division). In a real sense, though, modular forms are quite distinct from the other fundamental operations. It’s possible to visualise the other three, and in fact doing so is a standard technique when learning basic maths (‘if John has five marbles and gives Jane two, how many does he have left?’) You cannot, however, draw or even imagine what a modular form looks like, because it exists in the upper half-plane of hyperbolic space. Hyperbolic space is four-dimensional, which is profoundly difficult for humans to visualize, constrained as they are to percieve three dimensions at the most. But, just because you cannot visualize higher-dimensional space, that does not mean the theorem ITSELF is as unvizualizable.

There ARE whole number solutions to X^2 + Y^2 = Z^2 (you might recognise that as Pythagoras’ Theorem). In visual terms, it means you can find two squares that can be added together to form a third, larger, square. For instance, if you had a square made up of 9 tiles and another square comprised of 16 tiles, in total there would be 9 (3^2) + 16 (4^2) = 25 or 5^2. But what about the equation X^3 + Y^3 =Z^3? If you had one cube made up of 216 building blocks and another cube made up of 512 building blocks, can the sum total of building blocks make a single, larger cube? No. 8 times 8 times 8 = 512 and 6 times 6 times 6 = 216. Both are cubed numbers. But add 512 to 216 and you get 728, when what you NEED is 729 (the cube of 9 times 9 times 9). You ‘cube’ made up of all 728 blocks has one piece missing.

Ok now let’s visualize the impossibility of finding a whole number solution to X^4 + Y^4 = Z^4. What’s that? You cannot imagine what a four-dimensional object looks like? Oh, never mind, remember the cubes that didn’t quite have enough building blocks to construct a single, larger cube? This is the same thing, basically. Again, the ability to visualize more than three dimensions would make your understanding of Fermat’s Theorem quantively, not qualitatively, different. It’s all just examples of the same principle.

What the hell am I talking about? I am talking about a common fallacy that creeps up in books etc whose topic is ‘Technological Singularity’. Having explained why smarter-than-human intelligence implies profound unknowability, the authors then go ahead and imagine a post-human world anyway. Always, it is examples of technology, knowledge, and capability quantitively (but not qualitatively) different to life as it is today. Lifespans are extended. Computers are more powerful. Ailments that cannot be cured are less. The depth to which the senses can be immersed in virtual worlds is greater. The amount of material wealth the average family can accumulate is increased. And so on.

These forecasts are often attacked by critics as being hopelessly naïve. They are right, but almost always for entirely the wrong reason. The most popular objection is to point out the startling similarities between the promises of Heavenly Paradise touted by religion and the speculations of Life After Uploading (or molecular nanotechnology, or advanced genetic engineering etc) forecasted by the likes of Ray Kurzweil. We have long dreamed about this utopian world with not much to show for it beyond some suckers staring with rapture towards an imagined better future. But, really, the similarities of these scenarios with religious memes actually INCREASES the likelihood they will one day be realised. Why? Because they are deeply-held dreams, not likely to be diminished by the cold reality of failure after failure. This is comparable to another fantasy that was laughed-off by legions of naysayers as self-evidently absurd: The dream of achieving flight, which also managed to persist no matter how many failures the pioneers had along the way.

What makes these predictions naïve is not that they are unlikely (although, the notion that technological change can create a utopia certainly is). What makes them naïve is the belief that they give us a sense of what a post-human society will be like. Frankly, there are nuch better examples of transcendent knowledge actually existing RIGHT NOW in Sub-Saharan Africa. It is also an example that ridicules the anthropocentric backslapping that is the belief that human beings are the most successful and advanced species on the planet. We may pride ourselves to think we discovered architecture, engineering, central heating, agriculture, and huge economies from division of labour amongst specialists, but we did not. Termites discovered all these things an exceedingly long time before our species evolved.

Necessity is the mother of invention. Sub-Saharn Africa is much-too hostile an environment for termites, and this has pushed them to evolve incredibly complex societies capable of ingenious technological solutions. These are manifested in their nests, which are among the most spectacular of all animal constructions. If a termite were as tall as a person, their mounds, enlarged on the same scale, would be four times the height of the Empire State Building. Their enterior contains eleborate networks of tunnels, spanning a total length of six kilometres. These ducts and tunnels give the internal structure of termite mounds an appearance not unlike that of our own vessels and respiratory channels. Richard Dawkins has taken the analogy a step further, describing this network as ‘mud-fashioned organs, including an ingenious ventilation and cooling system’. There are spectacular combs of fungus growing in termites’ nests. Called ‘Termitomyces’, it’s a delicate and finickity organism that is absolutely dependent on the care and cosseting it receives from its hosts, who cultivate it as a source of food.

If termites could talk, one can imagine them saying, “humans! They think they are so clever, but their civilization is founded on non-renewable energy sources and they find it almost impossible to believe that renewables can supply enough energy. But we termites represent a biomass that far outweighs the human populace- 500 kgs of termites for every person alive!- and our 100 million year-old civilization uses construction methods that utilize nothing BUT renewable energy sources’.
Termites don’t hold conversations like that. They have no concept of ‘agriculture’, ‘architecture’ or ‘engineering’. They have never heard of energy, renewable or otherwise. You might wonder how they can design their homes to perform functions they cannot possibly understand, but then you built a system capable of conscious awareness despite the fact that what causes consciousness (or even what ‘consciousness’ is) are very much a mystery. To be precise, your DNA instructed sub-cellular mechanisms to build your brain and body, and therefore an entity capable of self-illumination.

The genetic instructions hard-wired into termites codes for a blueprint for constructing their elaborate home, distributed among the millions-strong workforce. Simplistic rules guide the behaviour of each termite, and these individual elements interact and organize themselves into a collective intelligence quite capable of meeting the challenge of constructing their impressive home and organizing their society. A termite mound has many of the attributes of a single large and voracious organism, complete with its own anatomy and physiology.

The difference between the collective knowledge of the colony and the individual insects that comprise it is profound. The theorems I talked about earlier also relied on collective knowledge. Wiles’ proof of the Taniyama-Shimura conjecture relied on the techniques of generations of mathematicians, and the Four-Colour Problem resulted in a deep collaboration with computers. Full understanding of either theorem is beyond the capabilities of most people. Less than ten percent of number theorists fully grasp Wiles’ proof, and the solution to the Four-Colour problem is one that NO human mind could get across in a lifetime. But, as we have seen, it is possible to represent both theorems so that they can be understood by the layperson, albeit in far less detail than a full reading of the proofs would provide. But when it comes to the termites, it would be simply impossible to describe the utility of their nest in terms simple enough for them to comprehend. It is quite beyond their capacity to understand.

Vinge’s ‘fish’ analogy argues that the same limitation exists for human-scale comprehension; that there exists knowledge of a complexity above and beyond a person’s capacity for understanding. Presumably the eventual engineering of vastly superior intelligences will yield minds as far beyond an un-augmented human’s as ours is to a termite’s. These powerful intellects will ask questions about phenomena as difficult for us to comprehend as String Theory is for an insect. But what does such a profoundly unknowable concept feel like when it enters a brain quite unable to model it? How CAN such a concept enter the mind if the brain has no capacity to model it in the first place? We know that people who have had certain neural pathways severed not only loose the ability to see colours, but also to remember or even imagine them. The profound unknowability implied by Vinge’s ‘fish’ analogy suggests an incomprehension beyond a blind person’s inability to imagine a blue sky. By any other measure, such people are as capable of understanding as the rest of us. Apart from a missing visual cortex, they share the same neural architecture as the rest of us, after all. But what would it be like to share an existence with intelligent beings whose neural architecture is as far beyond ours as ours is to an ant’s? Such creatures busy themselves, seemingly unconcerned about the awesome technological feats those God-like bipedal ‘post-ants’ have amassed around their nests. This is probably because their brains are simply incapable of computing such concepts, in which case (as far as the ant’s subjective view of the world is concerned) they don’t exist.

Some futurists imagine super-artificial intelligences will act as oracles for humanity, but why would post-humans waste efforts attempting to explain things we absolutely could not grasp, and if we cannot possibly express such concepts, or even hold them within the semantic primitives our brains evolved to model reality with, what reason is there to suppose the arrival of post-humanity will be a profound impact and not an uncomprehending silence that is equivilent to a singularity that never happened? Will we, in fact, just busy ourselves with our daily routines, as the ant’s society continues, uncomprehending and so caring nothing for the post-ant civilization?

We know that social insects like termites display a collective intelligence, to the extent that the nest itself could be thought of as one single entity. Similarly, Vinge speculated that ‘large computer networks (and their associated users) may “wake up” as a superhuman entity’, noting that the transition of the Internet from the world’s largest database to something more like a global brain was ‘proceeding the fastest and may run us into the singularity before anything else’.

Continues in part 2

Posted by on 08/26 at 01:15 PM