I'm Conscious, You're Faking

Would a Digital Brain Have a Mind?

No problem can be solved from the same level of consciousness that created it.

- Albert Einstein

The task of science is to stake out the limits of the knowable, and to centre consciousness within them.

- Rudolf Virchow

by Neville Holmes

Certain recent events caused me to doubt whether I know my own mind or not. Let me explain.

Last week, the first of our academic year, all first-year students in our degree program underwent a supervised test in which they pull an old computer to pieces and put it back together again. We give this test to put a healthy disrespect for digital circuitry - which is, at heart, only carefully polluted sand - into each student's mind as early as possible. We intend this disrespect to counter the superstition, held both by naive students and by members of the public susceptible to media persuasion, that digital machinery has much in common with the human brain.

Yesterday, I went to a lunchtime philosophy club lecture titled "Why the Body Is the Mind." Because some of the discussion related to consciousness, I recalled Giorgio Buttazzo's article, "Artificial Consciousness: Utopia or Real Possibility?" (see below). The juxtaposition suggested a strange contrast between computing people, who see mental capabilities in machines because they do not appreciate how complex the human brain is, and philosophers, who see complexities in the human mind because they do not appreciate that the brain and the computer share some simple and fundamental properties.

However, not being a philosopher, I find it difficult to be confident that I understand them when they discuss the mind. This uncertainty leaves at least three possibilities:\

	either I understand the basic nature of my mind and the philosophers don't,
	vice versa, or
	we share the same understanding but express it in mutually incomprehensible language.

I Later, I read Bob Colwell's provocative essay "Engineering, Science, and Quantum Mechanics" (Computer, February 2002, pages 8 - 10). Toward his essay's end, Colwell reported of entanglement theory that "the [photon's] wave function's actual point of collapse is when a conscious mind perceives the results" and that the collapse was caused by "the synapses of our brains, acting in concert to form our minds, at the instant we detected the photon."

Suddenly, I felt alone, isolated, out of my depth, and fearfully vulnerable. What follows is meant to enlist your sympathy and rebuild my confidence.

The Mind as Process

According to my Macquarie Dictionary (Macquarie Library Party Limited, Sydney, 1981, www.macquariedictionary.com.au), the principal meaning of mind is "that which thinks, feels, and wills, exercises perception, judgment, reflection, et cetera, as in a human or other conscious being: the processes of the mind." The principal definition of process is "a systematic series of actions directed to some end."

An action requires an actor, presumably the mind in this case. Why not the brain? The Macquarie defines the brain as "the ... nerve substance that fills the cranium of man and other vertebrates; centre of sensation, body coordination, thought, emotion, et cetera." Why so coy? Where is the mind in this brain?

The computer works like the mind-as-actor in that it functions as a device that processes data - conventional representations of facts or ideas. The circuits carry out the computer's processing by copying, transmitting, and transforming these data.

The mind, given the Macquarie definition, processes thoughts, feelings, intentions, perceptions, judgments, reflections, and so on. Neurons and glial cells process neural and hormonal representations of sensations past and present.

Although the idea of the mind as distinct from the brain has a natural appeal, defining the mind as an actor distinct from the brain invokes an unnecessary, even deceptive, dualism. Our mind thinks, feels, perceives, and so on, whereas our brain merely exists between our ears. The distinction can be useful and productive.

The problem lies in defining the mind as an actor rather than an action. If we regard the mind as the thinking process, it becomes distinct from the brain and becomes the systematic series of actions the brain takes as it processes sensations and reconstructs memories.

We can thus view the brain as substance, the mind as process. Likewise, we can view the computer as substance and its computations as process. The brain exists materially, while the mind arises as a property of changes within the brain. Similarly, the computer exists materially, while computation occurs as a property of changes within the computer. So far, so simple.

Source of Confusion

Consciousness seems to be the confusing factor. We associate self-awareness and identity with consciousness. Buttazzo writes, "Because we cannot enter another being's mind, we cannot be sure about its consciousness." This theme recurs in writings on the philosophy of the mind. But if a mind is a process, it's meaningless to talk about anything entering it.

Processes can only be perceived, thus inferring the operation of another mind from such perception must surely be sufficient. If anyone argues that we need certainty, we can counter that no one can be certain of anything, as the "brain in a vat" argument shows (www.artsci.wustl.edu).

Cycle time versus data rate

In speculating about the effect of cycle time on artificial consciousness, Buttazzo poses a curious question, "If consciousness emerges in an artificial machine, what will time perception be like to a simulated brain that thinks millions of times faster than a human brain?" What does "be like" mean here? In any case, the question confuses cycle time with data rate - the stupendous parallelism of the brain makes the cycle time of our present digital computers irrelevant.

However, Buttazzo speculates that the world might seem to slow down for a simulated brain as perhaps it does for a fly, "thus giving the fly plenty of time to glide out of the way" of a swatting hand. Let's apply some numbers to this speculation. The fly and I have much the same kind of neural signaling system. The average local fly measures about 10 mm, and I am roughly 200 times that length. My reaction time is about 1/10 of a second. A sub-millisecond reaction time for a fly is thus not at all mysterious, nor would much shorter reaction times in a digital computer be in any way puzzling. Where then is consciousness in all this?

Whence consciousness?

Is a fly conscious? Well, it's aware to the extent it can often dodge a swat - its perceptual neural system alerts it to the swatting hand so that it can dart out of harm's way. But what part of the fly's nervous processing is aware, and thus to some degree conscious?

Awareness must emerge at least from the transformation of perception into the intent or neglect of an action. The transformation of sensation into perception can be unconscious because it can be automatic: we remain, for example, cheerfully oblivious to the dramatic data compression our retinas carry out. The transformation of intention into motion can similarly be unconscious: we do not consciously stimulate each individual muscle in our mouth, throat, and chest as we speak.

For the fly, we might imagine that its nervous system functions like a computer system: its central processor "consciously" forms intentions on the basis of perceptions that its peripheral sensory system "unconsciously" produces, then its peripheral motor system "unconsciously" puts those intentions into effect.

Cache as cache can

We can transfer this analogy to the human nervous system to explain how consciousness arises from it, except that we have a more complex central processor than does a fly and much more occupies our minds than mere perceptions and intentions. We have extensive memory traces from which neural processes can reconstruct pseudoperceptions that pass through our consciousness. Intentions cannot be based practically on reconstructing all our possible memories at once. Part of our brain processes perceptions together with relevant pseudoperceptions to derive intentions. The processing of this area within the forebrain must be closely allied to our consciousness.

Human consciousness therefore strongly resembles the processing that a digital computer's central processor and its associated main store cache carry out. The cache brings relevant data close to where the current computation can use them. In this sense, then, our present-day computers are conscious, if their CPU has a cache.

Vive la Difference

Observing that "The human brain has about 10¹² neurons, and ... 10¹⁵ synapses," Buttazzo calculates that, using artificial neural networks, "Simulating the human brain requires 5 million Gigabytes" of data storage. Moore's law suggests that digital computers will have main stores of that capacity by 2029, although Buttazzo carefully qualifies this observation by adding that it "refers only to a necessary but not sufficient condition for the development of an artificial consciousness."

Using a digital computer to simulate an artificial neural network that simulates the human brain does not seem the best approach. Given the neural parallelism to be modelled, using analogue circuits to directly implement neural networks would seem a better alternative, one that might bring the feasibility date well forward, if research could divert the circuit manufacturing industry to this cause. But this begs the question of whether artificial neural networks can be made comparable with real ones.

In an artificial neural network, each node, or neuron, has an activation value that the network passes forward to other nodes through connections, or synapses. The synapse to each forward node has an associated weight that modulates the incoming activation value's effect on the forward node's activation. The weights can be adjusted in various ways likened to "learning." An artificial neural network's nodes mimic the classical neuron - but very roughly - with an axon down which the repetition rate of an action potential, the spike, passes an analogue value, dependent on the activation of the neuron's main body.

Disregarding as a mere production problem attaining 10¹² neurons and 10¹⁵ synapses in analog circuitry, where do artificial neural networks fall short of the real thing?

	The activation of a classical neuron is not an arithmetic sum of the synaptic effects. Rather, a complex process involving the intervals between action potentials at individual synapses and their relative timings between synapses determines the activation. The more neurons are studied the more such complexity is revealed.
	The human nervous system contains many different kinds of neurons and many kinds of glial cells. The glial cells provide more than support because they signal and have synapses just as neurons do.
	Action potentials alone do not control nervous signalling. Graded potentials and hormonal signalling also play a part, as does the great variety of different neurotransmitters and hormones.

Creating an artificial consciousness does not require simulating these complexities. But human consciousness lies far beyond any presently contemplated artificial one. We have, for example, developed a highly complex and utterly human consciousness of our physical bodies. Likewise, we have a highly developed consciousness of other people and of our society, whose collective consciousness shapes our development as humans.

Obvious parallels exist between the brain and digital computers. To fulfill their responsibility to themselves, and to others who might be misled by journalistic hyperbole, computing professionals should have well-founded opinions about the extent of these parallels. The profession should refrain from applying humanistic names to its mechanistic endeavours, and it must be conscious always of the essential differences between people and computers.

Perhaps in 50 or 100 years, our machines will acquire a humanlike consciousness and intelligence. But such machines will be utterly different from the puerile imitations we now have or can realistically design. Getting to such machines will raise professional and philosophical issues quite different from those that reflect on the nature of what human and machine consciousness can generate.

And what about the mind and entanglement theory? If a CPU and its cache possess consciousness, we could leave looking at photons to them. In any case, I've now put entanglement theory into the back of my mind under the shade of the tree in Bishop Berkeley's quad.

Neville Holmes is an honourary research associate at the University of Tasmania's School of Computing. Contact him at neville.holmes@utas.edu.au or www.comp.utas.edu.au/users/nholmes/prfsn.

Artificial Consciousness: Utopia Or Real Possibility?

by Giorgio Buttazzo, University of Pavia

Since the beginnings of computer technology, researchers have speculated about the possibility of building smart machines that could compete with human intelligence. Given the current pace of advances in artificial intelligence and neural computing, such an evolution seems to be a more concrete possibility. Many people now believe that artificial consciousness is possible and that, in the future, it will emerge in complex computing machines.

However, a discussion of artificial consciousness gives rise to several philosophical issues:

	Can computers think or do they just calculate?
	Is consciousness a human prerogative?
	Does consciousness depend on the material that comprises the human brain, or can computer hardware replicate consciousness?

Answering these questions is difficult because it requires combining information from many disciplines including computer science, neurophysiology, philosophy, and religion. Further, we must consider the influence of science fiction - especially science fiction films - when addressing artificial consciousness. As a product of the human imagination, such works express human desires and fears about future technologies and may influence the course of progress. At a societal level, science fiction simulates future scenarios that can help prepare us for crucial transitions by predicting the consequences of significant technological advances.

Robots in Science Fiction

Since the early 1950s, science fiction movies have depicted robots as sophisticated human-crafted machines that perform complex operations, work with us on safety-critical missions in hostile environments or, more often, pilot and control spaceships in galactic travels.[1] At the same time, however, the film industry has portrayed intelligent robots as dangerous entities capable of working against humanity in their pursuit of self-serving agendas.

Robotic rampages

The most significant example of this archetype, HAL 9000, is the main character in Stanley Kubrick's 1968 epic, 2001: A Space Odyssey. Although HAL controls the entire spaceship, talks amiably with the astronauts, plays chess, renders aesthetic judgments, and recognises the crew's emotions, it also murders four of the five astronauts in pursuit of a plan elaborated from flaws in its programming.[2]

More recent films, such as James Cameron's Terminator and the Wachowski brothers' The Matrix, present an even more catastrophic view of the future in which robots become self-aware and dominate the human race. For example, Cameron's 1991 film, Terminator 2: Judgment Day, begins with a scene depicting a horrendous war between humans and robots, which onscreen text forecasts will occur in Los Angeles, 2029 AD. According to Cameron's script, the fictional corporation Cyberdyne becomes the us military's largest computer systems supplier. It then morphs into Skynet, a powerful neural processing network built to execute strategic defense decisions. The network, however, becomes self-aware; when human engineers try to deactivate it, the network retaliates by unleashing the US's nuclear arsenal on its creators.

Robotic rescues

Few movies depict robots as reliable assistants that serve humans rather than conspire against them. In Robert Wise's 1951 film, The Day the Earth Stood Still, Gort is perhaps the first robot - albeit an extraterrestrial one - that supports a humanitarian agenda by helping its alien owner deliver an offer of peaceful coexistence to Earth. Likewise, Cameron's 1986 film, Aliens, shows a synthetic android that acts on behalf of its human owners despite their suspicion of it.

When SF and reality collide

Strongly influenced by theories on connectionism and artificial neural networks, which seek to replicate processing mechanisms typical of the human brain,[3] Cameron's Terminator represents the prototypical imaginary robot. The robot can walk, talk, perceive, and behave like a human being. Its power cell can supply energy for 120 years, and an alternate power circuit provides fault tolerance in case of damage. More importantly, Terminator can learn. Essentially, a neural processor - a computer that modifies the robot's behaviour based on past experience - controls it. Intriguingly, the neural processor is so complex, it learns at an exponential rate, eventually becoming self-aware. Thus, the film raises important philosophical questions about artificial consciousness. Can a machine become self-aware? If so, how can we verify that an intelligent being is self-conscious?

Turing Test

The work of computer science pioneer Alan Turing may help answer these questions. In 1950, Turing tackled a similar problem focused on intelligence. To establish whether we can consider a machine as intelligent as a human, he proposed the now-famous Turing test. The test uses two keyboards - one connected to a computer, the other positioned in front of a human operator. Both computer and operator are hidden from view, with only a monitor visible to display their output.

An examiner inputs questions on any topic that comes to mind. Both the computer and the human respond to each question. If the examiner cannot with confidence distinguish between the computer and the operator based on the nature of their answers, we must conclude that the machine has passed the Turing test.

In 1990, the Turing test received its first formal acknowledgment. Hugh Loebner, a New York philanthropist, and the Cambridge Center for Behavioral Studies in Massachusetts established the Loebner Prize Competition in Artificial Intelligence ( www.loebner.net). Loebner pledged to award a $100,000 prize for the first computer whose responses could not be distinguished from a human's. The first competition took place at the Computer Museum of Boston in November 1991.

Although the contest was constrained to a single narrow topic for some years, since 1998 the questioning's scope has been unlimited. After the conversation, each judge scores the interlocutor on a scale of 1 to 10, in which 1 means human and 10 indicates computer.[4] So far, no computer has given responses totally indistinguishable from a human's, but every year the computer's scores edge closer to an average of 5. Nevertheless, current computers can pass the Turing test only if we restrict the interaction to highly specific topics - like chess.

Deep Blue conquers chess

In 1997, for the first time in history, a computer beat reigning world chess champion Garry Kasparov. Like all computers, however, the victor - IBM's Deep Blue - does not understand chess; it simply applies rules to find a move that leads to a better position, according to an evaluation criterion programmed by chess experts.

Claude Shannon estimated that the search space in a chess game includes about 10¹²⁰ possible positions. Deep Blue could analyze 200 million positions per second ( www.research.ibm.com). Exploring the entire search space for Deep Blue would therefore take about 10⁹⁵ billion years. Nevertheless, Deep Blue's victory can be attributed to a combination of speed and a smart search algorithm, which gave the computer positional and material advantages.

Although mathematics clearly shows that Kasparov succumbed to brute-force computation rather than sophisticated machine intelligence, in interviews during and after the match he expressed doubts that his opponent was a computer and reported that, at times, he felt as if he were playing against a human. Kasparov also remarked on the beauty of the computer's moves, ascribing aesthetic motivations to what were, essentially, the results of raw computation. If we accept Turing's view, we can say that Deep Blue plays chess intelligently, but we must also admit that the computer no more understands the meaning of its moves than a television understands the meaning of the images it displays.

Computers challenge humans in other domains

In addition to chess, computers have begun approaching human ability in an increasing number of other domains. In music, for example, many commercial programs can create melodic lines or entire songs according to specific styles, ranging from Bach to jazz. Other programs can also generate solos atop a given chord sequence, emulating jazz masters like Charlie Parker and Miles Davis much better than an average human musician could.

In 1997, Steve Larson, a music professor at the University of Oregon, proposed a musical variation of the Turing test. He asked an audience to listen to pairs of classical compositions and determine for each pair which one was written by a computer and which was the authentic composition. The audience classified many computer pieces as authentic compositions and vice versa. A loose interpretation of these results could indicate that, with regard to musical composition, the computer passed the Turing test.[5]

Computers now approach human levels of understanding in continuous speech, electrocardiogram diagnostics, theorem proving, and aircraft guidance. In the future, we can expect similar levels of computer performance in areas that include complex tasks such as driving, real-time language translation, house cleaning, surgery, surveillance, and law enforcement.

However, even if machines become as skilled as humans in many disciplines, such that we cannot distinguish between their performance and that of humans, we cannot assume that they have become self-aware. At the same time, we cannot assume that such machines are not self-aware. In fact, while intelligence is an expression of an external behaviour that we can measure with specific tests, self-consciousness is a property of an internal brain state, which we cannot measure. Hence, to resolve this question, we must turn to philosophy.

Philosophical Views of Self-Awareness

From a purely philosophical perspective, we cannot verify the presence of consciousness in another brain, either human or artificial, because only the possessor itself can verify this property. Because we cannot enter another being's mind, we cannot be sure about its consciousness. Douglas Hofstadter and Daniel Dennett[6] discuss this problem at length in their book, The Mind's I.

Nevertheless, we can develop theories regarding the nature of self-awareness based on different philosophical approaches.

Pragmatic

We could follow Turing's approach and say that we can consider a being self-conscious if it can convince us by passing specific tests. We base our belief that humans are self-conscious on our inherent similarities: because we have the same organs and similar brains, it is reasonable to conclude that if each of us is self-conscious, so is everyone else. If, however, the creature in front of us, although behaving like a human, were comprised of synthetic tissues, mechatronic organs, and neural processors, we might respond differently.

The most common objection to granting electronic-circuit-driven computers self-conscious status is the perception that, working in a fully automated mode, they cannot exhibit creativity, emotions, or free will. A computer, like a washing machine, is a slave operated by its components.

Logic demands, however, that we must apply this reasoning to machines' biological counterparts. At a neural level, the same electrochemical reactions present in machinery operate in the human brain. Each neuron automatically responds to its inputs according to fixed laws. However, these mechanisms do not prevent us from experiencing happiness, love, or irrational behaviours.

With the emergence of artificial neural networks, the problem of artificial consciousness becomes even more intriguing because neural networks replicate the brain's basic electrical behaviour and provide the proper support for realising a processing mechanism similar to the one adopted by the brain. In Impossible Minds: My Neurons, My Consciousness, Igor Aleksander[7] addresses this topic in depth and with scientific rigour.

Religious

If we remove the structural diversity between biological and artificial brains, artificial consciousness becomes a religious issue. If we believe that divine intervention determines human consciousness, no artificial system can ever become self-aware. If, instead, we believe that human consciousness is a natural electrical property developed by complex brains, realising an artificial self-aware being remains an open possibility.

DualIstic

Many believe that consciousness is not a product of brain activity, but rather a separate immaterial entity, often identified with the soul. Seventeenth-century philosopher Rene Descartes developed such a dualistic theory about the brain and mind. However, this theory raised several unanswerable questions:

	If a mind is separate from its brain, how can it physically interact with the body and activate a neural circuit?
	If the mind operates outside the brain, how does it move atoms to create electrical pulses?
	Do mysterious forces activate neural cells?
	Does a mind interact with a brain by violating the fundamental laws of physics?
	If a conscious mind cali exist outside the brain, why do we have a brain?
	If emotions and thoughts come from outside, why does a brain stimulated with electrodes and drugs respond by generating thoughts?
	Why does the patient experience severely affected conscious behaviour when a portion of a diseased brain is surgically removed?

These and other concerns caused this theory to fade from popularity in the philosophical community.

Reductionism and idealism

To resolve dualism's inconsistencies, researchers developed alternative theories. At one extreme, reductionism did not recognise the existence of mind as a subjective, private sense-data construct and considered all mental activities as specific neural states of the brain. At the other extreme, idealism tried to reject the physical world by considering all events as a product of mental constructions.

With the progress of computer science and artificial intelligence, scientists and philosophers developed a new approach that considers the mind a form of computation emerging at a higher level of abstraction with respect to neural activity.

The major weakness of the reductionist approach to comprehending the mind is that it attempts to recursively decompose a complex system into simpler subsystems, until at some stage the units can be fully analysed and described. This method works perfectly for linear systems, in which any output can be seen as a sum of simpler components. However, a complex system is often nonlinear. Thus, analysing a system's basic components offers insufficient understanding of its global behaviour. Complex systems contain holistic features that cannot be seen at a smaller detail level, appearing instead only when we consider the structure and interactions among the system's components.

Paul Davies, in God and the New Physics,[8] explains this concept by observing that a digital picture of a face consists of many coloured dots - pixels. But the shape takes form only when we observe the picture at a distance that allows us to see all the pixels. Thus, the face is not a property of the individual pixels but of the set of pixels.

The brain as ant colony

In Goedel, Escher, Bach,[9] Douglas Hofstadter explains this concept by describing a large ant colony's behaviour. Ants have a complex and highly organised social structure based on work distribution and collective responsibility. Although each ant has minimal intelligence and limited capabilities, the whole ant colony exhibits a highly complex behaviour. In fact, building an ant colony requires a large and intricate design, but clearly no individual ant can conceptualise the entire project. Nevertheless, a scheme and a finalised behaviour emerge at the colony level. In some sense, we can consider the whole colony a living being.

A brain resembles a large ant colony in many respects. For instance, a brain consists of billions of neurons that cooperate to achieve a common objective. The interaction among neurons is tighter than among ants, but the underlying principles are similar - work subdivision and collective responsibility. Consciousness is not a property of individual neurons, which automatically operate as switches that respond to input signals. Rather, consciousness is a holistic property that emerges and flourishes from neural cooperation when the system reaches a sufficiently organised complexity.

Consciousness and matter

Although most people can accept the existence of holistic features, others still believe that consciousness cannot emerge from a silicon substratum because it is an intrinsic property of biological materials such as neural cells. Thus, we can reasonably ask: does consciousness depend on the material that comprises neurons? In Beyond Humanity: CyberEvolution and Future Minds,[10] Gregory S Paul and Earl Cox write:

It would be astonishing that the most powerful information-processing tool can be derived only from organic cells and chemistry. Aircraft [are] made out of different materials from birds, bats, and bugs; solar panels are made out of different materials from leaves. There is usually more than one way to build a given type of machine... This just happens to be what genetics was able to work with... Other elements and combinations of elements may be able to do a better job processing information in a self-aware manner.

If we support the hypothesis of consciousness as a physical property of the brain, the question becomes: when will a computer become self-aware?

A Prediction

Attempting to provide an answer to this question is hazardous. Nevertheless, we can determine at least a necessary precondition without which a machine cannot develop self-awareness. This precondition derives from the assertion that, to develop self-awareness, a neural network must be at least as complex as the human brain.

Why this assertion? Because it appears that less-complex brains cannot produce conscious thought. Consciousness seems to represent a step function of brain complexity and the human brain provides the threshold, as Figure 1 shows.

How much memory would a computer require to replicate the human brain's complexity? The human brain has about 10¹² neurons, and each neuron makes about 10³ synaptic connections with other neurons, on average, for a total of 10¹⁵ synapses. Artificial neural netWorks can simulate synapses using a floating-point number that requires 4 bytes of memory to be represented in a computer. As a consequence, simulating 10¹⁵ synapses requires a total of 4 million Gbytes. Simulating the human brain requires 5 million Gbytes, including the auxiliary variables for storing neuron outputs and other internal brain states.

When will such a memory be available in a computer? During the past 20 years, random-access memory capacity increased exponentially by a factor of 10 every four years. The plot in Figure 2 shows the typical memory configuration installed on personal computers since 1980.

By interpolation, we can derive the following equation, which gives RAM size as a function of the year:

bytes = 10^{(year - 1966) / 4}

For example, from this equation we can derive that in 1990, personal computers typically had 1 Mbyte of RAM, whereas in 1998, a typical configuration had 100 Mbytes of RAM. Assuming that RAM will continue to grow at the same rate, we can invert this relationship to predict the year in which computers will have a given amount of memory:

year = 1966 + 4 log 10 (bytes)

To calculate the year in which computers will have 5 million Gbytes of RAM, we substitute that number in the equation above. This gives us the year 2029. Ray Kurzweil,[5] Gregory S Paul and Earl Cox,[10] and Hans Moravec[11] derived similar predictions.

To understand the calculated prediction, we must take into account several important considerations. First, the computed date refers only to a necessary but not sufficient condition for the development of an artificial consciousness. The existence of a powerful computer equipped with millions of gigabytes of RAM is not in itself sufficient to guarantee that the machine will become self-aware.

Other important factors influence this process, such as the progress of theories on artificial neural networks and the basic biological mechanisms of mind, for which it is impossible to attempt precise estimates. Further, some could argue that the presented computation was done on personal computers, which do not represent top-of-the-line technology. Others could object that the same amount of RAM could be available using a computer network or virtual-memory management mechanisms to exploit hard-disk space. In any case, even if we adopt different numbers, the computation's basic principle remains the same, and we could advance the date by only a few years.

How about Moore's Law?

Some may object that the 2029 prediction relies on mindless extrapolation of current trends, without considering events that could alter that trend. Gordon Moore, one of Intel's founders, noted the exponential growth of computing power and memory in 1973. He predicted that the number of transistors on integrated circuits would continue to double every 18 months until reaching fundamental physical limits. That prediction proved so accurate that it is called Moore's Law.

But how much longer will this law hold true? Chip companies estimated that Moore's Law will be valid for another 15 to 20 years. However, when transistors reach the size of a few atoms, the conventional approach will not work, and this paradigm will break down. What's next? Will the microprocessor evolution come to an end around 2020?

Computing power's exponential growth

Some people, including Ray Kurzweil[5] and Hans Moravec,[11] noticed that computers were growing exponentially in power long before the integrated circuit's invention in 1958, despite the hardware used. So, Moore's Law was not the first but actually the 5th paradigm to track computing's exponential growth. Each new paradigm came along precisely when needed, suggesting that exponential growth will not stop when Moore's Law is no longer valid. For example, scientists are investigating new technologies such as three-dimensional chip design, optical computing, and quantum computing[12] that promise to extend computing power's exponential growth for many years to come.

Related Issues

Other important aspects of artificial consciousness include the possibility of achieving seif-awareness in sequential machines and the notion of time for a fast-thinking mind.

Sequential machines

Could a sequential machine develop seif-awareness? If consciousness is a product of a highly organised information-processing system, that property does not depend on the hardware substratum that performs the computation and thus could also emerge in a sequential machine. Indeed, sequential computers simulate most artificial neural networks today because such computers are more flexible than a hardwired network.

Someone could argue that a simulated process differs from the process itself. Clearly, this is true when simulating a physical phenomenon, like a thunderstorm or a planetary system. However, for a neural network, a simulation and a physical network produce the same result because both are information-processing systems. Similarly, the software calculator applet that runs on personal computers is functionally equivalent to its hardware counterpart and can perform the same operations.

The notion of time

Does consciousness depend on the response time of the computing elements? It is hard to say because we cannot change our brain's speed. Experiments on real-time control systems suggest that processing speed is important to meet the timing constraints the physical world imposes on our actions, but it should not affect the results of a computation. In other words, a different neural speed would certainly give us a different perception of time but perhaps would not prevent us from being self-aware.

If we could ideally slow down the neurons uniformly in our entire brain, we would perhaps perceive the world as a fast-motion movie in which events occur faster than our reactive capabilities. This sounds reasonable because our brain evolved and adapted in a world in which the important events for reproduction and survival fall within a scale that spans a few tenths of a second. If we could speed up the events in the environment or slow down our neurons, we would not be able to operate in real time in such a world and probably would not survive.

Conversely, how would we feel with a faster brain? Today a logic port is six orders of magnitude faster than a neuron. Biological neurons respond within a few milliseconds, but electronic circuits can respond within a few nanoseconds. This observation leads to an interesting question: If consciousness emerges in an artificial machine, what will time perception be like to a simulated brain that thinks millions of times faster than a human brain?

It is possible that, for conscious machines, the world around them will seem to move slower. Perhaps insects view the world this way, so that, to a fly, a swatting human hand looks like it is moving in slow motion, thus giving the fly plenty of time to glide leisurely out of the way.

Gregory S Paul and Earl Cox[10] address this issue in Beyond Humanity: CyberEvolution and Future Minds:

...a cyberbeing will be able to learn and think at hyperspeed. They will observe a speeding bullet fired from a moderate distance away, calculate its trajectory, and dodge it if necessary... Imagine being a robot near a window. To your fast thinking mind, a bird takes what seems like hours to cross the field of view, and a day lasts seemingly forever.

Why should we build a self-aware machine? Except for ethical issues that could significantly influence progress in this field, the strongest motivation for constructing a self-aware machine is the innate human desire to discover new horizons and enlarge the frontiers of science.

Further, developing an artificial brain based on the same principles as in the biological brain would provide a means for transferring the human mind to faster and more robust support, opening the door to immortality. Freed from a fragile and degradable body, a human being with synthetic organs, including an artificial brain, could represent humanity's next evolutionary step.

Such a new species - a natural result of human technological progress - could quickly colonise the universe, search for alien civilisations, survive to the death of the solar system, control the energy of black holes, and move at the speed of light by transmitting the information necessary for replication to other planets. As has proven the case with all important human discoveries - from nuclear energy to the atomic bomb, from genetic engineering to human cloning - the real problem will be keeping technology under control. Should self-aware computers become possible, we must ensure that we use them for human progress and not for catastrophic aims.

References

[1.] G Buttazzo, "Can a Machine Ever Become Self-Aware?" Artificial Humans, R Aurich, W Jacobsen, and G Jatho, eds., Goethe Institute, Los Angeles, 2000, pp.45-49.

[2.] D G Stork, ed., HAL's Legacy: 2001's Computer as Dream and Reality, MlT Press, Cambridge, Massachusetts, 1997.

[3.] I Asimov, I, Robot, Grafton Books, London, 1968.

[4.] M Krol, "Have We Witnessed a Real-Life Turing Test?" Computer, March 1999, pp. 27-30.

[5.] R Kurzweil, The Age of Spiritual Machines, Viking Press, New York, 1999.

[6.] D R Hofstadter and D C Dennett, The Mind's I, Harvester/Basic Books, New York, 1981.

[7.] I Aleksander, Impossible Minds: My Neurons, My Consciousness, World Scientific, River Edge, NJ, 1997.

[8.] P Davies, God and the New Physics, Simon & Schuster, New York, 1984.

[9.] D Hofstadter, Goedel, Escher, Bach, Basic Books, New York,1979.

[10.] G S Paul and E Cox, Beyond Humanity: CyberEvolution and Future Minds, Charles River Media, Rockland, Massachusetts, 1996.

[11.] H Moravec, Robot: Mere Machine to Transcendent Mind, Oxford University Press, Oxford, UK, 1999.

[12.] L Geppert, "Quantum Transistors: Toward Nanoelectronics," IEEE Spectrum, September 2000, pp. 46-51.

Giorgio Buttazzo is a professor of computer engineering at the University of Pavia, Pavia, Italy. His research interests include real-time systems, advanced robotics, and neural networks. Buttazzo received a PhD in computer engineering from the Scuola Superiore Sant' Anna of Pisa. He is a member of the IEEE. Contact him at buttazzo@unipv.it.

Source: Computer: Innovative Technology for Computer Professionals July 2001