• Contribute
• What's New
• Submit News

Issac Asimov wrote many tales of robots being put in sensitive situations. In one of these stories a lawyer must help a judge to decide if the robot deserves the equal legal representation of a human. Most people object to this notion because of the idea that artificial intelligence is not yet conscious. The irony is that humanity is unwilling to settle on a definition of consciousness. This is because it is impossible to prove another being, human or artificial, is conscious and hence is impossible to have an artificial being “become conscious”.

To many, simply saying consciousness is impossible to prove isn’t enough. Many require proof that it can’t be proved. As ironic or circular-like logic as this may be, I shall attempt to prove this by explaining the three main theories of consciousness; classicalist, connectionist, and naturalist. After I explain each theory I will give practical examples and describe the problems with these examples. The prime argument against each theory will be presented after this. However, my focus will be on not destroying the mystique of consciousness, but rather help redefine it. To quote Edgar Allen Poe, “They who dream by day are cognizant of many things which escape those who dream only by night”.

The first potential view of consciousness is the classicalist’s outlook. The view is that the underlining structure of the human brain is a system of singular representations and correlations between those representations.Often times a classicalist will bring up the idea of speech to illustrate their point of view. For example, in the sentence “The dog ran fast” classicalists believe the mind separates out the key points (dog ran fast) and looks for the first idea (dog). After the system finds this idea it then looks for the second representation (ran) and any additional modifiers (fast).

Current computational practice for the classicalist system would be Natural Language Processing (NLP). NLP allows for a better level of communication between a computer and a human. Humans naturally speak and think using their native language, which the computer can parse, separate, and define. Such systems include chatbots, robotic interfaces, and even higher-end computer games.

One of the main arguments about successful NLP is the processing power required to parse a large database using multiple techniques. This is invalidated when the process is split among multiple computers. This method is known as parallel computing.

Often times the classicalist viewpoint is also linked with the idea of the “Cartesian Theater” because of the structures classicalists assume about the brain. The Cartesian theater is the idea of a center of communication within the brain between the mind/soul and the body. It is called the Cartesian theater after it’s creator Rene Descartes. This argument however opened the first “can of worms” when finding the definition of consciousness.

To say that one believes in the Cartesian theater has become a sure sign of naivety within the scientific community. This is because of logistical implications to this theory. Most neurologists understand that the brain spreads an idea out over a network of neurons because of the inefficiency of having a center of command. In order for a connection to be made using the Cartesian method, through any significant amounts of data, a computer needs to have a great amount of speed. This becomes more and more of a burden when the number of casual links increases because the complexity increases exponentially. These systems then begin to suffer a rapid decline in quality because of partial or different information. (Hutchinson) “Economy is important to biological systems, as the maintenance of neurons and their connecting fibers consume valuable resources.” (Spitzer, 1999, Pg 121)

This draws many neurologists towards the connectionist theory. This theorem states that fundamental thought is spread across many localized points using a massively parallel computation. This allows the brain to not only connect separate localized points, but also compensate for slow neuron speeds by sending several million signals at once. (Hutchinson)

The current applications for the connectionist theorem are more subtle than that of the classicalists. Using Artificial Neural Networks (ANN) a computer system can decide anything from the amount of laundry in a washer to the amount a robotic arm must move to achieve it’s goal in a factory. This method can also be used to help a robot adjust its motors as it learns to walk, run, or even climb up stairs. It also allows robots to find edges and shapes, which can be used for self-navigation.

Critics point at the idea of common sense against ANN because common sense “needs no training time”. However, this is not logical when put in the light of our own minds. No human has a genetic memory and therefore common sense must be learned somewhere. Therefore a real neural network needs training time just as much as an artificial one does.

Although usually reserved for arguments against NLP, the “Chinese room” also demonstrates the essential flaw in the connectionist theorem: creating information and resolving information are two completely separate things.

Imagine two rooms with a window in between. A person is put in a room and told that when a set of symbols come through the window, match the set of symbols with it’s corresponding second set of symbols by using a key. After that, write down the second set on a piece of paper and put it back out the window. Now if those symbols were actually Chinese, it would appear to the first writer that the man responding knows Chinese. This however, is not the case as the responding man is simply resolving information and not creating it. (Mathews, 1999)

The third theory is that ideas are not created, but simply evolved. Darwin’s theory of evolution has three main components. The first is that mutation causes a new idea/gene. The second component is that the new idea or gene allows the being to thrive in its environment. The last component to this theory is that the being spreads this gene to its offspring. This theory is referred to as either Darwinism or naturalism and is mostly used in computers by applying a genetic algorithm.

Most naturalist applications are used for drafting. Using components based upon real world transistors, capacitors, and resistors a virtual engineer can create a circuit board specialized for certain economic and performance factors. A programmer may also use this idea to test several versions of code against themselves to make the key parts work better together. An entrepreneur may use this idea to find the cheapest method possible of manufacturing their product. A pharmacological company may test multiple molecular combinations while creating a new drug. A composer can finish a genius’s symphony.

One could argue about the lack of true computational randomness, especially with regard to when new generations are created in a genetic algorithm. However, even a human can’t pick a truly random number because of the bias of different mental faculties presented upon their creation. Someone who is more mathematically inclined may choose a larger “random number” then one who is not.

This theory’s Achilles heel is the assumptions that must be made when first forging a new genetic algorithm. In order to successfully extrapolate the naturalist theory into practice one must assume they understand the development and variation rates needed (also known as the mutation per generation rate). A high mutation rate will create high diversity and broadly reduce the potential of a well-rounded result. A low mutation rate will take too many generations of permutations in order to produce any number of significant results. The key to success is the balance of gain and mutation. (Horst, 2001)

As with each of these three main theories, balance is the key. With a classicalist system the need to verify that the connections between ideas to decrease computational overhead. The connectionist system needs balance so as to not overly specialize, which could cause a lack of adaptable change. As stated before, the naturalism system needs a fine balance in its mutation/generation ratio. However, balance only ensures gains from an artificial system and does not fix the holes that are still plaguing classicalist, connectionist, and naturalist theories.

“If we can’t establish precisely what a person is, it matters little whether we are one or not.” (LeDoux, 2002, Pg 19) We know a person is not based on a set of singular representations and correlations because of the power and quality of connections needed for this system. We know a person is not based upon a neural network alone as this type of system only allows refinement of data and not its creation. We know a person is not simply their genes because of an ever-shifting need for mutations. Humans compensate for this mutation need by learning new skills and acquiring multiple jobs in their lifetimes.

If you ask a person on the street if they are conscious, a likely response would be yes, followed by any one of several explanations. However, if you begin to look at these multiple explanations, soon you would realize that they are very different from each other. How is it that we can inherently know something without being able to articulate it? Does this mean consciousness is an illusion or simply a theory that has eluded us? Whatever the case may be, until we can fully define the essence of consciousness we cannot expect our creations to magically gain an artificial conscious. This is the prime reason why artificial consciousness remains a myth to this day.

LeDoux, J. (2002). Synaptic self. Penguin Books.

Spitzer, M. (1999). The mind within the net. MIT Press.

Hutchinson, L. Classical and Connectionist Cognitive Models. Retrieved April 25^th, 2004 from http://artificialintelligence.ai-depot.com/Essay/Cognitive.html

Mathews, J. (1999, December 13) Philosophical Arguments for and Against AI. Retrieved April 25^th, 2004 from http://www.generation5.org/content/1999/ai_phil.asp

Horst, S. (2001) Evolutionary Explanation and Consciousness. Retrieved April 26^th, 2004 from University of Phoenix, Online Library, Keywords Evolutionary Consciousness: http://mycampus.phoenix.edu