An event such as the creation of a computer (IBM WATSON) capable of playing Jeopardy that consequently played and won against two human contestants whom were seasoned and acclaimed Jeopardy champions, brings to mind what advancements have been made in modern computer science and to which extent the discovery of machine learning reshaped our views on Artificial Intelligence. The match between the humans and the machine was yet another first in human history as a computer competed against man in a language game, a language that some have said could not be understood in any meaningful way by a computer programme as asserted by Searle (1990, pg.27) in his claim that “symbol manipulation by themselves are not sufficient to guarantee the presence of meanings.” It would be prudent to bring to the forefront of our attention once more the implication of these advancements and ask the question: Can machines think?
In 1939 Alan Turning proposed what would be the basis of what is called a computer which is referred to as the Turning machine (Baum 2004). The conception of such an idea laid the ground work for modern computer sciences and changed what humans would traditionally refer to as machines. Turing (1950) contemplated the question, “Can machines think?” Examining the question itself, Turning formulated what he termed The Imitation Game. Simply put the Imitation Game was a test where an interrogator tried to distinguish which of two participants is a machine from which is human.
Baum (2004) argues that the mind is a computer programme and that an adequately programmed machine could indeed think. If the issue of whether machines can think is raised in a purely philosophical context and one is a Functionalist or maybe even a Behaviourist, then one might be inclined to be of the view that machines can think. In light of this, I will analyse the view that machines can think and in doing so, assess whether this is a sound position to hold for those whom have this view.
What is thought? In trying to answer a question such as this in philosophy might pose some problem as it is difficult to come up with necessary (required) and sufficient (guarantee) conditions that tells us exactly what thought is. I can however clearly point out what it is not. If I was the say that when asked- What is one plus one? You must first think about it to know the answer is two. This to me would be a mistaken view and is an example of what thought is not. If I was to say this is an example of thinking then it could be easily concluded that computers can know the answer is two when asked and therefore computers can think, while at the same time raising another issue “Can computers know?” One might even have issues with the assumption that one even needs to think about the question to know the answer, it could simply be that the question requires an assertion of facts and there is only one fact that exists-the answer is two. From this we can rule out thought being anything that comes before the mind or processed by the brain. Dewey (1910) expounded on a different notions of thinking as, “thought denotes belief resting on some basis, that is, real or supposed knowledge beyond what is directly present” (Dewey, 1910. p. 4). I will modify Dewey’s definition somewhat, and replace the word belief with the word conclusion. This is a fair modification because to say thoughts are beliefs is quite limiting because I can think of something I do not believe in. This kind of thinking can be analytic or critical thinking. My modification should give a better understanding of what it means to think in general. Thinking is thus: drawing conclusions resting on some basis that is real or supposed knowledge beyond what is directly present.
The idea of what is meant by machines is a much simpler task than the former. Rather than talking about machines in general it is better to talk about machines in specific as related to kinds of machines. Why do this you might ask? The answer is simple, defined generally man are machines and man can think, therefore machines can think. My argument would end there and it doesn’t seem very satisfying. So to specify, machines are: synthetically made bodies lacking biological components and constituting of digital circuitry. In this definition, a robot with a human brain wouldn’t be deemed a machine; neither would a synthetic biological body.
In light of the case being machines can think! One might feel inclined to ask: Can machines think? Turning reformulated the question into what he referred to as using “relatively unambiguous words” (Turning, 1950, p. 433). The new form of the problem was described as a game termed the “Imitation Game”. In it there were three people A (man), B (woman) and C (interrogator). The goal was for C to determine the sex A & B. A was permitted to cause C to make the wrong identification while B was permitted to help C. The new form of the question was thus: “What will happen when a machine takes the part of A in this game?” and “Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman?” (Turning, 1950, p. 434). Turning (1910) went on to predict that an average interrogator would not be able to certainly distinguish A from B in five minutes, if A is a machine. If an answer was to be attributed to Turning on whether machines can think, to me it’s apparent that it would be in the affirmative. However, to be charitable to the claims of Turning, the assertion of a thinking machine would be dependent on whether that machine passes the Turning test. A machine passing the test was indeed possible to Turning and he believed at the time of his writing that within fifty years time it would be possible for a machine to pass his test (Turning 1950).
Eric Baum is a proponent of what is called Strong Artificial Intelligence (Strong AI). Strong AI is the view that any system whatsoever not only might have thoughts but must have thoughts, provided only that it is implemented with the right programme, with the right input and output (Searle, 1990). Eric Baum (2004) clearly makes this point in even asserting that evolution and life itself are programmes. His main thesis was that “Semantics is equivalent to capturing and exploiting the compact structure of the world, and thought is all about semantics” (Baum, 2004, p. 3). Baum underlying view is that complex things arise from simple things strung together, for example, the human body from a collection of base particles Adenine, Thymine, Guanine and Cytosine (DNA). The DNA thus would be the compact structure of life or compact programme which would then be executed according to rules. The execution of a programme to Baum is pure syntax. In other words, given the right input or rules DNA makes a man or even a tree. All life has the same compact structure that is DNA but the rules or syntax is what matters in the execution of the programme. Meaning for Baum is thus the execution of a programme since by itself DNA has no meaning. Given the right rules then meaning arise from the outputting of a man or tree. Thus all mind is to Baum is meaning. The mind works by exploiting semantics. In other words the mind is a well written programme for manipulating the meaning that arise from compact structures and exploiting those structures. The programme of the mind is so well written because evolution had a long time to write it. Thus for Baum’s, a machine is capable of thinking if given the right programme.
John Searle, disagree with the ideas of Strong AI but did not out rightly deny that machines could think. His view was that the mind could not be a computer programme and wasn’t. One only think his mind is running a computer programme but once a conscious agent is going through the steps of the programme then it ceases to be an implementation of the programme at all (Searle, 1990). Searle argues that brains cause minds, and any system capable of causing mind must have causal powers equal or more than brain. He clearly states after these claim that, “this does not imply that only biological systems could think” (Searle, 1990, p. 29). He went on to say, any artifact or artificial brain capable of producing mental phenomena must be able to duplicate the specific causal powers of brains and it could not do that by just running a programme. Searle also thought that semantic could not be equated to syntax. He aptly put it as, “syntax by itself is neither constitutive nor sufficient for semantics” (Searle, 1990, p. 27).
Robert A. Freitas Jr’s idea is on sentience because Intelligence as a measure is so varied not only in forms but degree that it is quite difficult to make a definitive assessment base on it. There are in fact humans that have the intelligence of monkeys. Freitas came up with a measure of consciousness itself or awareness that should be applicable to any being of intelligence. He called this the Sentient Quotient (SQ). SQ is based on the rate of processing information in response to a stimulus. According to Frietas the dumbest brain possible has a SQ of -70 while the Smartest would have a SQ of +50. Humans have a SQ rating of +13 and all insects and mammals cluster within several points of this range. On the other hand computers such as Apple II have a SQ of +5 and Cray at +9. Vegetation SQ ranges around a SQ of -2. From speculations of other scientists regarding computers, Freitas calculates a SQ of +23 for a Superconducting Josephson junction electronic gate computer (Freitas Jr., 1984). The point Frietas makes with all this it that if humans and animals level of awareness or SQ vary so little yet we consider ourselves superior to animals and even more so plants. We cannot even understand the way of life of animals without long term study and effort or relate to them. Then what could we say about a SQ computer of +23 or better yet what could we possibly say about a computer that passes the Imitation Game.
Turning was trying to map the workings of the brain of a mathematician as he tries to solve an equation. The mathematician, Turning refer to as a computer steps in solving the problem was accurately written down using simple rules and could solve problems as well as the thinking mathematician. Because something derived from thought was replicated using programmes then therefore it would follow that thoughts can be replicated using programmes because programmes are minds. This would be one argument made by proponents of Strong AI but still it lacks something to me. It is not lacking in the sense that it does not assert that machines can think but it is not a sound enough argument. The proof that computers can think, in the arguments of Strong AI are not that minds are programmes but rather thoughts are programmes. Thoughts are thus one of many properties of the mind and in this regard Baum’s argument is only partly correct and needs some adjusting. In my adjustment of Baum’s claim, thoughts would be the programme and not the mind. In this manner machines can have thoughts, it is yet to be seen if they can produce them. Baum is too caught up in his ideas that programmes are minds so much so that if you should follow his arguments closely you are led to conclude that the universe has a mind with the compact structures being atoms. With this it would follow that all things have a mind. The adjustment of his idea as stated, means thoughts can be transferred but, not all things would have thoughts because they lack a programme.
The origin of thoughts would be a good basis to question my ideas as related to the adjustment of Baum’s argument. With this, Searle point of being consciously aware of the programme and its workings would make it not a programme is a strong point. From this he could argue that consciousness gives account to the origin of thought while my adjustment of Baum’s argument was not. He could also say that I’m just playing with words in that programmes are not thoughts just as much as they were not minds. Searle could argue that thoughts and programmes being the same still don’t accounts for meaning. Since programmes are syntax then so must thoughts and thoughts clearly have meaning. I suppose Searle would argue this since he argues it against the claims of Strong AI however, I do not think this argument is a very sound one. He supposes that thoughts have meaning in themselves by his assertion that syntax has no meaning and because this is so any claim that syntax is the same as semantics would be false. I believe in saying this Searle is ascribing properties to thoughts that they don’t necessarily have; namely meaning. Thoughts rather are bestowed with meaning from consciousness just like syntax. With this said, meanings are subjective because it can be seen that the same objects evoke different meanings; Therefore, an analysis of meaning to decipher whether a machine is thinking or not is pointless. Both Searle’s and Baum’s arguments are not sound enough to get to the issue in a definitive manner but they both give some insight.
The only approach is ascertaining whether a machine can think or not is a behaviouralist approach. This approach could be best found in the “Imitation Game”. Why this view you might ask? Baum took the ideas of Turning to a functionalist perspective that caused more minds popping up than we can even be sure of. This was all due to his definitions and likewise for Searle. Turning circumvented the problem of definition and reduced the issue of a thinking machine down to a simple observable event to which he asserted that the question “Can machines think?” is pointless. This was an ingenious approach because it circumvented issue of other minds or solipsism thus simplifying the issue. He reduced thinking to a matter of intelligence and therefore anything intelligent must can think or do something analogous to it. By think I mean as defined earlier.
After the discovery of machine learning, advances such as the creation of WATSON has been made. Watson might not be able to pass the “Imitation game” but it does show signs of intelligence. By intelligence, I mean the capacity to learn. If a machine can learn from previous actions to the extent that they operate differently than before in similarly situations then the argument is this:
1) The new reaction is indeed a conclusion
2) The basis was previous interactions
3) The new conclusion was beyond the programming of the computer since it wasn’t ever programmed for that reaction.
An analysis of WATSON suggests that the questions asked required some interpretation even by a man. WATSON was able to correctly interpret the questions and provide answers based on those questions which showed some signs of Intelligence. Since any intelligent thing must have thoughts then machines can think. The only thing of significance I would consider is the SQ of the computer. Given that its SQ is lower than that of a man then the relation to machine thinking might be difficult to make. I would suppose that given an Increase in SQ then it will become more certain that machines can think because it will be able to pass the Imitation game.
Baum E. B. (2004).What is Thought. Illinois: Bradford Books.
Freitas Jr., R. A. (1984). Xenopsychology. Analog Science Fiction/Science Fact, Vol. 104, April 1984, pp 41-53
Searle, J.R. (1990). Is the Brain’s Mind a Computer Program? Scientific American, Vol. 262, No. 1, January 1990, pp. 26-31.
Turing, A. (1950). Computing Machinery and Intelligence. Mind, Vol. 59, No 236, October 1950, pp. 433-460.