All life is nucleic acid. The rest is commentary. Isaac Asimov Consider the following aspects of human intelligence: consciousness, memory, abstract reasoning and emotion. Discuss the relative difficulty of creating an artificial version of each in a computer. Consciousness, Memory, Abstract Reasoning and Emotion: Four of the pillars upon which we humans lay the foundations of the claim upon which we take so much pride, that we are sentient beings. We are able to remember what happened in the past, thus, to predict the outcome of an event on the basis of how similar ones developed previously, and that is an example of abstract reasoning. But more than that, perhaps even more a sophisticated type of thought, Emotion and Reasoning. What are emotions? Generations of philosophers, scientists, poets and psychologists have attempted finding an answer to the question, so as to that of Consciousness, but as of today, the question is still very much disputed, and no final answer has been found, or is close to having been found. Despite that, all agree that both consciousness and emotion are a fundamental constituent of any human being. This essay will first attempt to hypothesize how Memory and Abstract Reasoning could be implemented in a machine; It will explore some of the possibilities, and challenges which could be encountered in the process. Thereafter, we will move on to tackling Consciousness and Emotion, attempting, if not to define, at least to describe what these are and whether they could be reproduced on artificial foundations (computer hardware). 1
Memory, the ability to recall events which occurred in the past, but beyond that, also the ability to remember facts, information, experiences, feelings; Summarizing, the possibility of examining the trace that something which happened in the past left with us. Some may argue that present day machines do in fact possess memory, they can store and retrieve information rapidly and, possibly, more accurately than we can do. They can retrieve information about past events (Eg: A system crash, a user logging into a system) and even about the external world (Eg: Temperature monitoring of internal components, power loss ) and should, therefore, be considered capable of memory. However, a more in-depth reflection may encourage us to believe that memory goes beyond simply storing and retrieving information, and should include also other aspects to this related. As for one thing, whereas a computer will memorize all and only the information it is instructed to memorize (until it runs out of storage space), and retain it until instructed to do otherwise (or forced to, for example by a physical damage), the human mind s way of memorizing isn t exactly as straightforward. We could in fact say that our brain prioritizes certain memories over other ones, more easily retaining some specific memories, and more prone to fade out other ones. Exactly how this selection works is still mostly unknown, and it is not to be excluded that it may vary from person to person, but it is likely that aspects which are evaluated when considering a memory s importance include: Age of the memory, that is, how far back was the memory captured, more recent memories are more likely to be recalled than older ones. A possible explanation for this is that events which happened in the nearer past are more likely to be relevant to the present moment.; 2
Uniqueness of the memory, how many other memories has the mind already captured that are similar to this one? A possible explanation for this is that the mind clusters memories. By grouping together all the memories of bus trips in one month, it can recreate a memory of an average bus trip. Being there no particular reason to remember each individual trip, the average memory will suffice in providing coverage for that set of events; Emotional impact of the memory. In fact, we could almost think of memory competition, with different memories competing among each others to be stored. This leads to the final part of the discussion on memory, and that is, subjective preferences. Whereas age of memory and uniqueness of memory are mostly objective, the emotional impact is indeed subjective. Therefore, whether realizing it or not, humans would be probably be making an active decision in deciding which memories to promote retention for, and which, on the other hand, they are fine to forget. Whereas it may be possible to implement certain aspects of this type of memory in computers, we would first have to discuss whether it would be advantageous. Imagine a computer which deletes older bits of memories because they have not been used in some time. Imagine these bits being your files. True, some computers do optimize disk space by proceeding to the removal of older information, but these techniques are often very much limited to a well-defined set of information, and very hesitantly implemented as they increase the risk of eliminating important data. Imagine a computer creating an average representation document out of all your files. The key element here is that computers are not able to identify the significance of documents and files, therefore to decide which are ok to united in an average copy, which can be deleted, and which others must be preserved at all costs. 3
On the other hand, if a computer was able to evaluate the significance of a document, of a file, and of implementing a system of memory similar to that of the human mind, this would probably make the entire process of artificial information storage more efficient. This leads to the second aspect taken into analysis: Critical Thinking, or Abstract Reasoning. The two are linked it that it is difficult to imagine Critical Thinking possible without the ability to pursue Abstract Reasoning, therefore, if Critical Thinking could be implemented, it would follow Abstracted Reasoning had been implemented too. Again, it is hard to give the term Critical Thinking (CT from now) a clear and final definition; for the aim of this paper, however, we will content ourselves by saying that Critical Thinking is the ability to evaluate the best course of action needed to move from Condition A, the present state, to Condition B, the desired state. This should be done by taking into account limiting factors such as external impedances, physical limits, resource availability Therefore, CT is the ability to hypothesize a series of solutions to a problem, and to evaluate the best one. As stated earlier, the issue of implementing CT in a machine inherits all the issues which implementing a more advanced form of memory, a selective memory, already faced us with. The simplest form of machinery is that which, given an input, performs a series of pre-defined tasks upon/based on such input, consequently developing an output. However, this form of thought can by no means be defined as Critical Thinking as, in fact, the machine has no part in devising or the method, it simply executes a series of predefined instructions. The machine does not have to generate an algorithm to solve the problem, nor does it have to evaluate its efficiency and, in fact, this type of machine is the most likely to generate an error based upon an invalid input. We could summarize as follows: Critical Thinking could be defined as achieved when a machine was able to make a genuine attempt at understanding and solving an unforeseen problem. 4
Lacking a true CT mechanism, instances which required a simulation of such resolved to use a particular technique to emulate it; this, I will call exhaustion of cases, and involved the programmers enumerating all possible scenarios the program could have been faced with, therefore instructing the program on how to react to them. This could have been useful in projects which only had to give them impression of Critical Thinking, and which had a limited number of possible scenarios. (Eg: Artificial Intelligence in videogames.) Of course, the main weakness of such technique was that, in fact, no CT was actually involved, only an emulation of such. The second technique, which I will call trial and error, is more similar to the human way of learning. It involved the user submitting an input, and the software applying an operation to it, therefore returning an output. The user will then tell the software whether the output is correct or, preferably, whether the output is closer to the correct answer than the previous one was. This method assumes that by learning more and more about its environment, the machine would familiarize itself with the obstacles but it put forward, and therefore be able to derive solutions to them. This method may be helpful in controlled environments, such as teaching a computer the game of checkers, in that at each move the computer does we could tell it whether it is a valid move, and whether somebody has won the game, but it would be less useful in a full-world analysis where, to say, input comes from a webcam, a microphone, and various sensors (eg: gyroscope, proximity sensor) In this second case, many issues would arise: 5
How do I define Input? In a simulation of checkers, an input would be considered a move on my part. That is a discrete input. It has a beginning and an end, and can be easily defined. But what about if I wanted the computer to learn to recognize a gesture I make with my hand? How would the machine know that the input began or ended, or which part of the input to consider? (Eg: My hand, rather than a small insect flying in front of it.) The human brain acquires these abilities over the first few years of birth,, however, the way in which this is done is still largely unknown, and a better understanding of this would make it easier to simulate a similar process within machines, hence, making machine learning possible and, consequently, the development of CT. This leads, however, to the next question: A machine may be able to accurately and efficiently remember, it may be able to critically think and devise strategies to solve unforeseen problems, but why would it do this. A brief reflection, could easily allow us to see that computers will not do anything unless they are explicitly instructed to do so. This is quite different from living beings, which will perform several tasks/actions without the need of any superior guide. This self-motivation derives from the presence of stimuli, which in turn come from necessities. The most primordial: Food, Water, Shelter, Reproduction and all physiological ones, and more elaborate ones such as shelter, protection, love and self-realization. (See Maslow s Hierarchy of Needs) These are, of course, absent within a computer. One may argue that the reason for which they are absent is because a computer doesn t need food, water, reproduction etc. but the argument is, in fact, invalid. We could easily imagine other theoretical needs which would apply to a computer such as: Constant Electric Supply, Efficiency of Internal Components, Free-from-Corruption data, Internet Access 6
It can be noted that all these needs or stimuli appear to be aimed at one thing: Preservation. Living beings appear to have a want and a need to remain alive. Because of this, stimuli are enacted whenever a condition necessary to the continuation of life is lacking. Eg: Lack of water within body, thirst. Similarly, stimuli are also enacted to promote positive action. Eg: Sexual pleasure for reproduction. More complex beings also presented more elaborated forms of stimuli, which are commonly defined as emotions. These occur when the psychological aspect of the being is threatened, rather than the physical one. Breaking up from the loved person doesn t necessarily harm the body in a physical way, but will almost undoubtedly cause emotional pain. Similarly, an A* in an exam will just as much not affect you in a physical way, but will cause pleasure and excitement. The reason for this is probably that, as organisms become more and more complex, psychological balance becomes as important as physical one. Whereas it is unclear why we appear to be programmed for self-preservation (An answer to that would probably imply the answer to life ), it seems clear that these stimuli, which are often elaborate enough so as to be called emotions, have a fundamental role in instilling self-motivation within a machine. Following, just as memory implied the need for critical thinking, so do emotions imply the need for consciousness. The very expression, to feel pain/joy/etc states that you need to perceive the emotion acting upon yourself, to be aware of it, and the very moment you are aware of the emotion, you are aware of it affecting you, therefore, you are aware of yourself. We could therefore state that, once consciousness is recreated, and we can assume consciousness to be based on a biological structure, then by affecting such biological structure (Eg: Hormones on Receptors) we could recreate emotion. By appropriately administering emotions, or even better by creating an autonomous system which regulates them, we could recreate self-motivation, which could be fulfilled through the use of aspects such as the previously described memory and critical thinking. 7
The final question, therefore, remains that concerning how to implement consciousness. But, before answering that, a further questions demands answering: What is consciousness? The generally accepted definition of consciousness suggests that consciousness occurs when a being can not only analyse inputs from the external environment, but is aware of being doing so; that is do so, when the individual can look upon their own thoughts, can reflect upon the very fact it is reflecting, therefore is aware of its own existence. Whenever that occurs, seen the being is aware of itself, it may also decide upon what it wishes to do with/to itself, this possibility can be further promoted, as seen earlier, by enacting emotion. The question of how to recreate consciousness is, however, perhaps the most difficult of those analysed so far, as well as the most important, as memory, critical thinking and emotion relay on it. In truth, a full understanding of how consciousness works is still lacking. Neuroscientists have still been unable to recognize how consciousness works, or exactly which processes in the brain are responsible for its development. Moreover, contrarily to the other three aspects which, as we have seen, are advantageous for the survival and development of an individual, there appears to be no evident reason for which consciousness would give it any advantage. (Eg: A philosophical zombie appropriately programmed and built may be just as likely to survive.) 8
So far, lacking a consistent understanding or how consciousness works, or even fully of what consciousness is, it is probably too early to attempt devising plans of how to implement it within a machine. On the other hand, however, emotion and consciousness, which as we have seen should be the driving force for critical thinking, therefore memory, could be substituted for by simply hardwiring motivation into a machine, therefore forcing it to apply critical thinking in pursuing tasks, without it needing (or being able to) produce its own motivation. A combination of old school programming, hardwiring pre-constructed instructions within a machine, a AI school programming, involving critical thinking a the previously discusses concept of memory, could be implemented to create a hybrid AI/procedural machine. 9
Bibliography Cabanac, Michel. "What is emotion?." Behavioural processes 60.2 (2002): 69-83. James, William. "II. WHAT IS AN EMOTION?." Mind 34 (1884): 188. "How Does Memory Work? The Plot Thickens." NIMH Â. N.p., n.d. Web. 10 Nov. 2012. http://www.nimh.nih.gov/about/director/2011/how-does-memory-work-the-plotthickens.shtml Crick, Francis, and Christof Koch. "Towards a neurobiological theory of consciousness." Seminars in the Neurosciences. Vol. 2. No. 263-275. 1990. Word Count: 2506 Words 10