Sharon Rose Alterman

I think that the theme that unites the readings and ideas from this first week of artificial intelligence is fear and change, both the fear of what we do not know and the fear of what we do and how little these two fears have changed. Both laymen and professionals in the field are facing this obstacle. Professionals, because they know what machines can and cannot do, are afraid of what we may or may not be able to accomplish. Will we really build the robot of science fiction fame? Will we build a machine on which we will not be able to pull the plug? Conversely, is it possible that we will never be able to build a machine that can be considered intelligent? Is this a good thing or a bad thing?

This dichotomy of views was portrayed very well in the film. Neither the host nor the expert really knew where artificial intelligence was going, or what we would be able to accomplish in the next thirty to forty years. The expert had more technical knowledge of what was going on with the examples of machine intelligence, but that did not give him a crystal ball with which he could predict the future of artificial intelligence, and what it could do to or for humans in the future. The host did not have the technical knowledge of the workings of the machines, and yet, he really did not have any better or worse idea as to what the future of AI might hold.

That film was made in 1961. In 2000 we still do not know where the future of AI could lead us. Although more people interact with computers on a regular basis than was the case in 1961, the average end user knows little more about the workings of 3the magic box2 as the host did in 1961. Most people still do not know what artificial intelligence is capable of doing. Professionals in the many fields that overlap and coexist with artificial intelligence still do not know what machines may be capable of in the next several decades.

This leads me to wonder if all that much has really changed in the last several decades? Have the technological advances we have made brought us any closer to the answer to artificial intelligence. To reference The Hitchhiker1s Guide to the Galaxy by Douglas Adams, do we even know what the question is? When we find it, will it invalidate the question? That, I find to be a cause for fear.

Rohit Apte

The movie highlighted the optimism shared by many early AI researchers ­ that we could have "thinking machines" in as little as 10 years. While AI research has made incredible amounts of progress, most of this has been in understanding certain functions of the brain or Machine Intelligence tasks that today are not considered to be AI relate anymore. We have computer that regulate traffic, programs that can handle a lot of mathematical processing, robots that can work an assemble line, and even optical "eyes" ­ cameras that can recognize basic shapes and colors. But we don't have machines that can "think". I agree with the MIT professor that until the computer produces something new, I will not accept it as being intelligent.

I think the main problem AI faces today is that computers are not really capable of independent intelligent thought. I have seen a small demo of Eliza (the computer that poses as a physiatrist) and it doesn't take long before one realizes that you are not talking to a real person. Moreover, I think we are as far from achieving "thinking machines" as we were when the movie was made ­ a startling fact, considering the technological progress we have made in the last 50 years.

Durell Bouchard

Pre-first week of Intro. to Artificial Intelligence my conception of artificial intelligence was that it was divided into two areas. The first area is most obviously attributed to all of the science fiction that I've been exposed to over the years. The common fantasy is that in the future there will exist machines with artificial intelligence that simulate human intelligence. My second notion of AI comes from what I see in video games, where the computer simulates intelligence with respect to some small predefined world. However, now it's so less clear what AI is. The first weeks readings all seem to start off similarly with a few simple sentences defining what AI is, which for the most part just restatement of what is obvious from the name artificial intelligence. Also most of the readings also go as far as to point out that it is difficult to really define AI, because it is difficult to define intelligence. What throws me through a loop is that AI is used to assist in the study of intelligence. So it is the study of something that may not be defined, and simultaneously it is also a physical creation of it.

Brianne Brown

The part of this week's readings that I found most interesting was the notion that Artificial Intelligence has both engineering and scientific goals. (Nilsson, p.2.) I guess what we're primarily going to be studying are the engineering aspects of AI, but I have to admit that the scientific goals, the notion that we can use AI to understand how intelligence functions in humans and animals is what really interests me.

Along that line, I was interested by the claim Nilsson presents on pg. 4 that "Conventional computing machinery is based on true-or-false (binary) logic. Truly intelligent systems will have to use some sort of "fuzzy" logic." This implies that the best model will be modeled on how humans think, and it assumes that humans use "fuzzy" logic. Is this, in fact, an undisputed claim? Is it possible that humans, like computers, use a sequence of binary logics in making decisions, or not? I don't know anything about this but I'm curious.

Another claim on pg.4 is that "We'll have to build new varieties of parallel computers to make progress in AI." I found this idea of mimicking the way that the brain computes interesting, but on the "Basic Questions" site at Stanford
the author claims that parallelism "presents no advantages." Neither of these authors elaborates on their claims; I'd be interested to hear what other people in the class know about parallel processing and its potential utility.

Also, on the same Stanford website, the author addresses the fact that current AI technology can create a very effective Chess machine but can't create a very effective machine to play GO. The author claims that this is because "a position in GO has to be divided mentally into a collection of subpositions which are first analyzed separately followed by an analysis of their interaction." I don't completely understand this explanation, and I wanted to bring up another point: I've heard it said that AI can't create a very good GO player because, unlike Chess, the point of GO isn't to win but to reach an acceptable level of compromise with the other player. Obviously, "acceptable compromise" is much harder to model than simply winning or losing. (Maybe because it requires too much world knowledge?)

Dan Crown

Prior to the beginning of this class, I knew very little about Artificial Intelligence. My experience with it was (so I thought) limited to infrequent computer gaming and reading a few papers on the subject. After reading the first week's materials and watching "The Thinking Machine" in class, my thoughts have already been drastically altered.

My earlier thoughts about AI were primarily limited to the subject of futuristic robots interacting flawlessly with humans and, in effect, "becoming" humans. "The Thinking Machine" helped me to see the simpler side of AI. What we are currently developing in labs is far from the futuristic scenes I had in mind before. We are creating machines to learn the alphabet or to answer questions limited to a very small area of expertise. We are creating children, animals, and insects, not robots with plans of world domination.

Nilsson's introductory Chapter 1 has also helped me to refine my opinions about AI. While it is, as I have said, simply an introduction, it gave me a view of AI from the inside. It relates some of the development stages of AI and some of what we are going through now. For example, Nilsson focuses on the difference between signals and symbols instead of telling us about a robot that can serve drinks and carry on a conversation.

The first week of class has brought my thinking process to a lower level and inside AI. My original thoughts, however, remain in the background as the goals and dreams of the field.

Renee Findley

Artificial Intelligence is a small and greatly unexplored field within computer science that is really only limited by imagination and pocketbooks. But in all of the readings, videos, and texts, two phrases have stood out beyond all others.

The first is the fundamental question: can machines think? However, this is not as easy a question as it first appears, when everything depends on the definitions of machine, think, and can. First of all, what is the definition of a machine? Is the machine supposed to be a robot, or a computer? In which case, can a machine designed for a very specific purpose, like Deep Blue, really think? Is thinking responding to a single task? Or many tasks? Is thinking a form of interaction? If it is a form of interaction, as the movie from class suggests, than any computer that can play a game can think. But if 'can' means 'able to independently' then no machine can think, because first it must be built and programmed.

In which case, some might say that in all of its years of existence, AI has failed to meet any goals. Arguably, artificial has been created, but not intelligence.

At this point, the second of the two phrases comes to mind, from Frederick Hayes-Roth's article: "Artificial Intelligence: What Works and What Doesn't," which said that creating AI machines, from the birth of the idea to the actual control becomes "exponentially difficult". The more we discover about the limits of our present forms of Artificial Intelligence, the more complex and demanding our definitions of 'can', 'think' and 'machine' become, creating problems with difficulty to exponential difficulty. Has artificial intelligence failed? No, on the contrary, nearly all the original definitions of 'can', 'think' and 'machine' have been realized, but our definitions have changed along with our technology.

Thus, failure is as subjective a term as any of the others. AI's goals and purpose are still liquid in the mold. AI appears to be a field which is in a state of constant flux, more about breaking barriers beyond present conceptions than at ever arriving at some determinable goal. Failure, short of stopping, is impossible, and success is always one more barrier away.

Scott Goldstein

Artificial intelligence appears to be a far more complex problem and discipline than I thought coming into this class. The various methodologies alone are enough to confuse me for years to come. The issue that seems to be the most interesting to me is one that is brought up at the end of chapter one in the textbook, and was alluded to in the film: the repercussions on humans if there are intelligent machines.

The Matrix presents a post-apocalyptic view of a world that developed Artificial Intelligence, and in fact many of the science-fiction horror stories in popular culture focus on AI run amok. I share this fear to some extent. Computers can do much to improve the quality of human life, yet what should happen if they ever gained the ability to truly mimic human intelligence? Could machines of equal intelligence, and greater physical ability to us not replace us?

The attitude of the film's host was of skepticism, but it was tainted with fear. He was worried about the possibility of being replaced by a machine, just as intelligent and creative as he, but without the human weaknesses of a man. Fortunately, this problem will not be a practical reality for quite a while. It was made quite clear by every source to which we were exposed that the technology and knowledge of thought processes available today is far behind what would be needed for true artificial intelligence to develop. It is, however, a fascinating concept, and one that should continue to be explored until it does actually pose a real and tangible threat to human society and the social hierarchy.

Maria Hristova

There are many things that I feel can be said about the field of Artificial Intelligence and I think that we got a chance to see how diverse the field is through our discussion in class and through the responses that everybody had. I had the feeling that everybody in the class had a different reason to be taking a course in AI and I find this very appealing because it means that not only the field itself is diverse but everybody from our class will have different points of view to present when we have a discussion. I think that the mixture of CS and non-CS students will make the class more interesting and challenging for all of us. I found the movie that we watched during our second class very interesting too because I knew that it was filmed only 40 or so years ago but technology has gone a long way since then. It seems so weird to think that in the world that our parents were growing up in computers were not an everyday / every-hour reality. I am sure and that very few people back then could imagine having a computer in their dorm room that they could access an incredible amount of information through the web. I also think that a very important point that was maid in the movie as well as in class is that the field of AI evolves constantly and subjects that might have been considered a part of AI 30 years ago have now become trivial or evolved into field of their own. Another very important point that the reading makes is that AI has an influence on many other science fields as well as being influence by them as well. I think that this is very important for the stage at which science is today. I think that after being specialized into separate fields for a long time science is becoming a uniform field again where all the different branches of it are becoming closer and closer again and I think that AI is a very good example of this process that has been going on. The reading was very helpful because it defines some very important terms that give explanations to what the basics of the field are. Terms like integrated systems, autonomous agents and problem solving seemed to some up in the textbook often so I found it very useful having them defined separately from each other.

Agata Jose-Ivanina

Today computers learn, they teach us about ourselves, they perform so many tasks better than we do. Unsurprisingly, people start wondering whether we would ever reach the point where machines will replace people. Can human mind create something more intelligent than itself? Looking at some examples of technical advances one could say: "Of course! Look at what the calculator does!" There are very few aberrant individuals who can compete with this tiny instrument that ameliorates our lives at times (although at times it makes us lazy and then the level of our skills actually decreases). But is that intelligence? What should we be afraid of? Of a machine that would be able to learn everything and will thus become a solution to any AI-complete problem? Will this machine and others like it ever attempt to replace people? Should MIT professors be scared of finding their daughters married to robots?

Perhaps, the most scary thing is that we don't understand ourselves and we use the machines to shed some light on the arcane work of our minds. Can a machine learn to develop an independent thought process that will go beyond our understanding? Then instead of assuaging the struggle to comprehend the secrets of Nature we will create new ones that might turn out to be hazardous. So the most important thing is to be able to control the development of the machine intelligence. But then it is impossible to stop that one curious being who will not confirm, who will start thinking "outside the box" and who will cross the forbidden line.

However, will the machines ever be the same as people? Perhaps, a DeepBlue defeated Kasparov (but the team
worked on that computer for several years). I can imagine one that could compete against Tiger Woods - it could probably make more accurate calculations. But will a machine be able to withstand Venus Williams' speedy and flexible attacks? Will it be able to "loosen up" enough? Can machines be as adaptable as humans? Well, perhaps, they will be able to compromise with each other better than we since they won't have pride and vanity - the cause of all our misfortunes. On the other hand, they might develop these qualities and thus prove that they are inseparable from higher levels of intelligence. Then we will have the advantage of all the experience of dealing with the proud and vain individuals that we have been accumulating for thousands of years now.

Also, why are we scared of the machines? Perhaps, we can start teaching them the rules of "good" behavior. And then maybe they would make this world better? Perhaps, they could calculate the solutions to all the economic problems in the world and all countries will become members of the first echelon? It looks like we are setting the rules, they will start where we want them to, so it is also our responsibility to make sure they are harmless.

Archana Joshee

In the movie "The Thinking Machine", when the computer came up with different plots for plays, it produced the outputs based on a set of input patterns that specified what was reasonable and what was not. When those rules were not specified, we were given an example of a plot, which seemed ridiculous to our way of thinking. But if the rules are always specified, will the computer ever be able to come up with a creative plot? At the same time, if the computer is left entirely on its own, will those stories ever be of any interest to us? I guess what comes up here is the necessity for the ability to reason things out. What makes stories interesting is the author's use of creativity. The story might have plots and ideas in it that had never been thought of before and that is exactly what captures the readers' attention. The author is capable to come up with such ideas due to the environment he is in, his interactions with the environment, his feelings, emotions etc. So I guess my question is, is a machine capable of experiencing such feelings, emotions and interactions? And without that will a machine ever be creative?

In one of the readings in the web I came across this statement. "Machine learning is said to occur in a program that can modify some aspect of itself, often referred to as its state, so that on a subsequent execution with the same input, a different (hopefully better) output is produced." What I find interesting in this statement is that, if a machine comes up with something new, will it always be intelligent? When humans come up with new ideas and inventions, some are intelligent, while others are not. What about a machine, will make it capable of coming up with intelligent ideas?

Also in the reading, I was very surprised by the number of areas which are considered AI-complete Tasks. I am still confused about exactly how a natural language system defines an intelligence system.

Kip Lewis

Before hearing of this class, I thought the study of AI would fall almost entirely within the realm of computer science. Seeing the class cross-referenced in philosophy suggested that it might be somewhat broader than I had originally thought. Not until this first week of reading did I understand, however, the diversity of fields required in tackling artificial intelligence. Besides the obvious computer scientists and engineers needed for making the programs and the agents that exhibit AI or aspects thereof, philosophers are also needed to tackle the questions of "What is intelligence?" or "What is a machine?". Psychologists are needed to figure out the human thought process, and biologists must explain just what is going on inside the human head on a molecular level when we think. Essentially, any study that attempts to figure out how humans, and especially human brains, work can march under the banner of AI.

One of AI's many debatable questions is "How do we know when an agent is intelligent?". Nilsson asks, "Should behavior be judged as intelligent independently of how it is computed so long as it is the "right" behavior?". My answer would be "no." In my opinion, if we only look at the behavior of an agent, then we are only seeing if it appears to be intelligent. Along that vein, it surprised me to hear that some people consider artifacts such as thermostats to be part of the field of AI. Simply making an autonomous decision does not merit any labels of intelligence. As I said before, displaying the right behavior, in this case keeping the room at the right temperature, is in no way being what I consider "intelligent."

I don't pretend to know what goes on inside our head when we think, perceive, and recognize, and I'm not even sure if our method of thinking is the only one that should be termed "thinking" or the only one that can be called intelligent. I do know however, that if someone could convince me that a machine could think in the same way that we do, I would cease to call myself an agnostic and would henceforth be an atheist. Any belief in God on my part hinges upon the existence of a divine spark in all people - a divine spark that enables us to think and feel and have emotions in the way that we do. If our thought process could be replicated by silicon and logic gates or some sort of neuron simulator, then we obviously wouldn't require any sort of divine spark. Likewise, if someone could prove to me that no machine could possibly replicate the way we think, then it would suggest to me that humans must contain a divine spark to think and be self-aware, and I would therefore believe in God.

Martin Lukac

The thing that got me thinking the most from the first two classes was when Prof. Kumar mentioned the thermostat. I had always thought of AI as what was shown to me on television and in movies: either a large computer that seems to do more than just run algorithms that help it interact with people or a humanish looking piece of machinery that was like a person. But, when Prof. Kumar mentioned the thermostat, I realized a very obvious thing that I had never thought of before: that machines can have some broken down form of artificial intelligence. In this case, the thermostat acts like a human in that when it gets too cold, it turns on the heat, and when it gets to hot, it turns on the AC. This is a basic function that comes to humans: when its hot we turn on the AC, and when its cold, we turn on the heat. What has kept me thinking about this example is a couple questions i had: can what the thermostat is doing be considered thinking
(reasoning)? obviously its just a reaction, but if there is no thinking involved, can it really be artificial intelligence? what about humans? are we actually thinking about the temperature, or is it as simple reaction just like the thermostat had?

When I listen to what people purpose AI will be like in the future, i think that there has to be something more that a program/algorithm that causes the machinery to actually learn and react/respond randomly in a human way. Its as if there is something that has been missing from the tremendous computational speed/power and amazing algorithms that are being predicted in future AI. I suppose i think this because to me the difference between thinking and reactions isn't really clear. There is something more in what humans do that make it more than simple reaction or recalling learned situations. There seems to exists some sort of untraceability in what humans do that i can't conceive of as being turned into an algorithm.

Reshma Menghani

After completing the readings required for this class, ironically my understanding of this field seems more obscure now than it did before. I have to admit that this field is "a lot more than I imagined". I've always compared AI with robotics, but I guess the media is to blame for that. Instead it's one of the many subfields in the Artifical Intelligence field.

One of the many bewildering questions that still stays on my mind at this point is what the consequences of such a field may be. The movie I feel didn't help much. Or at least it seemed to give a very positive outlook in such a field. I guess considering when the movie was made perhaps such a question was not asked as much as it is now.

The way I see it I think most people would agree that such a field especially in the future would have both positive and negative impacts on the future. I think the real question is how far will one outweigh the other. "the third (the period since 1945) is the period of deepening philosophical confusion and the emerging concept of the "artificial". (Ethics of Virtual Reality, Bearden). It seems as if such a question was asked and predicted a long time back in our history.

I think such a question is given more importance day by day as wetechnologically advance in such a field. We watch kismet in its social interactions with other people. How do we know that such a cute little robot could indeed be highly disadvantageous to our society? I guess we'll see what happens......

Todd Miller

/chapter one of book/

I'm interested by the author's dismissal of the term "machine intelligence;" when I learned it, the contention was that AI applied to thinking systems with less than human intelligence, and MI appplied to human or higher intelligence -- machines to which one could plausibly grant citizenship...

/Thursday movie/

I'm amused that the big demonstration of learning was handwriting recognition, because it /still/ doesn't work all that well. (Noting that the Palm, IIRC, doesn't recognize the Latin alphabet, but 'graffiti.') It's also /very/ telling that the difference between experience and programmed knowledge was glossed over, because that's the key sticking
point in most AI projects -- a simple definition of intelligent behavior is teaching! (To reverse the notion of learning as a definition of intelligence.) Software that can write software from a natural language description in the general case is almost certainly AI-complete. It was also somewhat amusing that /men/ program and are programmed. I didn't see the rubber/steel difference in the demo; maybe I didn't have enough time?

I also contend that intelligence is NOT rule following -- Godel's incompleteness theorem necessitates the ability to break rules. Incidentally, 'great men' like Newton and Galileo probably /didn't/ think like the rest of us. It's also kind of sad that the SAINT problem (symbolic calculus) is no longer an AI problem.

Maria Pace

One thing that struck me about the movie "The Thinking Machine" was that there was not one real thinking machine in the entire film. It amazes me how people back then considered the ability to play chess, solve logic problems, and put together an "original" screenplay from a list of choices and rules provided by a human programmer a sign of human intelligence. True, until these computers were built, only humans were capable of doing these things. However, to me a truly intelligent entity, be it human or machine, must be capable not only of carrying out instructions and following orders, but of knowing when and how to break these rules-- being able to reason outside of the given situation.

I wonder if scientist of the 1950's would marvel at the use of artifical intelligence in modern machines that we take for granted? My parents own a small appliance called a "Smart Toaster". It is so smart that when I insert two identical slices of bread into its dual smart-slots and push down on the smart lever, it either doesn't toast the bread at all, or burns it to a
crisp, just like Einstein could have done. Often, one slice will be charred, while the other is still raw. Maybe the problem is that the bread is un-intelligent, since it is neither man nor machine. At any rate, until I see more substantial evidence, I cannot be convinced that any mechanical device is capable of intelligent thought.

Heather Palmeter

Watching the movie, it was quite easy to tell that it was made in an era quite different than our own. Aside from the total lack of personalities and chain smoking, one of the most noticeable differences (and most relevant to the class) was how computers were regarded. Today you will find people who like computers, people who don't like computers, people who understand computers, and people who don't have a clue. You can even find people that seem to function better with computers than with people. What you aren't likely to find is a person who is afraid of computers. I'm not talking about the person who doesn't want to use the computer for fear of breaking it. I'm talking about the person that fears for the fate of humanity and the world because computers are on the rise. However, it is the viewpoint that seems to be indicated by the movie and it is certainly a mindset which sets the present apart from the past.

I think that this can be attributed to the lack of general knowledge that the average person possessed about the computer. Any doubts about this were quickly shredded in light of some of the interviewer's questions. Most of them could be answered today by almost anyone you meet on the street. This knowledge seems to lend the population a feeling of control over computers which allows computers to take on the persona of a tool rather than a malevolent force lying in wait to take over the world and rule mankind. For the most part, the movie left me with a feeling of how far we've come, if not in our knowledge of computers (though I would be hard pressed to argue that the understanding of computers has not grown considerably in the past years) then certainly in our ability to allow them into our society as a valuable tool (and source of great entertainment).

Of course, having lived in the computer era my entire life, it is difficult for me to imagine a time when computers were though of as anything other that a mainstay in both one's business and personal life. However, until recently, it never occurred to me to think of computers in broader terms. They were programmable tools that could either make your life easier or drive you to an early grave, depending on your point of view. Now, though, there is this drive towards artificial intelligence that I am now apart of, if only in the confines of this semester. Suddenly computers are much more. They have become a driving means of dissecting human intelligence in ways never tried before. Instead of simply studying human intelligence we have put ourselves in the unique position of trying to recreate it and, hopefully, by doing so garner an greater and more detailed understanding of human intelligence.

First they were feared. Perhaps it was because they were new or perhaps it was because, even in the beginning, it was plain to see what computers would someday be able to do. Then they were embraced, if not by all, than certainly by many, as tools that could help to change life, hopefully for the better. Now our perception is changing once again. Now they are tools to allow us to see ourselves in greater and clearer detail, something that I never would have envisioned of them at the very beginning. Though, people have feared the rule of men for centuries and the fact that the same was feared of computers since almost the beginning suggests that maybe the image of the computer as a window into the human being was not quite as unexpected as I would like to think.

Megan Rutter

The first week of class has given me a lot to think about. It has also
raised mnay questions about machine thinking and learning.

Can Machines Think?

The movie, textbook, and handout all address a question that came to my mind during our first AI lecture. Can machines think? This question has been baffling me for several days now, and I haven't been able to come to a conclusion yet, but I have learned some interesting points from the different perspectives given via the movie, textbook, and handout. I especially like how the textbook covers the material by looking at the words individually to define what they mean.

Consider the word "can." One's definition of the word "can" can change the entire meaning of the question. This way of looking at the question changed my perspective on it. Does it mean can existing machines think now? Or are thinking machines possible someday, but we don't have the technology/time/money to make them now? There are so many ways to look at the question, making the idea of machine-thinking much more plausible, though I'm still not convinced.

A couple of other questions came to mind throughout reading the assignments. Can machines really learn? The handout stated, "Why build an intelligent system when we could just build a learning system and send it to school?" Is this really possible? Do we know enough about human learning processes to program computers to do the same? I ask the same questions that were asked in the 50s and 60s. Can machines really produce novel things? I don't think that the example with the westerns proves the theory that machines can produce new things and be creative. While a particular play may never have been produced before, each indiviual part was pre-coded into the computer. However, the flipside to that argument, as pointed out in the movie, is that humans are precoded too. The idea that instinct is just another form of programming was brought up and discussed. Just as the computer had to follow the rules programmed into it, a human playwright must follow rules that make a play make sense. Are our plays a result of precoding as well? How creative are human plays in comparison to the computer's western?

Another issue that the reading made me ponder was the idea of computer vision. Exactly how does a machine see? In humans, the brain interprets what the eyes see. How does this work for machines? What can they use to interpret visual input? And related to the issue of sight, if computers can feel, see, and hear, can they taste and smell as well? This is a very intriguing question that I hope we discuss at some point during the semester.

Overall, the reading didn't really answer any questions that I had after class lecture and watching the movie. It did give me some additional perspectives on the topics at hand, but it also gave me more to think about, many more issues to question. For now, I'm still stuck with the question: Can machines think?

Brian Simms

The reading we did for this past week seemed mostly historical stuff and definitions. Stuff that needs to be read but isn't particularly interesting. Until I got to the AI-complete stuff I was fairly bored. I never realized that there were classifications of AI-completeness. I'm more interested in learning about neural nets, genetic algorithms, and fuzzy logic. Those areas seem to lead to more mimicry of human intelligence. Using normal algorithmic programming doesn't seem to help progress into AI. Autonomous Agents seem to be the closest thing to the vision of AI we see in the movies and look for in science fiction stories. What will separate humans from computers once we get to the point where we can create autonomous agents? And what will stop autonomous agents from building themselves? Will there be some kind of survival "instinct," as there is in humans? Will the agent be able to reprogram itself, or just react to stimuli and not "remember" previous encounters? What kind of techniques are available today for this kind of programming? These are just a few of the questions that pop up immediately...

Matthew Spigleman

My strongest response to the movie was the result of opposing comments made by two of the scientists interviewed. The first was that "Intelligent behavior is rule obeying behavior." and the second dealt with the quandary that the great minds of history have come to their brilliant ideas only by throwing away the laws the were brought up with (such as Darwin, Newton and Galileo). The first of these opposing views seems to be a rather narrow view of intelligence. The example talked about in class of the thermostat is quite telling in this regard. I do not think that the thermostat, which shuts off the heat at an exact temperature and turns it back on when the heat drops back to below the set temperature, is an example of AI.

However the second idea, that great minds can "think outside of the box" is also seems limited as there are have been many great advances and much had been done with traditional thinking and the expansion of previously held views. I came away from the movie with the feeling that the scientists of 40 years ago did not yet have a sense of when intelligence begins and memorization ends.

The distributed article on AI cleared up many of these questions for me with its discussion of the so called "AI-complete" projects. These had distinct goals, yet each goal is so critical to the very nature of human intelligence and development that its mastery by a machine would have uplifted that machine to a level of consciousness never before seen in computers. What I did not understand in the article is the apparent definition of AI as "human-level intelligent behavior." While this is presented as the historical definition it also seems to find its way into modern definitions as well. For instance I would think that AI is one of the best tools available to understanding the clicks and chirps of dolphin communication and that in some ways that is a higher goal than human language. I would be curious to know how much work is being done on the understanding of lower or differently evolved life forms.

Andreas Voellmy

The approach of machine intelligence appealed to me much more than that of computational psychology or computational philosophy. Here is my reason:

I don't think it is possible for artificial intelligence, or computer intelligence to ever equal human intelligence. Human intelligence draws on such a wealth of experiences and emotions that are peculiar to humanity. Our perception of the world and our animalistic responses to this world inform our intellectuality to some degree. A computer will never have the same primal understanding of the world that humans do.

Also, I believe that the human mind has levels of consciousness that transcend intellectuality. Perhaps there are more "spiritual-" or intuitive-oriented levels of mind that do not operate on the principles of logical reasoning alone. These levels of consciousness may also affect the way humans interact with their world, especially in the extreme creative cases (such as Newton, as referred to in "The Thinking Machines").

I think it would be silly to try to construct a set of rules or a set of circuits to mimic this kind of human mind. It seems that what computer intelligence may be, is different than what human intelligence is. Perhaps the two only overlap in certain areas, such as the area of rational, logical reasoning and certain types of problem solving. I don't mean to detract from what computer intelligence is or may become, in fact I mean to say that it will naturally be something different from human intelligence yet equally as exciting. I think the important, most pertinent approach would be to abstract the idea of consciousness from the human or animal consciousness and to attempt to implement this in machines. Then computer intelligence should be applied to solve the kind of questions that are natural to itself. This is why machine intelligence appeals to me most.

About the movieIt's interesting that scientists and philosophers of the 1960's recognized the profound ramifications of the invention of the computer. Today, we seem much less preoccupied with the philosophical/psychological meaning of computer technology, though its use is much more pervasive. It might be that thinkers of the 60's were unduly frightened by this new technology, but I think part of this change in attitude comes as a desensitization to computer technology. We think of computers as aids in accomplishing specific tasks, such as paper-writing, searching, and cataloguing. However, we don't view computers as entering the psychic space in general. I think this is a misperception and I think that computers are changing the psychic landscape in a remarkable way. As more and more people are born into the computer world, the sense of the profound historical change computers effected is lost.

Ruthie Worrell

In the book and in class, two approaches to AI have been delineated: symbolic and subsymbolic. Upon reflecting on the differences between these two approaches, I'd like to see if one comes out as making more sense to me than the other.

The symbolic approach involves first telling the computer the knowledge it needs, then representing the knowledge as symbols and explaining to the computer how the symbols interact with each other. It would seem that once this method is used to create an intelligent machine, much of the problem would be solved. This top-down approach starts at the top level, the complex phenomenon known as "knowledge," and works down from there, ending at the lower level where symbol processing actually takes place.

To me, this method does not seem like a very good way, because it doesn't make it possible to take "small steps." It seems, rather, that it is an "all or nothing" phenomonenon.

The other approach, the subsymbolic one, is in some ways the reverse of the symbolic approach. In this "bottom-up" approach, the beginning is at the lowest level, and as work progresses it moves upwards. The philosophy behind this approach is that it is necessary to lay a solid foundation before building any complex structures. Nilsson explained how supporters of this view point to the evolution of human intelligence as a long process. Starting simple and then moving on to more complex and challenging problems makes more sense, so that if one step fails, we are only set back that one step, rather than all the way back tothe beginning.

Nicholas Yee

At one point in the movie, the issue of preprogramming comes up. The question is whether preprogramming can ever lead to novel ideas or behaviors. Psychology has come a long way in understanding "preprogrammed" behaviors in human minds. Beyond the instincts and optical illusions they have know for several decades, psychologists have also learned about preprogrammed cognitive modules that can generate new information after assessing surrounding information. The language acquisition ability would be one of these. Languages obey underlying rules that transcend cultures. Linguists have set themselves the task of finding this Universal Grammar, which is not a rigid blueprint of syntax and semantics, but a list of "If then" clauses. For example, if a language is rich in verbal agreements (with gender/plurality etc.), then word order is unimportant. But if a language is poor in verbal agreements, then word order is strict. Thus, this would explain why word ordering is more free in romance languages, than in languages such as Chinese, where there is no agreement between verbs, gender and plurality. The search for Universal Grammar sheds light on the fact that programs can generate novel behaviors if constructed as flexible learning mechanisms, the way infants follow rules to learn and then utter a language. Infants are predisposed to picking up certain combinations of sounds, rather than others. This is why infants in urban cities don't learn to talk like motor engines. One might wonder what would happen if infants were taught a language that did not follow Universal Grammar. In fact, this has happened historically with Creole languages and sign languages, where the children grammaticized the crudely formed language of their parents. And thus new languages sprung up.

Back to CS 372 Course Materials | Back to Written Work Main Page