Week1 (Due 9/6)

CS 372 (Artificial Intelligence) Fall 2004

Week1 (Due 9/6) Reactions & Thoughts

Ioana Butoi

In order to consider an artificial agent intelligent it must do something people do that we consider intelligent. What do we consider an intelligent action? What is intelligence after all? We use the word pretty often but I do not feel that I can define it exactly. I used to think that artificial intelligence dealt only with machines being able to think. Now I understand that the field encompasses a lot of other things that for whatever reason I did not consider to be part of AI.

I like the old definition of AI from Shapiro’s article: “The attempt to program computers to do what until recently only people could do.” The consequence is funny, since as Shapiro said it leaves the field of AI only with failures since successes turn into other areas of Computer Science. I was really surprised that some of the things that we use pretty often are artificial intelligent agents. I never thought of the computer that shifts the gears in a car as being AI.

I am still confused as to when we consider a robot to be intelligent. When does the robot make use of AI. Do we consider it to be intelligent if it can go down a hallway? Would we consider a robot that assembles cars as being intelligent? The two differ in the sense that one can adapt somehow to the environment while the other repeatedly does the same exact thing. Would one be more intelligent than the other? Can we make such statements given the fact that we deal with two different behaviors?

I am still reticent to using the word intelligent in some situations. I still think that in order to call something intelligent it has to amaze me in a way or the other. In the above case I would consider intelligent the robot that is able to go down a hallway. The other one, since it cannot adapt to the environment, is just performing some programmed action without any alteration. Would it be possible to compare two different artificial agents in terms of their intelligence? Would it be possible do design an IQ test for artificial agents? Would it have a purpose that we could benefit from?

It seems that from everything that has been done until today, people are seeking artificial intelligence that resembles somehow human intelligence. Would we be able to recognize intelligence in any other form that is not similar to our intelligence? Are we constrained to things that resemble us? Our imagination is constrained, since we cannot create a completely new thing that has nothing to do with something that already exists. Is our ability to determine if something is intelligent restrained in the same way as our imagination is? Some people say AI is going to change the way people think about intelligence. It might help us understand it better, but would we be able to completely surpass its barriers? There is an endless number of questions on this subject, and most of them are far from being answered in the near future. AI is a long process and who knows where it is going to take us. We can only hope it is going to help understand us better.

Andrew Cantino

The in-class discussion has been interesting. The example of the computer-graded essay was personally interesting/timely because I took the GRE recently. I had read a little about how its grading process works, but it was interesting to see another example of computer-graded writing. It seemed to me that Microsoft Word could have done almost as well. I would like to see examples of computerized graders that look at things like content, originality, and organization, all of which seem harder to analyze then formal spelling and grammar.

Here are some other things that I have found interesting so far:

• I like the breakdown of AI into the three fields of computational psychology, computational philosophy, and advanced computer science that the handout provides. This goes along with what we have discussed in class, but somehow I find this division more intuitive.

• AI (in this case, the computational philosophy branch) is the generation of systems that solve problems that humans would require intelligence to solve. This seems like a good, general definition for this kind of AI.

• AI is plagued with always seeming to fail because whenever it succeeds, the results become a new area of computer science and are no longer attributed to AI. For example, the handout says that symbolic calculus systems were originally seen as AI problems, but now that they have been solved, no one attributes their success to AI. This is interesting and helps to explain why people know less about AI's successes.

This course looks like it will be fun and interesting. I am hoping that we will get to do some real AI programming and learn a bunch of interesting and useful algorithms, as I am personally looking to expand my programming knowledge/algorithms toolbox. I have enjoyed the overview of AI from this week, as well as the philosophical discussion. My sense is that this class will be both interesting philosophically and practically, and I'm excited to see where it goes.

I posted my recent AI examples on the Recent Examples of AI page.

Audrey Flattes

It will be interesting to answer these questions in class if they can in fact be answered. After week I find myself ready to think about the possiblity of Artifical Intelligence in a different way.

* On another I wanted to show everyone the lastest in Virtual Humans remember the robot that miraculously turned into a human well take a look at where we're at now with this article. Virtual Humans Proposed As Space Travelers don't forget to take a look at Laura Laura and the other interesting articles.

Christina Florio

Now that I feel I have a better understanding of what is considered AI, I have so many more questions. While The Thinking Machine made me question the definition of thinking (I still have not come to decisive definition), the readings have made wonder what are the definitions of machine and can (Nillson, p. 2)? Why must machines “think” like humans in order for AI to be successful? If people believe that machines cannot think, at least not yet, then can the machines mentioned be considered AI? Since machines have no “fuzzy thinking” (Nilsson, p. 4), could it be possible that machines have the correct way of thinking? I am very interested in finding out how we approach AI in class, beyond readings. Maybe, my main question should be, how does an Introduction to Artificial Intelligence class define and explore artificial intelligence? That may sound a little confusing, but my point is that there are so many avenues to explore depending on how one views AI.

Kathleen Maffei

There are a few reasons why I find artificial intelligence interesting.

For one, a broad definition of the word “intelligence” leads to an alternative programming perspective. To consider the intelligence of objects – tools we generally take for granted - is not only interesting philosophically, but it’s also a matter of programming in another paradigm.

I’ve always been fascinated with psychology and philosophy, so I love the interconnection of these with computer science. In my opinion, the mind is a biological computer, the understanding of which may lead us to new ways of computing. Studying the artificial and biological computers side by side is a wonderful way to gain insight into each. While I’m fascinated with the idea of creating a being that replicates human thought, I’m even more interested in what an artificial being with intelligence can tell us about alternative ways of comprehending the universe – about other kinds of intelligence.

I’m looking forward to exploring various kinds of AI research and also in working with different approaches to intelligent design. We went over some of these things in Cognitive Science, but there’s much more to explore, and I’d like to fill in some of the gaps in my understanding of these concepts.

Sara McCullough

I think that the question of what 'intelligence' actually is underlies all of AI. That is what I am having trouble getting past. How can we study something about 'intelligence' in machines if we cannot clearly define intelligence. I also think that this question correlates directly with the question of how the human mind works. These are issues I would definitely like to discuss more.

William Moss

Karen Pang

Yuna Park

Ben Root

I tried to submit but someone else had saved a copy while I was editing, and when I went back to try to copy and paste my entry it had disappeared from the edit box!

These are a sum up of some thoughts I had about week 1:

• The integration of AI into many more diverse technologies than I knew of before: Specifically, pilotless takeoffs and landings, Highway driverless navigation, and in many human interaction venues like telephone directory services and ATMs.

• The subdivisions of the fields of AI. I feel like an important scale on which these subdivisions can be placed is that between 'practical' and 'theoretical'. Where the practical subdivisions might be mimicing human behavior through AI, and a more theoretical one would be mimicing human thought processes. Human behavior like speech recognition and image processing can be used in the real world to do things that humans used to do. The use of a computer that mimics the triumphs and foibles of the human thought process however doesn't seem like it has advantages over a real human except in the theoretical sense. One can test hypotheses about thought process rule sets with a computer instead of giving people lobotomies.

• Lastly, I was presented with this question in a Biology class that is concerned with early development of the Central Nervous System in amphibians and I thought it fit in with the more philosophical elements of AI: ""Is the human brain intelligent enough to figure out how the brain works?""

Sandeep Singh

here

Last week we discussed how AI technologies were being implemented in various applications ranging from the most mundane things like solving crossword puzzles and grading essays to logistically planning the shipment of weapons and materials during the Gulf War. When I first learned that AI software was used to grade essays such as the one administered by the College Board, I found myself doubting my own beliefs. At the time, I was a senior in high school and the pressures of standardized tests were beginning to take their toll on me. An avid programmer and someone excited about technology, I found it rather ironic that the thought of computers grading a test that could potentially determine my future left me apprehensive and hesitant. During the last lecture, we also discussed how AI could be broken down: systems that think like humans, systems that act like humans, systems that think rationally and systems that act rationally. Breaking AI into these different areas provided a better understanding about what AI actually means. The example of applied AI that I chose is one that is becoming increasingly popular in movies. No, I’m not referring to the 80’s classic, Weird Science. The Lord of the Rings trilogy made exceptional use of AI software as mentioned in this article. The most awe-inspiring scenes of this trilogy were the fight scenes. The overwhelming fight scenes with hundreds of thousands of men and Orcs going into battle left every movie-goer breathless. Needless to say, this effect could not have been achieved without the help of AI. According to the article, a software development company gave each of the combatants a little bit of AI so that they could fight independently—meaning all of their moves were ‘autonomous’ and unpredictable. The idea that each of the warriors fought without the direct instruction from a director or programmer adds to the overall reality of the fight scenes. I believe Hollywood will continue to embrace the idea of AI in their movies and I think it’s safe to say that movies of this nature will become increasingly realistic over the next few years.

Darby Thompson

Before taking this class, I had a vague idea of what had already been achieved and what researchers were hoping to achieve in the world of Computer Science (and arguably in AI), however I did not have a clear vision of what characteristics a machine, software, or merely an algorithm must have to be considered AI. I must admit that this is still an unclear area for me, however class on Thursday helped enormously.

The most useful thing that I have learned this week was to separate AI into the subcategories “Think like humans”, “Think rationally”, “Act like humans” and “Act rationally”. Now it is much easier to see what kinds of systems are under the AI umbrella. I still think that it is completely ridiculous that we use a word (“think”), that we cannot define universally, to be a in the title of two of these subsections. Surely these sections should again be broken down into much more clearly defined areas?

As for the article, I thought the ‘Machine Intelligence’ section was interesting, particularly when the author explained how self-defeating the goal, “to push outwards the frontier of what we know how to program on computers, especially in the direction of tasks that, although we don’t know how to program them, people can perform”, is.

Your name here

Immutable page (last modified 2005-05-23 02:47:07)

Or try one of these actions: LikePages, LocalSiteMap, SpellCheck