Sharon Rose Alterman

I was interested, in reading the chapter for this week to see the author's explanation of how an agent might use predicate calculus to reason about its world. I had always thought about predicate calculus in terms of abstract reasoning, and action planning, and these were discrete entities in my mind. In the examples in the book and in class, I was able to put those two domains together in my mind, leaving me with an understanding of the purpose of the semester I spent learning about first-order logics.

I thought that the exam was fair, although it showed me that I did not understand alpha-beta pruning as well as I thought I had. I think that will make more sense when I try to implement it in the project. Overall, I felt that I had a handle on the material presented, and was not upset by my efforts.

In terms of the project, the due date of which is far to close for comfort, I am actually looking forward to implementing it. I have thought about it some, though I've not done any work at the computer. I think that it will be very important to what I want to do with the understandings I get from this course, to not only understand but actually implement this algorithm. Thus, I'm somewhat daunted and scared, but excited and happy.

Rohit Apte

Durell Bouchard

Brianne Brown

Grace Chou

Dan Crown

When I studied the mathematical logic two years ago I was very impressed that an entire course could be based on creating an abstract model of mathematics. I won't claim to fully understand the logics we studied in the course since, besides the fact that it was two years ago, it was the most difficult class I've ever taken in anything. However, one thing that I remember I did notice was the lack of a means to relate an object to another in a simple way. There were ways but these are pretty complicated and rely on a number of other variable parts of the logic we were studying. These methods also were more useful in classifying a group of elements as similar in some way. Propositional Calculus seems to share some of these shortcomings.

Predicate Calculus, however, has saved the good parts of Propositional Calculus and added relation constants. These objects give us a straight-forward, simple way to identify how two objects are related. For example, in class we spoke of the Parent relation, which could be used to specify that Bob was the parent of Alice. Other simple relations can obviously be added to the domain. This makes both grouping and ordering of the elements possible without having to use complex notation and properties of any specific logic.

Renee Findley

The only thing that struck me as unusual this week in AI was a discussion of the breakdown of a wff in prepositional calculus, involving the statement, if P then Q, and if Q then R.

And then we had to prove if P then R.

I was impressed at how much harder it was to prove it for a computer to go through the basic laws of logic, as I had been taught in sixth grade that one could apply modes ponens to both phrases,

if P, then Q.
P
Q
and if Q, then R
Q (as has been proven)
R.

The end. I saw why this was necessary, but that was what struck me as interesting.

Scott Goldstein

This week really was not all that much in the way of thought. I have found many interesting articles for the paper assignment, but acquiring them is proving more difficult than I thought. Once again the BiCollege libraries prove their weakness in my time of need.

The predicate calculus was not very difficult, especially since I have done it before. This was much simpler than the examples we had to do in the semester long course I took. I enjoy the logic problems, and I believe that I am beginning to understand how a computer could use those symbols and rules of inference to actually interpret the world it is presented, and act accordingly.

Due to the exam, there is not much more to say for this week.

Maria Hristova

Chapter 15 from the textbook makes a very important statement about AI and the understanding and implementation of predicate calculus. Nilsson says that the "big problem for AI is what to say, not how to say it."(p.248) I think that this is a crucial point in the understanding of logic and its implementations in the world of the given agent. An agent using predicate calculus as the tool to communicate and reason about the world has to conceptualize it through the world's objects, relationships between them and functions they are involved in. In a simple world this can be easy because there are only a couple of objects to be involved in the process of conceptualization thus keeping the number of operations low. In a more complicated world however, the number of objects can be large and the relationships between them can produce a complicated and exponentially growing structure which can be hard to handle especially if the agent needs to respond to a constantly changing environment. Nilsson makes an important statement that the agent's conceptualization need to be grounded. Instead of producing every possible conclusion about the world around it the agent only focuses on some thus constructing a finite perception of the world which can be used by the agent as a knowledge base for its actions or decision making process.

Agata Jose-Ivanina

Ever since we started to talk about logic, Cassie comes to my mind. Perhaps, the most striking thing at the time of the presentation was the fact that Cassie was able to answer that question about Mary's and her roommate's phone number. So I was trying to figure out what should the rules have been.

However, there are still too many questions that I am not about to answer. I tried modeling the roommate relationship:
Roommate(X, Y); Telephone(X). But I get lost at trying to encode the fact that if X has a phone number, and Y is X's roommate then they have the same number. Could it be: Roommate (X,Y)
Telephone(X) implies Telephone(Y)

Then by Modus Ponens Telephone(Y) is true if X and Y are roommates.

However, it will then take an enormous amount of memory, thinking and coding to write down the rules even for a simple situation that doesn't take us a long time to think about.

Archana Joshee

The lecture on Predicate Calculus helped me more in seeing the application of Logical Reasoning in an agent's world. Predicate Calculus seems to have more depth to it than propositional calculus since it allows us to talk more specifically about the agent's world and the things in it. I was a little confused while reading the book, but hopefully they will become more clear after the lecture. Unfortunately, I haven't been able to spend much time on the final project. I am sure I will have a lot of questions once I do.

Kip Lewis

So, there certainly wasn't a whole lot to react to this week. We had a test. We're not really supposed to talk about that, so I won't go into any details. We went over some more logic. I saw a news piece on robot pets being sold in Japan. They looked fairly interesting, and were selling for $1500. I talked to my grandfather about Artificial Intelligence. He argued that you have to start by looking at the human brain and how it works, and while I agreed that that would be a fine place to start, I brought up planes flying differently than birds, and argued that maybe there are other ways to be intelligent. We both agreed that parallel processing was important in the search for AI, however, since it seemed that intelligence couldn't just be sequential.

Creence Lin

I guess I was wrong about the "imples" thing being opposite of subset. (ie. apple is the subset of fruit, but we never say that fruit implies apple). The predicate calculus does make it so much easier for representing the three blocks in the world. In math land everything seems so clear cut, but in the real world there are messy real world problems and the ability to bridge that gap is really useful. It seems like it would be much quicker to look at a picture of the block worlds, than to go through the predicate calculus statements, but the pictures cannot easily be used for other interpretations the way wffs can.

Martin Lukac

I was thinking about the logic we were going over in class, and it struck me that this is in a sense very similar to how we interpret the world. The thing is, this logic or rather the sets of rules we come up with is _a lot_ more simpler than the set of rules that we use to interpret out world. If we take the block example, we know that the whole block world can be descrinbed with a few rules, and everything that can be done can be described with a few more rules. In our world, the rules for object lying on the floor and being stacked on each other have the same basis, but they are much more encompasing for different objects and different placments. I thought for a second that there really are no rules, beacuse we can basically configure anything any way we want to and say that everything else is just goverened by physics, but the logic rules do exists beacuse you don't need to know physics to be able to say that you can't stand an elephant on top of a drinking straw. This just shows that everything can be reduced to sets of rules.

Thinking about how these rules are put in computers, I see that its diffcult becuase we need to forget what we know (all the different possiblities and combinations of possibilities in our world), and start building a whole new world from scratch, with simple rules that allow whats usefull to be defined. The more i think about it, the cooler it seems that we can reduce our environment to logic.

Reshma Menghani

Predicate Calculus seems to be a complete reiteration of Introduction to Logic offered here at Bryn Mawr College. It happens to overlap with alot of other math classes that I have taken here too.

The book mentions at one point that formulas can be used to represent knowledge that an agent has about the world. It seems that there is a relation between the number of formulas and the world(s) that you are looking at. As the number of formulas increase you are looking more specifically at certain worlds over others. If you have enough formulas thus having enough knowldege, you have enough information for one particular world.

Reading through the rest of the text, exsistential and universal quantifiers seem almost redundant. Yet a point to be taken about this is their meanings in a given formula. Putting a universal or exsistential quantifier in a formula can drastically change the meaning of a particular formula. Using either also has a great difference.

Universal instantiation and exsistential generalization seem easy to follow. However it would be nice if we could go through these rules using object constants.

Todd Miller

Maria Pace

Heather Palmeter

As class only consisted of one period this week, we didn't go over the same volume of material as we did in other weeks. What we did cover was the introduction to predicate calculus, a topic covered by this weeks reading as well.

While doing the propositional calculus the week before I was left wondering why the logic felt incomplete to me. I definitely felt that something was missing. When we began to study predicate calculus, I realized that it was the Universal and Existential qualifier. I never realized what a powerful tool they were before, especially when you are trying to describe a world. The simple concept of "for all" and "for every" allows a much finer distinction to be made with in the logic, and consequently, the description that is being built of the world is that much more detailed and precise.

Though this is certainly an exaggeration, propositional calculus, with out these qualifiers, was like trying to do math without the number zero. There's only so far you can go before you hit a wall.

Megan Rutter

I had no idea what to write as a response for this week since we only had one lecture (and it was on predicate calculus!) and we didn't have any labs to do. So, I checked out the AAAI website for some ideas.

I found...
KONANE-Hawaiian Checkers From Peter Ingebretson. "About the Game: Konane is an old Hawaiian game, similar to many varieties of checkers. Strategy is reasonably simple, but the game is difficult to win against a talented opponent. The AI playing this game uses a simple minimax algorithm, with alpha-beta pruning to reduce the size of the search tree. As such, it is quite challenging when searching moves deeper than five or six turns in advance."

And there were many familiar names in the Hall of Fame. I found Edina's game easier to play, possibly because of the simpler graphics, but I liked the fact that this web version showed you your playing options. I hated the quotations at the bottom of the screen (only because I was playing poorly and feeling stupid about it) but the idea is really good. It makes for a good laugh. I spent a long time playing Konane, and I am horrible! Absolutely horrible! I hope my program will be nothing like me. I did get to add my name to the Hall of Fame list, but it's for level 2! How embarrassing!

Brian Simms

We didn't cover too much last week since we had the test. The predicate calculus didn't seem to be too different from the propositional calculus, so I'm not too worried. I'm pretty sure I've seen it before, so it's not too big a deal. At this point I'd like to get some work done on my Hawaiian checkers program. And I'm looking forward to Turkey Day. That's all for this week.

Matthew Spigleman

Andreas Voellmy

Nicholas Yee

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