Those are the specifications for a final project in a course I'm taking: "design a piece of intelligent computer software." I was wondering if anyone had any suggestions on where to go with this? Well, first, a bit of background. This is a philosophy course titled: Minds and Machines, so something extremely complex isn't really neccessary, just that it be interesting (I chose this as my project from a list of possibilities). I have three weeks with which to complete the project and I have a slightly better than beginners understanding/experience in programming AI. My first thought was a game, something like connect four or maybe checkers, but when I mentioned to the professor that I was thinking of something along those lines, his reaction was less than enthusiastic. I haven't ruled out a game completely, just not the "standard" ones mentioned above, noughts and crosses, and the like. So, again, any suggestions? Thanks.

how about The Game Of Life(Life is simple, yet produces pretty neat results)? Here is a page that describes the rules to "Life"(this thread may be of interest to you, it's probably a bit more complex than what you want to do, but it's still interesting)

How about a really simple implementation of classical conditioning? As in, the computer has to make some decision, (like say, pick a number between 1 and 5), and you can punish or reward the AI's decision, which will affect its behavior. On the back end, this can be implemented with a really simple neural network.

Also, maybe you can add something where if you punish the AI too much, it becomes neurotic and develops an eating disorder. =D

hmm, that might be good too, but I still recommend Life, it's very simple to make, yet gives complex results look here to see a Java applet that has a life program, and a few cell setups(I recommend you look at the "Gosper glider gun" example), and, of course, you can also set your own cells in the applet

Simple chatbot. (you can write things like this in hours, not days and most definatly not weeks).

Simple statement/responce, allow it to issue multiple statements, and learn from the user (looks hardm but takes the sum total of about 10-20 lines of code....)

Simple net... (i've fogotten the name, with links and nodes. node(monkey) link(eats) node(bananas), are some things i remmber about these.

Have a rather simple rule based system, modififying the simple net (above). Something like
If something is a banana, a monkey likes something.

Add a "humanistic" typer (timer + a few ifs and rand statements, 2-5 lines at the cost of about 2 mins coding and 30secs testing), and you've got yourself a very good project.

Well done!

From,
Nice coder (and yes, i do make things like this all the time, so it may take a little longer for you. But its well worth it, and its loads of fun! [grin]. Nice_coder.Pm'abble = true)

Animal mineral vegetable (or is it 20 questions in the US) is easy to program with a binary tree yes/no question and answer system (where the computer guesses what you are thinking).
If it can't guess all it needs to do is a) ask what the answer is, b) what question would differentiate it from the leaf node reached, c) whether the answer is yes or no to the distinguishing question, and d) add it to its data tree.

Thanks for the ideas guys. I think I'm going to run life by him. I'm afriad a chatbot wouldn't be acceptable. Speaking of classical conditioning though, could anyone think of anything I could do with genetic algorithms aside from the standard ants simulation? We did discuss evolution quite a bit and I think that might go over well.

Another thing you may want to try is the 8 puzzle game using A* pathfinding and such. Lots of information online and it s an inteligent program however the teacher may get those alot. So just an idea.

Maybe a data classification program?

What exactly are you looking for?

From,
Nice coder

Quote: Original post by wild_pointer
Speaking of classical conditioning though, could anyone think of anything I could do with genetic algorithms aside from the standard ants simulation? We did discuss evolution quite a bit and I think that might go over well.

Implement a Holland style Classifier System (CFS). Very easy to implement (about 200 lines of code tops) and implements unsupervised learning through trial and error plus evolution.

Timkin