I need a design idea for an artificial intelligence class. Do you have one

Question

I wanted to ask users of Qaru for a good idea for a project that could entertain another student programmer for a semester. Computer vision may seem interesting, although I could not say whether a project in this area can achieve something that can be achieved in 4 months. What do you think?

Answer 1

There is a story that in the early days of AI research, when significant progress was made in "hard" logical problems with the help of mechanical theoretical theorists, the professor appointed one of his graduate students as an "easy" problem to solve the problem of how vision, subject to significant contribution to the brain. Obviously, everything turned out to be much more complicated than the professor expected. So no, not vision in the general sense.

If you are just starting out with AI, there are several directions. The classic AI problems - logical puzzles - are solved using a mechanical theorem (usually written in Lisp - see here for classic text on solving logical puzzles). If you do not want to create your own, you can get a copy of Prolog (they are essentially the same thing).

You can also go with pattern recognition problems , although you want the initial problems to be fairly simple so as not to worry about the details. My dissertation included the use of stochastic processes for recognizing letters in a free-floating space, so I partially relate to this approach (don't start with stochastic processes, though, if you really don't like math). Right next door is a subfield of neural networks . This is popular because you can hardly learn NN without creating interesting projects. Across the whole domain (template processing), it’s great that you can solve real problems, not toy puzzles.

Many people love Natural Language Processing , as it is easy to start, but almost infinitely difficult. One of the defining problems is the creation of an NLP program for processing a language in a specific domain (for example, discussing a chess game). This makes it easy to see progress, albeit difficult enough to complete a semester.

Hope you get some ideas!

Answer 2

Cross-linking software, which most people implement as one of their first ANN applications, is character recognition (not necessarily hand-written characters).

Answer 3

You can do something simpler than general "vision", for example, point a webcam on a digital thermometer and read the temperature.

Answer 4

I can’t say without knowing more about you, your friend and the project. I think not".

I would point you to two sources. The first is Peter Norvig "Artificial Intelligence" ; the second is "Collective Intelligence Programming . " Perhaps they will inspire you.

Answer 5

One thing I've always wanted to do is improve the AI ​​of older video games. Take DOOM, for example. A very simplified AI, compared to what you see today. You can try to improve or even completely rewrite the AI ​​for enemy characters in a video game.

Or you could write your own little game that specifically focuses on the computer, making unexpectedly informed decisions.

Answer 6

Write a spam filter. Choose an interesting building for training and filtering (Twitter posts, Facebook wall posts, blog comments ...?). There are many ways to create a classifier, define training data, etc.

Answer 7

I always thought it would be interesting to write something that would look at the message (say, the question here) and predict how many votes it would receive. I initially thought about this idea in the context of looking at a blog or article and predicting bonuses / recessions on reddit (or, more simply, on forging, such as low, medium, high).

Answer 8

A program that plays poker, hearts or the like.

Answer 9

Make it interesting, for example, in an AI game, I know that a chess simulator can take some time, but maybe you can cook it a little (only 3 pieces, 2 rooks and a king on each side) ...

Answer 10

Turing tests are interesting. Here is a link to how LOLBot passed the turing test