AI Understanding and Practice

Question

I say good day to you!

So, having studied for a number of years, I finally had the option to gain the knowledge of amazing AI programming! Expecting to develop robots, games and logic to take over the world, my dreams were quickly dashed to reveal the reality of AI programming is far from what I imagined.

Though - still having major interest in AI, I ask the question:

While I have theoretical knowledge of the subject, what does one do to practice and gain understanding into the major areas of AI, be it Learning, planning, vision, etc... Furthermore, is using AI outside of the research field a feasible career path?

I would prefer real life experience answers to general opinion in this matter, but all is welcome.

Regards, Byron Cobb.

Answer 1

This is an extraordinarily skimpy answer to this question, but it may serve to whet your thoughts.

Suppose you have a Thungus. It has a considerable amount of configuration variables (think, say, 100). Each variable has at least 1000 different possibilities. E.g., [0,1000] in the integers. There is some first-order logic that goes into different combinations being considered valid/invalid.

A failed setup in production in your Thungus will instantly cause over $10,000 in damages.

You must validate, at worst, 100^1000 possibilities. To setup a configuration takes between one and ten seconds. Which configurations are valid? Which configurations will generate interesting test points?

Devise a search strategy to explore the state space that yields a workable set of configurations that can be applied to the Thungus.

note: this is a real-world scenario in a non-AI scenario.

Answer 2

This is another very skimpy answer. I would look at writing programs that use a Neural Net. I have read a little bit about this field, but I have yet to actually write one (due to lack of time with other projects). I think this would be a good project to learn some AI-programming (albeit a specific field). I'm hoping to do the same in the next few months.

Answer 3

"What does one do to practice and gain understanding into the major areas of AI"? why, read the relevant textbooks and practice writing your own programs, of course, the same thing one would do for any field really.

"using AI outside of the research field a feasible career path?"

Yes, they just don't call it AI. Some examples:

1. Google voice uses AI techniques to transcribe phone messages into text automatically (machine learning)
2. Microsoft's Kinetic uses AI techniques to correctly (we hope) identify the position of a human in 3D space in front of the XBox (filtering, learning, modeling)
3. Picasa recognizes peoples' faces in photos (image recognition)
4. Numerous video games (first-person shooters) use AI techniques so their bad guys are not total morons and can work in a group (AI planning, multiagent systems)
5. The algorithm the iphone uses to predict what word you are trying to type, that's also based on AI research.

Just some examples, I'm sure others can add more.

In short, when you are writing a program that most others think will never work, you are doing AI.

Answer 4

For robotics, check out Microsoft Robotics Developer Studio. The cost of entry is minimal and you can see if this is what you want to do. This will allow you to work with robotics software and the site provides links to hardware to see your efforts in action. If not Microsoft, many other robotics kits are available to start playing with.

Answer 5

Some of the coding for the Java Robots environment has become fairly complex and it well worth studying for AI implications.

Not only do they use active sensing and attacking, but also have advanced strategy selection based on observing enemy behaviors. On top of that, they allow teams of robots to interact and coordinate their strategy. Worth checking out.