Wednesday, October 12, 2011

Intelligence Is All About A Partially Observable, Stochastic, Continuous, Adversarial Space...Really!

At least that is what I learned in lesson 1 of Stanford's free, online, Introduction to Artificial Intelligence Course. 

Along with 160,000 of my classmates from all over the world, I am trying to learn the basics of artificial intelligence (AI).  Besides a long-time personal interest in AI (for my capstone project in law school, I designed an expert system that helped college students navigate landlord-tenant law (While it may seem trivial today, believe me, back in 1983, this was considered enormously sexy...)), at least half of AI is about how to better understand uncertain environments, something with which intelligence professionals are intimately familiar.  In fact, in lesson 1, as one of our professors, Sebastian Thrun, points out, AI can be thought of as "uncertainty management" -- words that should also resonate with most intelligence analysts.

My hope is that some of the formal ways in which AI scientists go about doing their business might have direct application to what we often tend to think of as the very squishy world of intelligence.  Since many AI applications are already concerned with what are traditionally intelligence problems, my assumption is that the language and systems used by AI professionals will help me understand and explain my own work better.

So far (I am well into Unit 2 of the course), I have not been disappointed.  While the production values are more Khan Academy than Nova, I find the short video clips, frequent quizzes and my own interest in the material to be enough to keep my attention.  The concepts that underlie AI are so embedded into almost any predictive system on the market or in the works that it is hard not to recommend the course to virtually all intelligence professionals.

While it may be too late to sign up for the course, you can view all of the videos on YouTube (I have embedded the introductory video below):

No comments: