|
Reasons AI Real-world computing Requirements
Caveats
|
Preparation for AI Principles of Artificial Intelligence provides a solid grounding in the techniques of modeling, engineering and evaluation that form the basis of modern AI. These techniques are essential for further research in AI, and Principles of Artificial Intelligence is a prerequisite for advanced courses such as Pattern Recognition, Machine Learning, Computer Vision and Computational Linguistics. We expect to offer Principles of Artificial Intelligence only in the fall semesters, so don't wait! Interest in Real-world Computing Principles of Artificial Intelligence is not bogged down by any grand vision for computers that exhibit human-level intelligence. It addresses a more specific and practical issue: how does software design, implementation and evaluation change when programs must take real-world data as input and take real-world actions as output? It is fantastically difficult to carve small and interesting tasks out of the world and solve them in a constrained way, the way you would carve up the problems induced in typical computer programs and write separate modules to handle them. (Some AI researchers have even joked about "AI-complete" problems, as though real-world problems are so complex and interdependent that any sufficiently interesting problem will effectively involve a solution to all of them!) But the use of real-world actions and real-world data is now an increasingly common requirement of cutting-edge computer systems from internet agents to embedded devices. Any commitment to the real world, no matter how constrained, brings with it the ensemble of perspectives and challenges to which AI research responds, and which this course addresses. Principles of Artificial Intelligence counts towards category B of the breadth requirement. Principles of Artificial Intelligence makes use of classic ideas from AI including
Principles of Artificial Intelligence also delves from time to time into techniques from linear algebra, calculus, probability and statistics. The class will be easier the more you know of these subjects, but proficiency is not essential. The mathematics is there because we don't just teach how AI algorithms work, we teach when and why they work. This time pays off in future research, where you often must use existing algorithms judiciously, or design new ones. But you need not master it all right away, since problem sets, programming and exams emphasize what and how, not when and why. If you haven't studied representation, state-spaces and search previously, and have no mathematical background, Principles of Artificial Intelligence may not be the course for you. CS 530 now covers the material that Matthew Stone covered in CS 520 in spring 2000 and in spring 2001. If you took these classes, you know the stuff. Otherwise, you probably don't.
|