|
Why Stanford AI class teachers are Ignorant Motherfuckers
By donnalee in donnalee's Diary Fri Oct 07, 2011 at 09:59:16 PM EST Tags: ai class, stanford, information wants to be free, multi-agent systems, closed fist of the teacher, aporia, standing on the shoulders of giants, obsolete education models, judgment, control freaks (all tags)
|
|
|
The on-site version of the Stanford Introduction to AI (CS 221) class's first programming assignment is to implement, in a Pacman game program, several well-known search algorithms:
http://robots.stanford.edu/cs221/progAssignments/PA1/search.html
|
|
|
|
|
|
Instead of assigning solved problems for which the teachers already have the answers, why don't Norvig and Thrun emulate Socrates and ask hard questions they don't know the answers to (http://en.wikipedia.org/wiki/Aporia)?
So in the Pacman assignment, instead of having students reinvent the wheel, why can't the instructors make all their knowledge (i.e. the solutions) transparent, and ask the students to solve the further problem of how the game can be made even more intelligent?
My method would be to allow the user to switch between algorithms at runtime. In some cases a depth-first search might be better, in others a breadth-first, in yet others A*, etc. Or some combination of the different methods? I would start by making an agent that would run the program as written, and then use natural language feedback and commands to switch among search agents without restarting the agent wrapper.
But since I'm missing the algorithms, and since Norvig and Thrun want to prevent the solutions from being posted anywhere, needless obstacles are placed in my way. Doesn't the still, quiet voice inside warn the Professors that restricting the free flow of information by withholding solutions and requesting that others do so as well, is contrary to the spirit of science? Why can't they think like giants, providing the shoulders for others to stand on, to see farther than they can?
Another way to make the program more intelligent would be to have it try different search algorithms itself and select among them based on some user-defined performance metric (total points, for example).
In fact the instructors might have the latter already for automatically grading homework submissions. But, unlike Buddha (see http://www.accesstoinsight.org/tipitaka/dn/dn.16.1-6.vaji.html Part Two, Paragraph 32), they want to keep their fists closed, and use the knowledge they selfishly hoard to pass judgment on others; instead they could share freely, and work with students to take AI to a new level beyond their wildest imaginings!
|
|
|