AI Class II

Ok then. They got the Search lessons up. And are promising to post a homework assignment by about 4 hours ago... Also the quiz-post-refusal thing seems to have been a server loading problem and I didn't have any trouble posting answers today. So still a bit behind the curve here, but moving in the right direction.

There are some slips, probably mostly on my part. Like the quiz question about whether a Depth First Search is guaranteed to find a goal and be complete. I forgot to remember that the lecturer mentioned that we were dealing with an infinite depth search tree for this particular incident. So, more minus-quiz points for me. Gotta hang onto every word apparently.

<Edit mode="stew">
Overnight I realized that there were two (my count) examples of lapses in pedagogical technique in the first week's videos -- three if you count not defining Rationality but then including it in the summary slide for Unit 1.

First is the Depth First Search question above. I replayed the lesson -- unit 2.20 -- and he does say  "...lets move .... to infinite trees..." 20 seconds before completion of the quiz question presentation. So I should have remembered it. But, if he had repeated the infinite tree condition at the end of the question I might have caught on to what he was getting at.

The second was in unit 2.31. He describes the simple "Vacuum World" environment and calculates the number of states in a two position system by writing 2 x 2 x 2 = 8. This is the correct number but not the right calculation, and -- my excuse for failing the quiz at the end of the next unit -- when the system is scaled up with more positions one needs to use the right calculation, which is: 2 x 2^2 (notice that 2 is one of two values where this is equivalent to the previous multiplication, ?maybe three if zero^zero is a number?). This is because there are X possible conditions for N positions -- every position can be either clean or dirty -- so the total number of environmental states is X^N, not X*N. I merrily went along with the multiplication paradigm when it came to scaling up to 10 positions and multiplied 2 times 10 instead of raising 2 to the 10th power. Again I might have caught on, and had a better understanding of the issue, if it had been treated more rigorously in the introductory case.
</Edit>

In a different example they present the idea that you can use an estimated-cost-to-goal value to guide a search in fruitful directions. This is called an "Heuristic". However they never defined the word but just started using it in the middle of describing some algorithms. Lucky me, I'd already read the Book so I knew. Just like the "Rationality" thing in Unit 1...

Have an online study group meeting tomorrow (Weds) night in which we are supposed to discuss homework confusions (among others). So I hope we get the homework in time...