Several presentations were given. One on the fact that people don’t give a shit about online security because they’re irrational and shortsighted, all they want is convenience. So if you’re planning on being an evil business man, don’t bother with online store security. I should have known.
Then the new ZFS file system by Sun for virtual server storage was presented, and compared to EXT3. I actually learned a lot from that presentation, like how RAID is used in parity checking. Also, the ZFS file system actually just uses a data pool for all drives and volumes (making volumes a thing of the past) managed by the ZIF I/O system, which is integrated with the actual partition manager. Partitions become invisible and irrelevant, and all the super blocks are managed by “uber blocks”. It’s all very interesting, JFGI.
I totally awesomified everything with the physics presentation. I feel like…. really good. It went really smoothly. Interesting because I had a nervous breakdown right beforehand and fucking puked. In the bathroom, but I do have a horrible fear of public speaking.
But then the Othello game team presented, and they were actually really good. The most amazing of the AI scripts shown by the advanced non-linear algorithms class was the one by Iain. It took the game timer and computed how many seconds it has to make this move (realistic approximation) then spends exactly that amount of time. It does a possible moves benefit/penalty difference tree search, but to speed it up it does alpha beta pruning to remove sub trees with first layer search expensive nodes to within a certain tolerance (each new layer search given the most promising avenues in Layman’s terms.) The result is a beautifully executed compromise between the fastest computed and most predictive model, where for each new search layer moves are eliminated until the best is given or time runs out in which case the best of the moves so far in the search is used. This means that the longer it is given, the better the move it will make, but because it is entirely scalable to time as it is ACTIVELY AWARE of time, it will always make a move at least one layer of predictive strategy deep, but ALMOST ALWAYS MORE.
Yeah. Fucking genius. It beat EVERYONE, and their relatively puny AI programs.
So, yeah, I would say that I may have been the highlight, but it was hard to to Iain’s AI program. I was blown away.
THERE.
Then, we headed down to Eric’s Aaron’s house (I don’t have their names down yet blargh!) and played rock band for a while. I got to sing Metallica lol. After a while we went to a dinner with Dr. Liow and some other professors I don’t know, Chinese food for all! It was pretty sweet. We talked about how awesome Chrono Trigger is and why A* is still quite often the best pathfinder if you prune the search properly for large maps with area zoning. Also, there was some discussion as to why romance novels are low grade girl porn, and agreed to have another AI face off. Plus there was specifically Colombian coffee at the place, that is my favorite variety! It was all around awesome. Then on the way home we were almost killed in an accident because of the freezing rain. But we didn’t hit anybody or vice-versa. We DID do some awesome donuts though.
musings of ninjarat 