The everyday blog of Richard Bartle.
RSS feeds: v0.91; v1.0 (RDF); v2.0; Atom.
11:08pm on Friday, 28th September, 2012:
Anecdote
I was at a Games and AI workshop all day today. I wasn't presenting, but it was at my home university, Essex, so I showed up anyway.
I've been to these things before, but this was a little different. In the past, it was mainly about AI, with games being just a cool application domain; this time, there was more about AI that's actually useful for games, both in terms of its novelty and its practicability. Game developers are famously (and all too often rightly) suspicious of academics telling them what to do when either they already know how to do it better or it's irrelevant to their needs; this time round, the presentations were neither of these. The main reason for this is that there was a focus on the Monte Carlo method as a means of providing a fitness function for search, which has a number of features game developers like: it's fast, it's anytime (that is, you get better results the more iterations of it you use but you get passable ones even for relatively few), it can be helped by heuristics, it can be adapted for stochastic results, and it works a treat for AI opponents of human players.
That said, I was struck how several of the presentations could have been given when I was doing my own PhD in the 1980s. They didn't really add much to search except that they were being used for an unusual search space; the only real interest was in their evaluation function (eg. how good is the game I have randomly generated?). They had the same old problems of good old fashioned AI planners to do with cross-level planning (they can't mix planning and meta-planning operators, which in this context means they can't create rules and then create reflexive rules or rules about rules). Still, it was generally interesting stuff.
The workshop ended with a meal at Wivenhoe House, the university's stately home that's now been converted into a hotel school (somewhat controversially, because hotel management is not a research-oriented subject). The food there was surprisingly good, although too expensive for me to consider going there for lunch during term-time. The hotel itself had a few teething troubles, though, including features such as not being able to get any hot water out of the shower unless the hot water tap on the basin was running, or having electronic keys that stop working — sometimes you can't get into your room and sometimes you can't get out. I'm sure these will all be fixed in time, though, as the possible impact of legal action is always an incentive to improve.
I met some interesting people, too. Having spent some time speaking to Néstor Romeral Andrésof Nestorgames, I'm starting to think I perhaps ought to manufacture some of my own board game designs to sell over the ol' Internet.
Hmm, perhaps n0t, though. I'm great at creating things, but less than successful in getting them to market. Still, if people are still buying games, perhaps it's worth a shot?
Latest entries.
Archived entries.
About this blog.
Copyright © 2012 Richard Bartle (richard@mud.co.uk).