Knowing how to win at Trillion Credit Squadron

A while back I used to be big into Traveller, the SciFi RPG, galactic empires and interstellar trade game. I was the type of person who found the rules interesting themselves (bonus geek points?) and I even used the planet generation rules to do my computer science project for school all  in BBC BASIC.

This was all in the past, I have been working in software for a while now, making my code smart enough to deal with most situations, ganining  insight into how the system worked. Ambling around the web in recent weeks I keep bumping into Doug Lenat’s Eurisko project. It was placed in a problem domain and was coded so that it could extract rules from the system. It was used on a few projects one of which was Trillion Credit Squadron. A Traveller scenario in which one designs a fleet for upto a trillion credits. A trillion sounded like a lot of money in those days. There were rules as to how ships could be built. No end of trade offs; powerful engines used more fuel, and so you needed a bigger ship, and by this time it wasn’t going go very fast. Then there were weapons to add on top of that.

What Eurisko did, and I have been digging around on the interweb for a while to find out more, was to concoct guidelines into what mattered in the game. I am not sure how it did it, but a certain amount of LISP was used. The insights from running thousands of battles and analysing the results led to a fleet formation which for two years it was unbeatable. In the end no one wanted to play against Eurisko so Doug moved on.

Doug went on to develop the Cyc project; AI being limited by the understanding of the English language Cyc has been entering millions of pieces of information. Painstaking work, now however the system can read and understand from the web (although it doesn’t believe everything). The link above was from two years ago, so should your lights start to flicker it may have reached conciousness and is devouring our national grid. An AI system that knows enough to understand and learn from what it reads is quite an amazing concept, what amazing things is it being used for now?

Now I wouldn’t mind working out how Eurisko worked. Might be an interesting project to code Traveller’s ship construction rules into Semantic web and compete against other AI teams. SciFi used to develop the future, it wouldn’t be the first time!


~ by zeristor on February 28, 2010.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: