I’ve been watching a new show on Food Network called Dinner: Impossible. Every week, a top-tier professional chef (Robert Irvine) is given a nigh-impossible challenge to complete. In the first episode it was merely to create a five-course dinner for a wedding reception of 200 in ten hours, but things have gotten much harder and more bizarre since then.
I’ve thoroughly enjoyed the show. Robert is a pro from Dover (possibly literally, since he’s English) and has never failed a challenge yet.
Now that I’ve watched about five episodes, I’ve noticed that once Robert gets on site and manages to map his current challenge to his formidable previous culinary experience he’s unstoppable. He’ll get into his groove and start barking orders and delicious meals will pretty much appear out of nowhere.
So the producers have ramped up the challenge by throwing him more and more curves on every episode. The most recent challenge had him cooking in a tiny train car with only one of his sous-chefs and the conductor of the train was constantly interrupting him with change orders. Once his food was nearly done, he had to transport it to the dining cars, reheat it and then serve it.
Now, as I was watching the most recent episode, I oddly enough saw a parallel to game AI. Specifically, the episode reminded me of the original Command & Conquer.
The AI for Command & Conquer was pretty weak. It didn’t respond well to feints and was highly predictable; thus it was pretty easy to beat. Of course, in order to beat it you had to get your economy going. You had to build your base, get your Tiberium stream flowing and start cranking out units. So later missions of Command & Conquer rasied the difficulty not by making the AI smarter (the AI is the same in all missions) but by making it more and more difficult to get your base built and get into your groove. One late mission had you landing on a beach with just your mobile construction vehicle. The beach already has emplaced NOD guns so your MCV starts taking damage right away. If you’re quick you can run right and just barely get away from the guns with a sliver of health left on your MCV, which you can then deploy near a Tiberium field. But a damaged MCV deploys into a damaged Construction Yard, so a lucky hit from a passing NOD bike is all that’s necessary to hear EVA say “Your mission is a failure”.
When the Ensemble guys got together to make Age of Empires one of their goals was to create a non-cheating AI that played at several levels of difficulty. For a long time I thought it was kind of foolish to spend so much time on a non-cheating AI when for the most part we players wouldn’t be able to tell the difference. But now I’m convinced they made the right decision.
As for Planitia…well, I think Planitia will be lucky to have any kind of AI.
wouldn’t it be extremely cool to have an open interface for the ai? you would have to make some kind of sensor functions and action functions accessible through a script language (i used squirrel through sqplus once and can only recommend it).
i would love to get my hands on some ai programming and you could even host some kind of ai contest 🙂
Actually, Ensemble Studios has provided AI scripting tools for a lot of their games and has held contests like the one you described in the past!
Odds of me doing it? Pretty slim. I don’t even have a scripting system implemented in my game yet.
The Car in Age of Empires I…
When I was in high school, the very first version of the Age of Empires and the expansion version, Age of Empires: The Rise of Rome was released….