I'm writing this post while waiting on a release build to compile. I'm using incredibuild and it looks like there are about 10 machines helping at any given point. All that compute power and still I'm only half done at the 12 minute mark.
I once worked on the IronMan game and it had compile times under 1 minute on POS machine. The difference was the size of the code base and the strict enforcement of compile times as the ultimate virtue in code.
At the time I thought it was the dumbest policy imaginable. Why use a typed language if your going to throw out type safety? Why hand code different variations when a template can handle the job?
Now as I sit here and watch blue bars change green past the 20 minute mark, I can't help but think the Secret Level guys had a point. Well the game sucked and they are out of business but man those compile times were amazing.
Wednesday, June 29, 2011
Tuesday, June 28, 2011
Next Generation Game AI
What will next generation game ai look like?
It will animate smoothly. This is not a trivial problem but it is one that is solvable with modern tools and techniques. You can see some of it already in games like Battlefield 3 and Madden. Up the animation budget by 10x and the processing budget by 100x and suddenly you can start using some new animation techniques that weren't possible before.
For starters search becomes a viable tool at runtime. Not sure what stop animation you can play when running forward aiming backwards and jumping over a red turtle? No longer do you need to bake that transition into a graph. Search over stop animation frames and pick the one that pose matches best. Pops and bad blends are now only an issue because of lack of data and not because your hand built graph didn't handle every possible case.
Fancy IK starts to be more reasonable. The players left foot is on a curb and his right is on the ground and he's strafing left. Well if his left toe crinkles then his hips will rotate by some massive jacobian * x. You laugh but this type thing will make a visual difference. It will help give actors that wonderfully overused artist term *weight*. Perhaps a better example might be handling two sword fighters clashing swords. Having the IK try to model the muscles involved would be complex but would give awesome results.
Just having more animation will make the AI look way better. Procedural hacks that slide guys around, melee animations that miss by several feet, walking near things without adjusting motion, these all go away when you can replace with animation. Well theoretically. It might become a cost issue as actor time for mocap becomes and issue. It could be expensive to have and actor climb over the 5ft wall and then the 4ft wall and then do it ten more times for variation.
So no, next gen game AI will not make you think you are playing with real people. It will not help you solve puzzles or make you feel like the world is filled with living things. It will animate smoothly and that is awesome.
Monday, June 27, 2011
The world is not going to end
Will the fear mongers please take a break. They have a track record of being wrong 100% of the time and yet they seem to be bellowing louder than ever. Global warming will not destroy the planet. It might change it but it won't destroy it. Artificial intelligence is not close and if it were I see no reason to jump to the conclusion that it would self improve at an exponential rate or want to do us harm. Neural modification is not close and if it were our humanity would not be cast aside willingly or easily. Nuclear catastrophe is unlikely and if it happened would not destroy the planet.
There are a few common threads here. One the world is massive and complex. The systems involved having many balancing forces. If they didn't then odds are they would have crashed already. The fact that they haven't means that there is likely not some tipping point that can be easily crossed. Two these doom scenarios are forseeable and preventable. I will not willingly walk off a cliff and most of semi-rational mankind will not either. The doommongers often point to the stupidity of mankind as the cause but that's just their bias based on some irrational sense of superiority. Three the doomongers have bad motives. They are either trying to get money, get power, get attention, or feed their own need to look down on the rest of mankind.
Tuesday, June 21, 2011
Crime and Punishment
The worst book I ever read or partly read as the case maybe. I found the main character completely unbelievable. I can see why it's a classic though. It's got all the trademarks of a classic. The main character is a victim. The main character is a whiny douche. The ratio of whining to action is almost infinity.
Honestly though why is this book a classic? Is it fun to read? Absolutely not. Is it a timeless tale that anyone can relate to? No! No one can relate to this story because it makes no sense and is completely unbelievable. Star wars is more real than this tale. Is the prose amazing? Maybe. I doubt it though because it was originally written in Russian. So what is it then?
To be continued...
Hola
Welcome to SomewhatCynical! This is my new blog to post commentary on whatever strikes my fancy. Some topics you might see here are politics, gaming, programming, AI, economics and books. Everynow and then I will probably throw out an NPR rant because goddamn I need some sort of frustration vent when it comes to NPR. Anyway welcome and hopefully it will get better from here.
Subscribe to:
Posts (Atom)