What AI Means for Game Development

Normally when people discuss AI (Artificial Intelligence) in video games, it’s usually in reference to how enemy characters behave. New research has suggested that the neural networks of artificial intelligence could have other effects on game development; in particular how characters are animated. Artificial intelligence can allow developers to create characters that move realistically in real time, even through a complicated game environment.

There’s new method for handling character animations that will be showcased at the SIGGRAPH conference by ACM this summer. It’s called Phase Functioned Neural Networks for Character Control. Typically characters in games move through “canned” motion capture. Players can expect to see their characters move the exact same way thousands of times during a game. Artificial intelligence changes all of this; it gives the computer the power to create animations based on user input rather than choosing a pre-recorded movement to play.

Pros Of the Neural Network Technology

There are plenty of pros to this kind of system. It adds a new layer of immersion and realism to a game. Making characters move more naturally pulls players in. It also drastically reduces the amount of storage space needed. Typical motion-capture animation requires the game to come with a large database of animations. Not only does it mean the game takes up more space, but it also slows down systems. An advanced neural network animation system means games will take up less space and will run smoother.

Another great advantage is that the system doesn’t need every potential movement to be planned into it. One demo of the technology tested the capabilities of it by excluding motion capture data for crouching and walking through rough terrain. The system was able to learn however. It combined the animations of crouching on a flat terrain with the animations of moving over rough terrain, creating animations for crouching through rough terrain.

Cons of the Neural Network Technology

There are some downsides to the technology as well, mostly related to how new it is. There are still some kinks that need to be worked out. The animations still look a little rough and unnatural right now. It also leads to some strange movements. There are also some limits to how far this neural network technology can go. It’s surprisingly adept at learning, but things can quickly go wrong if you throw something new at it. The only way to guarantee the neural network works right now is to provide it with all the information it needs.

It’s also a lot to ask people to wait 30 hours following motion capture for the system to train and learn, especially if you need to re-record something. It’s also difficult – if not impossible – for an artist to polish up a neural network animation, which is currently possible with pre-recorded canned animations. Another problem is that even though the neural network acts quickly, it can still take a little longer than using a pre-recorded animation. The creators of the technology are confident that it can be improved on and sped up in time.

It’s interesting to see where game development goes from here as technology continues to advance. It’ll be great to see not just where game development goes, but also something other than the same old animations every five seconds! 

Given that people are often reluctant to try something new, you might need to use casinopromocodes.org to encourage them to play your neural-network infused game, but they’re sure to come back for more after giving it a go and seeing the results for themselves.