Are Advances in AI Development a Thing of the Past?
I remember when realistic squad tactics and lifelike AI were significant selling points on games, and each game wanted to outdo its predecessors in terms of how convincingly its enemies attacked you. Half-Life was a groundbreaking game for two reasons: its immersive no-cutscene approach and its enemy AI that for the first time made you feel like you were fighting a coordinated team.
We haven’t really grown since that point and I’d like to discuss why enemy AI development has not just slowed, but will no longer be a significant focus for game developers.
Human intelligence will always be better
One of the main thrusts of AI development was to create enemies that could convincingly mimic human behavior. I remember multiplayer bot behavior was a big focus for a few years in the mid 90’s when FPS games took off but not everybody was on the internet.
The internet is so prevalent these days that bots are no longer required. If you buy Call of Duty 4, you’re not struggling to find a packed game. Why waste development time on bots that only a tiny minority will ever use when those resources could be spend making downloadable maps and weapons that everybody will appreciate?
Did he just try to flush me out with a grenade? Is that a flanking maneuver?!
People don’t care about AI
Another thing to consider is in what direction the industry is moving. If you look at the highest selling games of this generation, only two stand out as requiring real enemy AI: GTA4 and Assassins Creed. The remainder, which far outnumber them (Smash Brothers, Mario Kart, CoD 4/5, Gears 1 and 2, Madden etc.) are all predominantly online if they’re multiplayer and don’t depend so much on a cohesive world. This doesn’t even count the music games that have become such a large component of the industry.
I think it’s unrealistic to expect significant strides to be made in game AI in the near future when it’s clear that gamers don’t care that much about it and where it is required it can be hacked and scripted enough to be convincing. In a single player FPS or a game like Assassins Creed, scripted sequences can take some of the pressure off AI routines and make convincing and challenging situations. On further reflection, there are many genres that never needed great AI at all. RTS, RPG, Sims, MMORPGs and Platformers all require extremely simple AI. It was really only ever a concern for single player FPS games, and those have evolved for the most part into pre-programmed events and fleshed out multiplayer modes.
Is the AI dream dead?
AI that is indistinguishable from human behavior is a holy grail of programming and while we’re obviously nowhere close to that goal, games could have been a significant innovator towards it. Games have been a catalyst for improvement in computer hardware and internet infrastructure. I was secretly hoping for games to create AI that could cross over into other fields but it looks like as gamers we’ve chosen a different path. Hopefully the focus on AI development is more of a temporary hiatus instead of a permanent departure.
I’m interested in your thoughts on the matter. Hit up the comments!
