4 ms·
I do agree with the author about the lack of progress in FPS enemy AI. I follow AI much more than I following gaming. I definitely find it curious that enemy A
by joe_the_user 10y ago
I do agree with the author about the lack of progress in FPS enemy AI.
I follow AI much more than I following gaming. I definitely find it curious that enemy AI has hasn't made a lot of progress when game-playing in a different form is such a focus of deep learning.
- azernik 10y agoGame AI tends not to use machine learning approaches - it's all hand-coded strategies.
- e12e 10y agoI suspect machine learning (and other, older ai techniques) might work for games - but one would probably have to train them for the right goal: fun. I suppose for some players and some games, it makes sense to train for "good" - but even with chess - is it really fun to play an ai you cannot realistically beat - ever?
- stevenwoo 10y agoI haven't worked in video games since the PS3/XBox360 but everything was bespoke on every platform for every game except for maybe low level audio in my ten years on PC/Mac/PS3/XBox360 and I had to review/examine code for games other than the titles I worked upon. Even games that used game engines/physics engines have heavy customization. I only worked at smaller studios but the practice of starting from almost scratch every time can make it hard to have a standard toolkit of AI, the reuse incentive was low.
- Radim 10y ago1. HW resources: games run on extremely tight HW budgets. It's probably the most competitive, cut-through industry in terms of what you can afford to do with a given CPU/memory/bandwidth per frame. While DL is still in the "let's throw a thousand GPUs at it, wait for a week, see what happens". 2. Quality control: DL is a blackbox. Bad for the playing experience for obvious reasons. And even if just using pre-trained models, the lack of control goes directly against what both the designers and players typically want (predictable fun). 3. Finally, DL hasn't made THAT much progress here, actually. There have been serious efforts to apply DL to modern games (e.g. DeepMind with Starcraft), but what you've probably seen is "just" DL applied to some finite-state, 100% deterministic, complete-information, zero-sum games. A far cry from complex dynamic environments. Which is not to say it cannot be done. Just that the incentives (both business and technological) are aligned differently -- too little bang for a lot of buck.
- lloeki 10y ago> HW resources: games run on extremely tight HW budgets Come on, you could have decent AI back in the day that pinned me down and made me run for cover, then had me at that cling gling oh shit moment when a perfectly timed grenade lands at your feet just as you were starting to take a breath†. Today we're stuck with scripted sequences and playing whack-a-mole with sniper rifles. † only HL1, Halo and FEAR seemed to manage that.
- Radim 10y agoYou may be under appreciating how much effort goes into good old hacking in games, to make things look "natural" but with minimal HW footprint :) Cutting corners. Especially back in the day! It's more of an art. Applies to graphics, AI, audio...
- dkersten 10y agoQuality control: DL is a blackbox. I remember a discussion on gamedev.net once where exactly this was brought up. Its very hard or even impossible to tune an ANN to behave how the designers want it to. Simpler systems may not be as technically intelligent, but their behaviour can be tuned, encouraged, scripted etc to provide the experience that the designers wish. Beyond that, I remember they also talked about how it simply wasn't worth it because game AI isn't about being smart, its about being fun, interesting and "looking smart" (ie looking humanlike, human flaws and all). Machine learning is simply solving a different set of problems than what games have.
- drawkbox 10y agoSome games, if multiplayer and a bot is put in, can use paths learned by human players to appear more human. We did some of this for racing games which really have pretty easy AI when compared with character games or strategy. An example is take ghost runs from real players, essentially splines from real players good and bad, but adjust the runtime play to the current vehicles but try to run that line. That way runs appear more human and probably can't be replicated by AI as well. For racing there is also the typical rubber banding AI which tunes to the runtime play[1]. Most AI in games really come down to just scripts and action recipes. Zone based events/actions, path passed actions/events, follow/ranged/staged reactions, global actions/events, AI directors that can tune the game to the player or difficulty dynamically etc. Left for Dead probably has the best AI director out there for tuning some great co-op games[2]. It does med drops, weapons, enemies appearing all based on the current state of the players and skill. Music also changed per player based on their situation and intensity of that situation. The Director, sometimes referred to as the AI Director, or simply as AID is the artificial intelligence of Left 4 Dead that features a dynamic system for game dramatics, pacing, and difficulty. Instead of set spawn points for enemies, the Director places enemies in varying positions and numbers based upon each player's current situation, status, skill, and location, creating a new experience for each play-through. The Director also creates mood and tension with emotional cues such as visual effects, dynamic music and character communication. Moreover, the Director is responsible for spawning additional health, ammo, weapons, and Special Infected, like the Witch or the Tank. It should be noted that there is another Director in the game, which controls the music on a per-player scale, called the Music Director. [1] http://www.giantbomb.com/rubber-band-ai/3015-35/ http://www.giantbomb.com/rubber-band-ai/3015-35/ [2] http://left4dead.wikia.com/wiki/The_Director http://left4dead.wikia.com/wiki/The_Director
- meheleventyone 10y agoThe goals are very different. AI in games needs to provide a challenge to the player whilst still being comprehensible. Most of all it's designed to be beatable. Part of the reason games make a good deep learning challenge is that they are consistent systems that people can comprehend and reason about pretty easily. Most FPS games with the exception of titles like Alien: Isolation simply haven't needed anything complex. Essentially more complex AI with the same freedom wouldn't be noticed by the player in most FPS games. Given more freedom a more complex AI suffers from being less comprehensible as the player is less aware or unaware of the actions it took to do something clever. For example a clever enemy that flanks the player is experientially identical to a scripted event that spawns dumb enemies on the players flank. Further the advantage to the second technique is not only that the AI requires less work but the entire scenario can be carefully designed. Which is not to say a more intelligent AI wouldn't make for a fun FPS game it's just people haven't been making those games.