As you'll know if you've read our latest issue's - dare we say it - masterful feature on artificial intelligence, when people talk AI they're usually focusing on some facet, led astray by casual marketing speak about "dynamic" worlds and "unpredictable" combat. AI is a much broader, more nuanced topic than that, extending from soundtracks that dog your movements to cutaway killshot sequences, and if next generation consoles prove as powerful as rumoured, tomorrow's games will have minds of their own.
It's not that next gen AI needs to be "better", necessarily. The last thing anybody wants in Call of Duty is a 'bot that plays like, well, a Call of Duty player. Campaigns that already reduce us to acid tears would become truly nightmarish. What we're after, rather, is AI that convinces us whether it's acting dumb or testing our wits, AI that's so natural and seamless we don't even notice it. There's a whole range of ways developers could achieve this. Here are a few possibilities off the tops of our heads.




















































8 comments so far...
Thos. on 27 Jan '12 said:
8 pages?!? *closes tab*
Can you at least consider adding a Printable Version button, with all the text on one page, like everyone else does?
OXM ETboy on 27 Jan '12 said:
We'll look into it...
bedlum69 on 27 Jan '12 said:
Small articles on every page is really annoying,why do this?
OXM ETboy on 27 Jan '12 said:
For starters, multiple pages are appropriate when you're dealing with lengthy list features - we don't do this with scrappier lists, reviews, blog pieces, non-listy features and the like. For seconders, and to be perfectly frank with you, we do more page view traffic as a result, and we need to keep upping page views to make money.
Sorry if it rubs you the wrong way, Thos and bedlum, but it is what it is.
Wolf of Badenoch on 27 Jan '12 said:
I must say Ed, your honesty is refreshing.
Yes, it's annoying having to keep changing pages, especially when reading it on an iPhone. But, if it enables the site to keep running and sooths my major gaming addiction until I can get my hands back on my beloved Xbox I'll put up with it.
On the subject matter, I'd love to see the AI in shooters actively adapt to what your doing. For instance, CoD on Veteran is literally an exercise in frustration, often running the same 30 second segments 50+ times till you can JUST squeeze through and try the next 30 second segment 50+ times.
Imagine, if the AI had learned your patterns and had resistance waiting on your 7th run that wasn't there on the 6th, forcing you to try a different route. This would force you to try something different, perhaps stumbling across a different (easier) way to tackle the bottleneck.
Decent_Jam on 27 Jan '12 said:
I really want to see AI that does the sort of things you do occasionally, mess around for a bit of a life, do things which are a bit pointless just for fun. Generally in my case though this boils down to shooting, punching or generally harassing whoever I'm playing the game with though, so they might have to think of a few other ideas.
Grummy on 1 Feb '12 said:
I really don't see the problem with multiple pages. Are people so lazy now that they can be bothered to click their mouse button to make the page turn?
Anyway, AI, the problem with AI is that if it is pushed too far something could, hypothetically, stop being a game and start becoming too real. Looking at your AC example, one of the delightful things about AC is that it can be a relaxing experience just bouncing across rooftops and getting into random skirmishes then going back to bouncing across rooftops. The idea of an AI system that learns and adapts is sound, but sometimes people don't want hyper intelligent AI, they want an AI that is good enough to make the experience enjoyable and still let them play as they like. A dynamic learning AI could take your favourite relaxing game and make it increasingly more taxing. That *might* make a game more exciting, but just as likely is that it would make the game more irritating and spoil the experience. Personally, I wouldn't want the generic guard AI in AC changed, I think its perfect as it is because it works for the game.
That's the big challenge with devleoping AI, decing when to say 'enough', not just developing the most complex AI system you can just because you can, but developing a comprehensive AI system that fits the game.
Take Skyrim, you talk about a living ecology, which sounds good in theory, but think about it some more, how much time are you really going to spend following and watching the activities of a single bear? OK, sure you could perhaps see it kill something, but to become an a more real ecological AI system, it would then need to do something with its kill. Eat it there? Drag it to its den? How about a mating ritual for it, could we run into a couple of bears in mid coitus? What about a bear giving birth to a cub? Maybe we could get an achievement for assisting the animal in its birthing? Questions on this need to be asked. Where is the cutoff point? What level of ecological AI is too much? What fits the game and what doesn't? Does having an extensive AI for wildlife really matter in a game? Have you ever tried just hunting in Skyrim? It works great, it doesn't matter that the animals don't have a real and full AI, they move around naturally enough so for the job of hunting, it feels real, and THAT is the key. I have a friend in the indsutry who was telling me once about when one of his lecturers in Uni brought up the subject of developing AI, in a nutshell, his Lecturer said "why spend thousands of man hours and hundreds of thousands of pounds developing an AI when you can fake it for an effect which, on the surface, will appear real". His point is sound. Something peripheral like wildlife doesn't need a comprehensive AI because you don't spend all that much time in the game really dealing with it. It adds a dimension to the game, assists in the immersion, seeing wildlife around and busying itself, THAT is its purpose, and it can be faked enough to such an extent that it does the job well without developing a comprehensive AI. Don't get me wrong, personally I'd love for everything to have such a comprehensive AI system, but in reality it takes up a lot of time and money for something that is largely irrelevant.
Where I personally want to say stronger AI developed is in conversations, as you demonstrate in your ME example, but also in generally dynamic worlds.
With regards to conversations, it would certainly be a monumental task to get it working, but it is one of those systems that, when built, would always be used and just be slightly changed to add or take away certain words and phrases as required for the specific game, the bulk of the system would be lifted direct from one game to another, certainly with a company like Bioware that relies strongly on these conversation elements in 2 seperate franchises.
What I really want from this though is a seemingly small thing, I would love for games with such a system in place to be able to read and speak my characters name. If I type in 'Grummy' as my character name in an RPG (I never use Grummy in game btw) then I would love for NPC's to call me 'Grummy' rather than 'Hawke' or 'Guild Master' or 'Chick Chaser'. It's the same with ME as well, yes you can name your character anything you like, but his/her surname is still Shepard and is still called Shepard, for all the work Bioware go to to make the game interactive and make you fell like you ARE your character, you always know that everyone who plays it is being called Shepard. In many ways, I'd prefer to not be given the option of naming characters sometimes, in a game like ME especially. Just give him a full name and use it then make the game feel even more cinematic rather than clinging to this image of Shepard being our avatar in the game world.
A small aside here, but along with that, if Kinect i being built in to next gen Xbox, then lets start making games that can scan our faces and bodies, or faces and bodies we give it to scan, then build them in game for us. Some people would love see themselves built in game, and strive hard to do it but never get an accurate facsimile. Others would prefer a movie star, building the face of Arnie as your Nord in Skyrim for example. This really needs to be implemented.
As for Dynamic worlds, well, even though I say the ecology of wildlife doesn't need a vast improvement, I DO think that ecosystems should be improved. Again, take Skyrim, a game perfectly designed for a dynamic ecosystem, and in reality all you can do is earn money form completing simple pointless chores and increase available funds at merchants through perks. Well, why can't it be more dynamic? Why can't the people of Riverwood NEED someone to chop wood or hunt for them? Why can't that small village demostrably grow and change depending on how you treat it? If you spend more gold there, why can't the homes look better? Be better furnished? The residents wear better clothes? The stores stock better items which you can see on display? How about an issue with supply and demand? Riverwood, for example, has no direct source of food, a few chickens and a cow is about all, but no farm of any kind to speak of, and the only growing veg is a small selection behind the inn, so clearly they have need of a food source, you could well go hunting and feed them and get well paid for the food you bring back, or perhaps you could just periodically bring them other food you have in your pack for increased profit, but how about being able to broker an agreement between th residents of Riverwood and the Jarl of Whiterun to send them regular shipments of grain and fruit?
With a dynamic world you could create gameplay through your activities. Sure you could ignore it all with no detriment to yourself and play the game as it is and enjoy it like crazy, I know a lot of people don't like using the cooking or smithing activities and suffer no real detriment for ignoring them, but having them there adds more choice to the game. Similarly, yes, travelling across the map just to sell food to Riverwood because you earn more money there than from selling it at your nearest shop will for some people be a redundant feature, particularly as gold becomes easy to obtain later and serves no real purpose, but the option being there adds more to the game. Especially if it is used as an instrument of dynamic gameplay. You could just factor in a basic supply/demand element, see who needs what in different cities and setllements then change buy/sell prices to compensate, or you could make it grow. If a settlement needs food, you can get a regular job form them which will send you somewhere to hunt or gather. Its a small thing, but it would be there. If you could broker deals and arrange shipments from place to place, then you could see those shipments travelling the roads, perhaps hire on as a guard-gameplay created through activities. You've chosen to invest time in developing a settlements ecosystem, and it has developed a new gameplay options. The list of possibilities is huge, and the choices you make should make a difference. I personally think one of the best gameplay elements that could be added to a game like Skyrim would be option of developing a whole settlement. You stumble across a small camp of people who want to make a permanent settlement, and you help them through gathering food, wood, some animals and farming tools and they become a farming community. With those items gathered they would start cutting down treest, clearing areas and planting crops, building paddocks for the animals, building rudimentary homes. The settlement has grown from a rough camp to a small farming community. Then you can continue to provide more materials and at certain thresholds, they have enough to do different things, build better homes, a mill, a woodcutters etc. the end game would be building it up into a thriving settlement to match ones like Riverwood. With the growing of such a settlement you would have constant choices of jobs to hunt, gather, defend, and later on to help build their economy, perhaps hiring workers to build a pallisade or stone wall, meeting with the local Jarl to hire some guards for the settlement.
These of course are really AI that would help in Skyrim and other open world RPG's primarily, but couldn't you imagine a similar system in place in, say, the next RDR game? How about just about any sci-fi shooter at all that dared to be more than a linear corridor shooter.
Other factors of a dynamic world would be giving groups and gangs active mentality, let them choose to do things. So if you're playing Skyrim again, perhaps a tribe of giants could attack a small settlement because their AI has decided they should. This could reflect in a game like GTA, rival gangs start attacking each other seemingly at random rather than being generated htrough advancing the game. This sort of system would work really wel in a game like the Godfather ones where it has families at war and locations to take over etc. A dynamic world would see these things happening and you stumbling across them and dealing with them at any time rather at scripted moments.
I think Dynamic worlds are the real next step for AI.
CunningSmile on 1 Feb '12 said:
Good point Grummy; what happened to the dynamic economy that we were promised in Skyrim. A lot of interviews leading up to release talked about investing in a town making it thrive whilst destroying the local mill would destroy the town. I've seen nothing like that, except for one particular shop I use often has slightly more cash now. Even Fable 2, as stripped down and simplistic RPG as it was, had evolving towns so what happened to Skyrim's evolving towns Bethesda?