View Full Version : Future of gaming

Jun 26, 2007, 11:04 AM
So I finally got to see Die Hard 1 last night thanks to ITV. all throughout it I had daydreams of playing a game like this, and that not Nakatomi Plaza game on the PC from a couple of years ago. I'm talking about an actual game that tries to mimick real life.

Through school, then college and university the bus always took me through the same route. After 10 years on the same road I got bored and started coming up with this little idea of mapping out the whole area and introducing a problem that was seemingly random.
Such as travelling over a hill, an alien ship would crash land. Depending on a few variables you'd get the Police, Army or whoever involved as well. You would have to make your own weapon, or find one. Events would be randomised or influenced by minute decisions so the whole thing could be a strategy game, or all out action, or whatever. Hell just play cards with an injured soldier.
I suppose what I'm on about is a completely non-linear, absolute sandbox game. Though with some restrictions. No scripts, everything generated - story, game type, outcomes simply as the game progresses as it would if it were real.

We're all playing games with basic AI. Even the most "real" AI's only think in short bursts - ie. "How can we kill this guy in the shortest time possible?" or "Where to run!". In Die Hard I was inspired by the actions of the terrorist group and the dynamics you simply don't get in a computer game. Such a game would be so engrossing and personalised for the player, but would require tremendous amounts of processing power and AI.

If I wasn't so addicted to working on 2D games I'd give this a shot myself :D

Suppose I'll be waiting for next gen for all that.

Jun 26, 2007, 11:25 AM
We're all playing games with basic AI. Even the most "real" AI's only think in short bursts - ie. "How can we kill this guy in the shortest time possible?" or "Where to run!".

I disagree with this; I think it's more of a 80/20 where 80% of games are flawed by basic, rudimentary/mindless AI and 20% have advanced or elite AI. Games are starting to get to the point where enemies react accordingly to different situations, and (especially in first and third person shooters) will fall back, attack, flank, take cover/hide, etc. Resident Evil 4, Gears (and probably Resistance; I've never played it) as well as most of the Tom Clancy games, Metal Gear Solids and even Call of Duty have great AI.

Slowly but surely developers are grasping evolved AI and now they have the processing power to do it. Up until now, they were handcuffed by hardware that couldn't handle it.

Be patient Jimmi...I think games like what you are dreaming about will come soon enough. ;)

Jun 26, 2007, 12:25 PM
You want situations to have their own version of procedural generation (something like that), eh, Jimmi?

Jun 26, 2007, 12:34 PM
Exactly. I'm trying it myself to a much lesser degree in my current project (though with some fixed points like bosses and environment interaction), and it's surprised me to find so few games are truly open ended.

Though in order for any of this to work we're going to need a perfect physics engine and AI (or artificial stupidity in some cases :D ). It wouldn't need to rely on graphics so much, but would likely require a very beefy CPU. More than anything we've got access to in gaming now.

Jun 26, 2007, 12:37 PM
Just wait, I'm sure whatever Killzone 2 stuff is shown at the "it's not E3 anymore but we still call it that" E3 show this year will be hyped by Sony as being revolutionary and including AI on the level that you're looking for. :p

Ah, the almighty Sony hype. :D

Jun 26, 2007, 01:36 PM
Ah, the almighty Sony hype. :D

There's nothing more powerful ;) or flatulent :D

Jun 26, 2007, 02:26 PM
Your idea reminds me of the computer game they play in the book "Ender's Game".

Jun 26, 2007, 03:37 PM
hm a complete game generator ? i would guess that is still decades of AI developments away to get a complete random game

it might be possible to get a certain degree or scale working but across all game genres and also producing balanced games that would be a pretty fricking dinosauric task to write such a "random compiler"
after all it would also have to automatically create textures or models and animations

on increased horsepower providing advantages in _this_ generation: well i wouldn't be so sure about that .. after all the bottle neck of AI isn't on the hardware side but more on the programmer side
while more power is good it is still mostly used for increase other aspects like more enemies etc.

after all, where are the really great AI ? only in shooters and then all the can do is a limited fall back or counter attack or a flanking movement restricted by level design
after all enemies still have to make it easy for the player and learning agents are still mostly unneeded in games since you can get more performance by hardcoding and also most grunts don't leave long enough to it having effects ;)

self rewriting AI/agents won't be around widespread for quite some time
easily visible looking at this generation of gaming hardware: complex branch prediction, out of order executions etc. aren't in fashion at the moment instead we have multicore + lottsa simd units for lot of easy symetrical calculations
in general the bus systems between the units are still problematic for shared calculation of multiple AI units

good for physic calculations but for AI ? i'm sceptical about that
(damn there are still board games like go where AI still fails miserable against human amateur players

also a huge problem of AI is actually not the AI itself but more on the animation side and working together with environments

Jun 26, 2007, 04:15 PM
I suppose it's not entirely random. There would be area constraints too, the age ol' invisible wall. 1 event is bound to happen, be it an alien invasion or terrorist attack. The moment it happens and position is randomised in a great map.

Of course all buildings and houses will have to be completely furnished. Which is probably a good decade away...

The whole system would have so much replayability. Throw location based weather effects in and an accurate day/night/season cycle along with all that procedurally generated stuff and authentic AI. That is the next gen console, or next gen CPU I'll be fawning over.

Jun 26, 2007, 04:24 PM
yeah, but given that much open-endedness and replay-ability, it would come with a hefty price tag, because there is little incentive to buy other games if you have one game that can essentially do anything.

Jun 26, 2007, 04:52 PM
well it depends what you mean with authentic AI ? after all there will be always problems since concessions have to be made for the player to make it less frustrating

if or example would like to finally see squad tactics worth mentioning...
we still get way to often to see that one soldiers standing there ,after his machine gunner has fallen, firing with his single shot rifle

sure AI characters push more this days but how often do you really get rolled up by an enemy because of such an action (which _isn't_ coming in through a linear/hardcoded scripted event like in way too many games)
i still remember my time during the conscription service we are still so far away it's not even funny

Jun 26, 2007, 04:59 PM
I thought about hat too after watching Die Hard long ago.

I've always thought that it'd be cool to be placed in a position like John McClane where you need to stop these terrorists in this huge building and that your actions will dynamically affect theirs.

A game where you have to make very conscience decisions whether you should take out a foe or not when given an opportunaty, keeping in mind the fact that you have very limited ammo (because you can't just pick up ammo that is lying around the building for no apparent reason) and that if you miss, they may very well kill you, or they may not be alone and that may set off a fatal gun fight.

You have full freedom about the building, but of course the bad guys themselves may keep you from certain areas or not.

Jun 26, 2007, 05:13 PM
Also, I've just got through watching 24 from season one thru season six, and I would absolutely LOVE a good 24 game.

You have a 24 hour period to complete the game and foil some kind of plot. You primarily play as Jack Bauer of course, but maybe at times also play as Cloe at CTU.

It would have a intricate trust system for each major character (both good and bad) as well as a trust system for CTU and the group of bad guys in general. The trust levels will be crucial to keep high, in order to allow Jack to receive important intel. For example, it is good to keep CTU's trust level high in order to call in backup. Several objectives that are issued to you at times will affect these trust levels. There are also some personal objectives that if accomplished, will significantly gain trust for a particular character, but may hinder trust on another or on CTU in general.

When at CTU, the game could play in some ways similar to the terrorist headquarters levels in Splinter Cell: Double Agent, where Jack may have to obtain classified intel without permission.

When outside of CTU, you will have to freedom of GTA, scouring the city of LA, hijacking cars if deemed necessary or in pursuit of a suspect.

Again, depending on your actions will dynamically change character's attitude towards you as well as their actions.

Sorry that was a little long. I have a whole bunch more ideas on this, but I'm going to stop here.

Jun 26, 2007, 06:04 PM
A well designed multiplayer is the best we're going to get in this direction for some time. Coding and hardware aren't there yet...so playing with/against other players will give the best AI.

Plus as cool as a game like that would be, there's something to killing a real player over AI no matter how smart...I'm sure games will get more like this in the future...like an extension of Dead Rising, but I don't care in the end...still would rather lose to a real person in Chess over a Computer.


Jun 26, 2007, 07:05 PM
I never thought of that, a multiplayer game much really work. You have yourself, a bunch of terrorists and some hostages. Maybe some cops on the ground too.

I think if money was involved along the line then people would really play it properly too, without being too silly. Say it costs 5 per game, if you don't die and your "team" wins then you get some kind of reward.

Right now I'm thinking the Die Hard-esque game would be the easiest to make. Simply map out a full building and some surrounding buildings (without insides) for decoration, rather than the alien invasion game which would require a much larger map.
Oh, and I suppose it goes without saying but once the game has loaded there should be no more loading to keep the experience as smooth as possible. So hello, 16gb ram. :D

Jun 26, 2007, 10:15 PM
You should try to create a simple 2D mockup. That would at least give you an idea of what works and what won't work.

Although the hardware resources are probably have reached a point where you can create adaptive AI and non-linear games (we've seen to a certain extent already), no development team is going to invest that amount of time to code it all. And you definitely still have to put some kind of limits to your scope.

Jun 27, 2007, 06:23 AM
I am trying it now but on a much lesser scale, it's more a rethinking of how a non-linear game should work. Offering a single overall goal but countless paths to get there.

Jun 27, 2007, 07:09 AM
I think largely what you want is out there though not all in the same game.


Saints Row has the non-linear story line were you can do the mission mostly in any order. Though some missions open up new missions. A sort of flow on effect.

GTA series has been trying to achieve these lofty goals since the nineties. Similar to Saints Row in that mission can be done in any order it is just that some missions need to be completed to progress the story.

My opinion is that we have the technology. CPU's and GPU's are powerful enough to do this though the money for the development time is not. A game like you would like to see is going to take many man hours and each one cost $. A game developers need to justify risk verse reward. How much money can we spend and how much will we get as a return. How long can we develop this game in a constantly shifting marketplace. Everyday the next big thing is announced.

A new trend in game development over the last 5 years ( or more
) is you buy a game engine then all you need to do the art and a tittle( over simplified I no!). This means that a company gets a return to develop and sell the game engine as well as making the game itself. The Doom 3 engine, Quake 3/4 engine, source engine are all examples of this process.

This could be what you have been waiting for. A company develops a core game engine and builds new or allows new engines to be added depending on the style of game(pure speculation). So for a example company A develops graphics engine #, company B buys engine # and adds their own AI engine to engine #. Thus creating a game which was faster to make cheaper and with better AI.

p.s. I may be well off base :p

Jun 27, 2007, 08:10 AM
I'm hoping that Assassin's Creed will be somewhat like your idea. I hope we will be able to try missions over and over again, but with different ideas and paths each time. One time you may want to go right after the victim, or another time try to climb along the roofs and attack from behind or above. I could see myself playing that for hours, experimenting each time. I really hope they give ample time restraints and freedeom to choose how to complete the mission. From what I have seen and read so far, it appears they are meeting those goals.