Hard Light Productions Forums
Off-Topic Discussion => Gaming Discussion => Topic started by: BloodEagle on April 27, 2013, 06:35:08 pm
-
http://www.pcgamer.com/2013/04/22/x-coms-creator-julian-gollop-on-modern-blockbuster-game-development-overdesigned-uninteractive-paper-thin-illusions/
In a genial interview, strategy-game elder statesman and creator of the original X-COM Julian Gollop talked to us about his imagined alternate history of gaming, his preference for procedural systems, and how he feels modern games have abandoned the promise of advanced AI in favour of shinier visuals and reward mechanics designed to massage players’ egos.
It's way too short, but it's nice to hear the opinions of someone who's been involved in some of the more interesting parts of the industry.
-
I agree with the sentiment that games lack "soul" today. Important things (Such as AI and gameplay being at the forefront) are ignored in favor of cheap
drug graphic based games as the natural result of the oppressive and destructive capitalization of gaming. Which is why we should improve the FSO AI
-
I agree with the sentiment that games lack "soul" today. Important things (Such as AI and gameplay being at the forefront) are ignored in favor of cheap drug graphic based games as the natural result of the oppressive and destructive capitalization of gaming. Which is why we should improve the FSO AI
Exactly, and I hate it. Urrrrgggghhhhh!!! :mad:
:lol:
AI has regressed badly. When's the last time you saw anything praised for it's AI? Mainstreaming strips games of soul and personality.
-
Er... how is AI any worse than it was in the '90s?
-
Er... how is AI any worse than it was in the '90s?
FEAR had the best AI, and it was super simple and elegant.
-
Er... how is AI any worse than it was in the '90s?
FEAR had the best AI, and it was super simple and elegant.
Wasn't it just nothing-special code propped up by excellent level design and enemy-announced scripted ambushes/'reinforcements'?
...abandoned the promise of advanced AI
What promise? I remember 90s gaming being just as focused on bigger, better, more shiny graphics as it is now.
-
The code was special in that it did precisely what it was supposed to do in concert with the audio and level design. Game AI is nothing like 'real' software AI; it's a set of parlor tricks to provide the illusion of intelligence and the ability to surprise and challenge the player. FEAR worked so well because it built a set of apparently intelligent NPC actions out of a few very efficient, basic behaviors supported by other aspects of the design.
They immediately ****ed it up for the dreadful sequel.
-
FEAR's AI was interesting in that it had a goals oriented approach to performing the next action. It would determine what it needed to do then decide on a series of individual actions that would achieve it.
http://web.media.mit.edu/~jorkin/gdc2006_orkin_jeff_fear.pdf
-
I'm pretty sure that the point he was trying to make was that almost no one invests in (Game) AI when making games nowadays; as opposed to back then, when AI was definitely improving over time.
Of course, that could have more to do with Mechanics Regression and / or Reinventing The Wheel than any particular distaste for spending more resources on AI.
-
Galactic Civilizations II had some impressive AI. Half-Life 2: Episode 2 had some really solid AI as well. XCOM: Enemy Unknown's AI is fairly competent as well.
Can't really think of any others that haven't been mentioned.
-
The problem is not that developers/producers (put the blame on who you wish) put the importance on shiner graphics as opposed to more in depth AI and gameplay.
The problem is that it's what the masses of consumers demand.
As long as they demand it, they'll keep making more. Consumers are the problem. And I don't think there is a solution to this problem.
-
This is going to be a "HURR DURR Old games were so much harder new games are for noobs" thing, isn't it.
EDIT:
The thing is, I don't really think we need "better" AI. It is trivially easy to set up AI code so that it will give players an unbeatable challenge (Also known as "Cheating in the AIs favour"). What we need is more "natural" AI. The problem is that as graphics and sound and storytelling have advanced to make games much more real, the deficiencies of the AI become that much clearer. In an environment where we can expect things to feel real, moments where an AI opponent will do something clearly nonsensical (Like, say, storming in front of the player's gun barrel) will stick out like a sore thumb.
Of course, once you try to solve this problem, you'll get to another problem, namely "How do I make this a fun game again, now that the AI can kick a given human's ass pretty much all the time". Sometimes (See XCOM), that sort of thing is expected and taken into account by the fanbase. But elsewhere? I am not so sure if "Make games hard again" is really a good thing to do.
-
More in depth gameplay != harder
-
Consumers are the problem. And I don't think there is a solution to this problem.
In this case, all the hope is in indie game devs, who are "doing it for the art" not to make big bucks out of it. These days, that's where you get epic stories, memorable characters and unique gameplay. While most of the best games I know are old classics, I've also played a quite a few indie games that are up to the old standards (and what I consider my favorite series, ArmA, is actually getting better with each subsequent installment). My biggest complaint about many recent games is multiplayer focus. I'd rather have a good story, good AI and solid SP gameplay than MP balance. Making the game primarily focused on multi means the devs can weasel out of developing AI and writing a good story, and those (especially the latter) is what I'm looking for in a game.
Also, I really hate all those premiums and all that's designed to milk the player out of his money. They're there, because most people fall for it, but I prefer to pay for the game once. I only get DLCs if they actually add to the story.[/rant]
More in depth gameplay != harder
Pretty much true. Though in many cases, more in depth gameplay means the game may be easy to get into, but harder to master. Few games managed to pull off complex gameplay with a completely smooth learning curve. Also, since unlike in SP, in MP you get thrown into the game without much preparation, and many people play games for that, gameplay depth can put people off or never be experienced (if mostly confined to SP).
-
Well I know from working at Ubisoft they have hundreds upon hundreds working on Assassin’s Creed – more than 400. Assassin’s Creed III is absolute bare minimum 600 people, probably, were working on it for most of the time worldwide across many studios.”
This makes me feel pretty good at the stuff we produce here at HLP with the numbers we have.
-
More in depth gameplay != harder
It's also not equal to complex, either, which is another common mistake. The simplest games can have an incredible wealth of strategy and depth, and the most complex can easily just collapse to a single strategy which completely dominates the metagame.
-
For me it's not about making games harder, ideal AI is human-like AI. AI that will match and surprise you, not AI that has the intellect of a chessmaster, instant reflexes and perfect attacks.
Though I do like a hardcore type difficulty to be included in a game, just so the limits can be taken off the AI.
Someone mentioned Galactic Civilisations II, I remember that being praised for it's AI myself now. I have the game, but I've never been able to get into it to find out, but I remember reading some big articles on the game's AI, which is what made me buy it in the first place.
Some of today's game AI's seem either so dumb that they can barely even pose a threat to you at all, or have no more complexity to them than to simply charge at you. Or shoot at you when you come into range and not do anything else. Just dumbed down or underdeveloped so much. Play something like Streets of Rage 2 and crank the difficulty up on that and the enemy AI is pretty damn good and challenging. That game has surprised me on more than one occasion after playing some more modern game with inferior AI.
AI doesn't need to be complex, people praise the AI of the pacman ghosts a lot, but when it comes down to it, it's very simple, giving an illusion of intelligence. If you are under an illusion, it's just as effective as a complex AI. It makes you feel the same way about the in-game AI.
-
TbH, the AI in the article you're probably thinking of (http://www.computerandvideogames.com/161570/blog/galciv-2-war-report-final-entry/?site=pcg) was pretty much an accidental parlour trick. The devs have said that they never programmed anything capable of that sort of high-level inference; it was all being projected onto the AI behaviour by the author.
-
I guess I'm showing my age in that when I hear the name Julian Gollop, I think of Rebelstar Raiders, Laser Squad and Lords of Chaos.
-
TbH, the AI in the article you're probably thinking of (http://www.computerandvideogames.com/161570/blog/galciv-2-war-report-final-entry/?site=pcg) was pretty much an accidental parlour trick. The devs have said that they never programmed anything capable of that sort of high-level inference; it was all being projected onto the AI behaviour by the author.
I have read that one before, but not before I got the game, as the expansion wasn't out.
-
This makes me feel pretty good at the stuff we produce here at HLP with the numbers we have.
Yeah but Ubisoft and other studios have to be able to ship a game by a certain date. I'm sure Assassin's Creed could have been done with a lot less people but when the ship date approached, they likely hired a lot of temps to make up for the backlog of tasks and assets they needed completed.
-
Debugging :shaking:
-
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
-
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
I don't think that's true - Bioshock Infinite wasn't a very good game.
-
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
(http://upload.wikimedia.org/wikipedia/en/0/0e/DukeNukemForever.jpg)
You were saying?
Time does not mean anything if the game does not have a "soul." If a game does not try to add new things to the chessboard, it's just another clone. You can make games that don't have a multi-million dollar budget that are better than 90% of the crap you'll find in the industry: Example (http://www.moddb.com/mods/angels-fall-first-planetstorm) Here's a more up to date gameplay vid of that game: http://www.youtube.com/watch?v=tOvgLQspt6k
-
back to the FEAR AI thing, it never really struck me as being any different than other AI. except maybe that they ran away more.
-
The paper someone linked on the last page explains the distinguishing features in some detail.
-
This makes me feel pretty good at the stuff we produce here at HLP with the numbers we have.
Yeah but Ubisoft and other studios have to be able to ship a game by a certain date. I'm sure Assassin's Creed could have been done with a lot less people but when the ship date approached, they likely hired a lot of temps to make up for the backlog of tasks and assets they needed completed.
I'm just sayin' ya know
If I could clone myself 9 times (and my pc), WoD would have been done in two months :p
-
the mythical spoon-month
-
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
I don't think that's true - Bioshock Infinite wasn't a very good game.
I find it difficult to believe that it could have been better if they cranked it out in a year's time. It's more likely that the good parts of it would have been knee-capped leaving nothing good whatsoever, in that scenario.
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
(http://upload.wikimedia.org/wikipedia/en/0/0e/DukeNukemForever.jpg)
You were saying?
I think the main problem there was the multiple studios passing it around, rather than the development time itself (which was much longer in this case than what I'm talking about). Also, that's not one game -- it's a Frankenstein's Monster of a bunch of games.
-
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
I don't think that's true - Bioshock Infinite wasn't a very good game.
I find it difficult to believe that it could have been better if they cranked it out in a year's time. It's more likely that the good parts of it would have been knee-capped leaving nothing good whatsoever, in that scenario.
It would've been better if they'd made good design decisions, put them together into a good design document, and stuck to it. The reason it took so long is because they kept building huge chunks of stuff and then throwing them out.
In general my experience as both a consumer and a 'developer' (lol) has been this: if it's taking more than three years something has gone very wrong.
-
The funny thing is, if they were willing to produce these over a longer period of time, the costs would be substantially lower and the end product would be infinitely better.
(http://upload.wikimedia.org/wikipedia/en/0/0e/DukeNukemForever.jpg)
You were saying?
Time does not mean anything if the game does not have a "soul." If a game does not try to add new things to the chessboard, it's just another clone.
And if it adds/changes to much, it is BETRAYL.
No matter how much you muck with the formula, you'll never make everyone happy, because everyone will have their own idea of how the game "should" be.
And speaking of which, I now some haters will get their panties in a twist for me saying this, but when it comes to AI, SOTS 1/2 is an example of a good non-cheating AI.
-
SOTS 1
Yes, definit-
/2 is an example of a good non-cheating AI.
+ Fixed an issue where AI ships could fly into planets when performing fly-by maneuvers.
+ AI no longer retires ships constantly.
+ It is no longer possible to scrap enemy ships.
(http://i.imgur.com/KQ3XX.jpg)
nevar 4get
-
Well unless you're claiming that the game crashes if the player is doing too well, you haven't actually refuted Trashman's point. :p
-
the mythical spoon-month
Meaning?
SOTS 1
Yes, definit-
/2 is an example of a good non-cheating AI.
+ Fixed an issue where AI ships could fly into planets when performing fly-by maneuvers.
+ AI no longer retires ships constantly.
+ It is no longer possible to scrap enemy ships.
(http://i.imgur.com/KQ3XX.jpg)
nevar 4get
:lol: :lol: :lol:
-
Well unless you're claiming that the game crashes if the player is doing too well, you haven't actually refuted Trashman's point. :p
The game absolutely crashed when the player was doing too well, as well as when the player was not doing too well, or anything in between
(but note the patchnotes as well)
e: am i dyslexic
-
the mythical spoon-month
Meaning?
It's a reference to this book (http://en.wikipedia.org/wiki/The_Mythical_Man-Month).
-
This is going to be a "HURR DURR Old games were so much harder new games are for noobs" thing, isn't it.
EDIT:
The thing is, I don't really think we need "better" AI. It is trivially easy to set up AI code so that it will give players an unbeatable challenge (Also known as "Cheating in the AIs favour"). What we need is more "natural" AI. The problem is that as graphics and sound and storytelling have advanced to make games much more real, the deficiencies of the AI become that much clearer. In an environment where we can expect things to feel real, moments where an AI opponent will do something clearly nonsensical (Like, say, storming in front of the player's gun barrel) will stick out like a sore thumb.
Of course, once you try to solve this problem, you'll get to another problem, namely "How do I make this a fun game again, now that the AI can kick a given human's ass pretty much all the time". Sometimes (See XCOM), that sort of thing is expected and taken into account by the fanbase. But elsewhere? I am not so sure if "Make games hard again" is really a good thing to do.
I think what most of us would like is AI capable of problem-solving without cheating. Strategy games are great examples of this - SC2 (to choose a popular example) actively allows the AI to cheat to produce harder levels of difficulty. Without those cheating methods, AI is incapable out out-foxing a human opponent. Actually, this is one of my major gripes with RTS games in general - they often come down to efficiency and build order, versus any tactical or strategic-level thinking.
I agree that natural AI is the goal, but that requires the AI to actually be able to perform limited problem-solving; or for scripted sequences that trick the player into thinking the AI is problem-solving. All the games that get praise for the depth of their AI did exactly that.
Harder is not "better." Challenge is. It's easy to make a game harder - think of the lethal accuracy and damage ME3 MP's enemies put out compared to players - but it's not easy to provide a fair challenge. Anyone else tired of AI with ESP and auto-aim? I sure am. That said, it's also trivial to defeat because there has to be enough error built in to allow a human to win. Some balance is necessary.
I'm playing a lot of older games at the moment, precisely because their gameplay is often superior to modern games they inspired with better visual but less depth.
-
Actually, this is one of my major gripes with RTS games in general - they often come down to efficiency and build order, versus any tactical or strategic-level thinking.
WALB :toot:
-
I think what most of us would like is AI capable of problem-solving without cheating. Strategy games are great examples of this - SC2 (to choose a popular example) actively allows the AI to cheat to produce harder levels of difficulty. Without those cheating methods, AI is incapable out out-foxing a human opponent. Actually, this is one of my major gripes with RTS games in general - they often come down to efficiency and build order, versus any tactical or strategic-level thinking.
this is why i play RTS's (ok, the only ones i have are C&C) on easier difficulties, and i make it fun by setting personal "challenges" of defeating the AI in novel ways, or executing "perfect" tactics.
-
Actually, this is one of my major gripes with RTS games in general - they often come down to efficiency and build order, versus any tactical or strategic-level thinking.
WALB :toot:
Going to have to forgive me for not knowing that acronym...
-
Wargame: AirLand Battle.
-
Wargame: AirLand Battle.
We All Love Battuta
-
Also that, although I believe that game has some rather bad writing, truly atrocious gameplay, and a really weird reward loop.