GDC Europe

[Acert93]

IIRC, Killzone was panned for its AI, not lauded. Was anyone here when those videos were released a few months before launch? The AI was dumb as dirt and it raised quite a ruckus.

BTW, if any of you find this topic interesting, you might find Bungie's recollection of Halo's AI interesting. They faced the problem of making intelligent and understandable behavior and solved it in Halo, IMO.

I liked Halo's AI, at least the bits I have played.

I think one problem when discussing AI with consumers is that there is a difference between difficulty and solid AI. I can think of some games that by game design were easy, but the AI was fairly intelligent. On the other hand games that were VERY hard but the computer

-Saw everything
-Heard everything
-Had perfect aim at any distance

I remember crouching in the jungle across a lagoon in FarCry and being spotted when I did not even move! And of course their bullets hit me while they were standing and firing

Of course if that makes the game good then the AI is good. Just like a game like HL2 where the NPC are pretty weak and it is designed where you can kill 5 guys at once. Being hard or easy does not make good AI. Having AI that fits the theme and goals of your game design does.

In that regards I would say Halo, FC, and HL2 all have good AI.

The best AI I have seen is in CounterStrike:Source. The AI scales VERY well and has a very human feel to it. On easy they make human mistakes and their aim is off. As you up the level of difficulty you can tell they are paying attention to ques better--just like a human--and have better aim, but they use human like tactics (that work!) and even make mistakes!

I hate SP because AI is frequently very unbalanced and dumb. Dumb as in not working within the constructs of the game design... it "breaks" ... when a NPC goes back to roaming a hall way after spotting you that is lame! But the gems, like CS:S, make it enjoyable at times.

 

[MechanizedDeath]

IIRC, Killzone was panned for its AI, not lauded. Was anyone here when those videos were released a few months before launch? The AI was dumb as dirt and it raised quite a ruckus.

BTW, if any of you find this topic interesting, you might find Bungie's recollection of Halo's AI interesting. They faced the problem of making intelligent and understandable behavior and solved it in Halo, IMO.

I'm not talking about by gamers. There's a reason they gave a presentation on this very topic at GDC. Gamers sometimes don't know anything. PEACE.

 

[Inane_Dork]

I'm not talking about by gamers. There's a reason they gave a presentation on this very topic at GDC. Gamers sometimes don't know anything.

If gamers think the AI is dumb, the AI is bad. It doesn't matter how smart or cutting edge or whatever the AI code or algorithms are if it appears dumb.

Anyway:
http://ps2.ign.com/articles/561/561669p1.html
http://ps2.gamespy.com/playstation-2/killzone/562748p1.html

If there's a reason Guerrilla is making a presentation on AI, I would be glad to hear it. To my knowledge, no one has lauded KZ for its AI. I've only heard critics of it.

 

[Phil]

Sometimes you have to fall short on your goals because technology isn't there yet to be able to benefit at a later time. I think KillZone's AI looks very ambitious and I do remember a while back in an interview that apparently, the framerate problems were related to AI and not the graphics engine - which, now that I look at the presentation doesn't seem far fetched at all. I suppose the AI was too ambitious for this generation to be practical - but perhaps this will prove to be quite a thing come next generation.

The problem usually in development isn't that the AI can't be good - it's that there's no way of knowing which of the billion things the player could do to try to challenge it. When you develop AI, you have a specific behaviour of the player in mind - and if the player tries to play it different, suddenly the AI could be perceived as dumb. It's like playing Metal Gear Solid like a first person shooter - it doesn't work that well. Now if you try playing with the AI by sneaking, hiding etc, you really start to see how gold the enemy behaviour really is.

I guess this is the problem with KillZone if the AI is trying to simulate real combat behaviour. How many gamers and players outthere know how combat behaviour works? The "gamer" isnt' simply a fixed entity... each and every gamer plays his game different and that's where the problem is when judging AI for one.

 

[Acert93]

But on the other hand I don't think you can blame gamers for "not doing what we want".

If there are limitations in the AI then the developers, especially the map/level makers and story writers, need to take that into consideration and design accordingly. AI has always been a limitation, yet the good games learn to cope with those limitations and design gameplay that makes those limitations a design strength.

I would compare it in some ways to rendering engines. The D3 engine would make a poor FarCry game, and the Source engine would make a poor D3 game. Ditto physics. Halo 2 and Half-Life 2 both use Havok, yet their implimentation and the impact on gameplay are completely different.

AI is only as good as the end product. You could have fabulous RTS AI, but it would suck in a FPS, and if you try to plug it into a FPS title you are going to get low marks.

In all fairness AI is hard to make, and I believe it is impossible to make the noobs and the hardcore elite happy at the same time. As a "good" gamer I have learned to look at AI in games in regards to the theme of the game and how it was implimented. Did the game design enhance the experience or not? Did the AI break on a regular basis? I also try not to focus on the AI difficulty (like gun power, aim, etc) because I realize that not all games are created with the same target audiance in mind... and some games are more 1-on-1 and others 1-on-20.

Any game that fits the AI to their design and it "works" as intended is a success. A lot of games exibit broken AI. Either they are too optimistic, aggressive, rush the product, etc... but poor AI can really ruin a game.

But it is hard to please everyone because some people will enjoy aspects of a game's AI and othes will completely dislike it. I know this is common in racing games. Not much devs can do about that less cater to what their audiance wants.

 

[robofunk]

Using anything but what the game considers real combat tactics should get you killed then right?

 

[Phil]

Acert:

But on the other hand I don't think you can blame gamers for "not doing what we want".

I know, and I wasn't trying to. I think the point of my post was more to say "good AI" comes in various forms - either it gives a very satisfying experience or the AI is good from a technological POV. I think this is the case with KillZone, in which the AI is obviously very ambitious (technically) but fails to satisfy (practically) - or simply demand a different approach to the game than the average joe knows.

We already had a very similar debate with Gran Turismo 4 too. Play with the AI and it becomes horrendous, stupid and very predictable - however race the track as you would on a real track and it becomes a lot less obvious and you can tell that the AI at least takes their corners as professional drivers would too. It's all a question of target audience and what you're trying to achieve.

Heck, I've programmed silly TicTacToe (the game you play on paper with the grid of 3 by 3 with X and O and you have to get 3 in a line to win!) AI and at some point it was impossible to beat - certainly this would qualify as very good AI from a technical POV but from a gamers-POV hardly very satisfying. We do play to win after all.

I haven't played KillZone, so I don't know where the limiting factor was (to me), but I do know that in challenging games, the most challenging aspects becomes to figure out how to beat the AI. I will agree though that if the AI varies from brilliant to plain fustrating/stupid - it doesn't add much to the gameplay experience - but it doesn't mean the AI isn't good (technically). Controlling 14 AI enemies all with combat behaviour isn't an easy feat (and at the same time to let it run efficiently without using too much of performance) and as I said, hardware limitations certainly came into the equation at some point as well. Lets hope they can make it more playable come PS3.

 

[Acert93]

Not trying to argue or disagree, just exploring the other direction

Heck, I've programmed silly TicTacToe (the game you play on paper with the grid of 3 by 3 with X and O and you have to get 3 in a line to win!) AI and at some point it was impossible to beat - certainly this would qualify as very good AI from a technical POV but from a gamers-POV hardly very satisfying. We do play to win after all.

Off topic:

Had to share this story. When I was getting married I had to travel about 700 miles to the state my soon to be wife lived with her family. On the way there my truck tire blew out. Unfortunately the spare had rusted to the frame and could not be removed. So we sat in a small dinner waiting for a tire shop to open.

Bored, my best man who was with me asked if I wanted to play tic-tac-toe. I said, "Nah, I cannot lose" He did not believe me so we played for 3 hours!

And he never once beat me... I think I made him go nuts

 

[Shifty Geezer]

It's extremely hard for anyone with some experience of the game to lose or win at noughts-and-crosses. There's only one manoeuevre that guarentees success but that option is lost the moment someone takes the centre posiiton, which invariable the first of second go occupies. As a ameans for developing AI it's got to be a useless exercise as moves are forced to be defensive counter moves to block one's opponent. After three pieces have been placed the outcome of the game is predetermined with block and counter-block. A better AI challenge is 4-in-a-row where you have the chance to set-up future positive moves while still playing defensively.

 

[Phil]

Shifty, Acert:

Yeah I know. When I said good AI, I didn't really mean programming all the IF-case scenarios in which you can pretermine if a win is possible or not and let it run through statically - what I attempted was to program a simple AI that kept track of its game history and started to "learn" based on errors it had made and tried to avoid. There are different approaches like calculating all the possible ways of movements to base the probability of a win or lose situation etc. The challenge at the time wasn't to reach the target of an unbeatable AI on the easiest / quickest way but more to simulate a more "intelligent" and human like behaviour (learn/experienced-based behaviour). TicTacToe / Noughts&Crosses is a brilliant game that for one so simple can already be quite challenging to program "intelligent" AI. It was just an example anyway that I recall to be quite amusing at the time.

 

[Panajev2001a]

Gah.. second attempt at posting because IE just plain doesn't work for me here anymore - as soon as I submit IE just eats 99% CPU and never comes back... Yes, I use IE... I know, I know...

Anyway.

My best "PR Hat" guess as to the source of the numbers is:

1 thread dual-issues a VMX fmad and an FPU fmad
1 thread issues another FPU fmad

That would give 12 ops per cycle making the rather large assumption that such scheduling is actually possible without some form of magic fairy dust.

Note my assumption of 2 FPUs... bit bold but it's the only way I can see to make things work. Assuming 2 VMXs is excessive - it blows the numbers and frankly it would probably blow the yields too if they had to find space for another one of those...

12 ops per cycle is what is required for both sets of numbers - multiply by cores and clock for X360 and add a sprinkling of SPE goodness for PS3.

12 * 3.2G = 38.4
38.4 * 3cores = 115.2

12 * 3.2G = 38.4
8 * 3.2G * 7 = 179.2
38.4 + 179.2 = 217.6 ("218-ish")

We should make marketing people write code if they're so clever....

How about a Gekko style SIMD-mode for the FPU ? (2-way SIMD using the 64 bits FPU)

 

[one]

Sony charges €25,000 for Playstation 3 developer kit
http://www.theinquirer.net/?article=25855

Playstation 3 supports mouse and keyboard
http://www.theinquirer.net/?article=25858

Sony Cell chip has seven out of eight cores running
http://www.theinquirer.net/?article=25860

 

[Dave Baumann]

With the kit you get the documentation and you can imagine what kind of secrets about the hardware itself are in those documents.

Apparently not detailed register level information on RSX, or so I hear.

 

[MrWibble]

Sony charges €25,000 for Playstation 3 developer kit
http://www.theinquirer.net/?article=25855

Playstation 3 supports mouse and keyboard
http://www.theinquirer.net/?article=25858

Sony Cell chip has seven out of eight cores running
http://www.theinquirer.net/?article=25860

That's some of the most poorly written excuses for "journalism" I've ever seen. No new information, a bunch of made up "facts" and veering off into editorial opinion that demonstates a pretty obvious level of bias. Steer clear and wait for someone reputable to report what was actually said at GDCE...

 

[one]

That's some of the most poorly written excuses for "journalism" I've ever seen. No new information, a bunch of made up "facts" and veering off into editorial opinion that demonstates a pretty obvious level of bias. Steer clear and wait for someone reputable to report what was actually said at GDCE...

Haha, the last one "Sony Cell chip has seven out of eight cores running" is news to me!... NOT

 

[MrFloopy]

That's some of the most poorly written excuses for "journalism" I've ever seen. No new information, a bunch of made up "facts" and veering off into editorial opinion that demonstates a pretty obvious level of bias. Steer clear and wait for someone reputable to report what was actually said at GDCE...

Haha, the last one "Sony Cell chip has seven out of eight cores running" is news to me!... NOT

Sounds to me that he's being very sarcastic and quite pissed at there being no new information at this event.

I think you've both missed the point of this being a protest news report.

 

[mckmas8808]

Sounds to me that he's being very sarcastic and quite pissed at there being no new information at this event.

I think you've both missed the point of this being a protest news report.

Sounds like to me he didn't know this information and is being slightly against Sony. Almost like a Xbox fanguy.

 

[MrFloopy]

Sounds like to me he didn't know this information and is being slightly against Sony. Almost like a Xbox fanguy.

really didn't read it that way. Just found it funny almost like he was bored with the info...
but hey, I can't read his mind so maybe.

 

[pc999]

In chess there are predetermined moves and techniques, patterns, that can be used together to form tactics. Every piece has "rules" and in general effective strategies rely on following these patterns. We are at the point where a computer can beat the best chess players--as long as the follow the rules!

But we also are limited by the games rules, we cant do moves that there are not in games, like chess when I think in(e.g.) h2 how to beat a human oponent I think in his range of movements/firing/waiting/etc... they are already there it only changes the way you use it.

Imagine a chess match where you could move 2 pieces at once! or you jumped all the way across the board, or made two moves in a row. Chess is a simple game, yet if you allowed humans to break the rules and do things the computer has not been trained to expect it would get creamed and "look dumb".
That is what happens when you stick humans in a game. If it can be broke--they will break it!

We are talking about a 3D sphere with all kinds of variables. You add physics, guns, running, jumping, doors and halls, vehicles, etc... into the mix and you have hundreds of variables... and you have to code your AI to react "intelligently" to all kinds of scenarios.

Probably that only means that there is no real inteligence in the AI, but I will let this discussion for other time. On the other hand we would need some trully beasts in the CPU, I hope the Cell/XeCPU, can do the job.

So even a really impressive AI can look pretty bad. Broken AI sticks out like a soar thumb. It can be really good... until it breaks, and then you are cursing at it for being so horrible!

So in that regards a lot have stuck to simple scripting. It may not be as dynamic as a fluid situational awareness AI, but on the other hand you can make it do, what you want it to do, when and how you want. And yet it is pretty simple.

Yet, is because of that beacuse that we think the AI is bad...

Killzone's AI was quite clever. The reason that it appeared stupid, imo, is the delay between updates was too great. It was burning up too much cpu time to update more frequently, I suppose.

I dont remember if I played the game, if I do it should be for only a little time, so I cant coment in first hand, but I heard quite a few times bad thinngs about it and only a few good one, I remember some videos of it I thought that AI had did a few nice things, but from videos to games a lot change...but we do have a response.

The problem with game AI in general is that if the player can't understand what is happening, it just seems random, stupid or cheap.

The biggest challenge with complex AI is explaining it to the player in the context of playing the game. It's one of the reasons that the best AI does not equate with the best gameplay experience.

So the player is to stupid to understud the AI, which is trully belliaveble, but that let me afraid if we will ever see good AI, on the other side it seems hat MS and Sony are telling dev that good AI is what gamers want (from GDCEurope, Battlefied slides), I hope dev hear them, which did not happen in this case.

It seems we will have a good time till see good AI, anyway I will try to play the game by myself I am really interested now.

 

[Titanio]

A more indepth article on the Harrison presentation, with some better commentary around PS3 online etc.:

http://www.gamasutra.com/features/20050831/carless_02.shtml

He then notably described the PlayStation 3 as "fundamentally a network platform from the beginning", and continued by describing it as a platform where "consumers easily graze content dynamically, delivered digitally, unlike other consoles [such as the PlayStation 2] where network functionality was an add-on." Thus, although Sony has not yet announced its Xbox Live-like functionality, Harrison's comments make it clear they are taking the challenge much more seriously this time round.

In fact, when challenged, Harrison agreed: "Microsoft has done a lot of things right in this space", and suggested that Xbox Live was "certainly a good model for a lot of the consumer experiences we're doing on PlayStation 3." Although not giving a precise example of something that's planned, he also pointed out that: "maybe there's a button on your TiVo which will spit content onto your PSP", and hinted that this might be the type of interoperability that next-generation Sony consoles might help with.

However, this attitude came in for a little more buffeting in the public Q&A session at the end of the keynote, in which Harrison further reinforced that, technically, "More people have played online on PS2 than a Microsoft format", but admitted "But we did not provide the same experience as they did." He did state again, strongly, that the PlayStation 3 will feature commerce, communications, community features, and media exchange functionality. Some will be free, some will be premium-driven by game services and publishers themselves, but Harrison commented: "We want to provide an open platform as much as possible", and in a notable reference to Microsoft: "Distinct from our competitors, we are happy for publishers to make their own financial agreements directly with consumers."