PC Gameplay Benelux Fallout 3 preview

Well I didn't mean to imply neural nets or anything.. they don't have to "learn", you just stack the behaviours in such a way that they appear to learn.

It's simply a series of rules and motivators.. for example you apply a series of behaviours and motivators to the group (raider camp) and a individual level.

So in the raider camp level you have a group parameter of "Cautiousness". It is incremented higher for every death in the group. It degrades over time.. so the group as a whole becomes less cautious over time (they don't worry as much).

As a function of that you have a group behaviour that determines the number of lookouts. This could be ranked based on who has the most sleep (and individual parameter which is a little different for each person so everyone doesn't get exhasted at the same time). You set certain locations as guard points for that camp and they "cluster" around those points when they are "guarding". You still leave individual rules like when to eat and such.

Next thing you know you have a reasonably reactive camp.. you kill a guard or two at night and the next night they double guard and only a few people are sleeping.. over time they get exhasted and also become less cautious. The system feels pretty realistic, but is driven by only a few algorithms.

They stated that they have the AI schedule now use navigation meshes, so they don't have to walk the same exact path to work/eat etc. I think that will also make a big difference since the people will not look like they are on rails hopefully.

I really hope they have made some progress on the behaviour of the AI.. it's a hard thing to get right.

I saw a comment about scripted events. Sure they have their place in an RPG like this, but they really can't replace a good AI simulation system. I think we have yet to see a great execution of a "virtual world" system where groups react fairly realisticly. But I think bethesda is making some headway that direction. It might be silly to see people talk about mud-crabs for the umpteenth time, but it is a step in the right direction IMO.
 
Xenophile said:
But I think bethesda is making some headway that direction. It might be silly to see people talk about mud-crabs for the umpteenth time, but it is a step in the right direction IMO.
Click to expand...
Perhaps it is right direction, but than again, i dont care. I just want to have a good experience playing a game and scripts ensure that characters behave as designed by the authors (% bugs) and can be reasonably tested i believe.

Now if there is a degree of freedom and learning ... well i pity guys who are going to test that and when product ships npc will act stupid anywas becouse they learnt sth new ... that is not the right direction for computer games (imo).

And maybe bethsoft was innovative with their radiant ai thing, but why am i to be a lab rat its tested on? and pay for it?
 
kyle said:
Now if there is a degree of freedom and learning ... well i pity guys who are going to test that and when product ships npc will act stupid anywas becouse they learnt sth new ... that is not the right direction for computer games (imo).
Click to expand...

Part of my point is that they don't "learn" anything.. "Learning" AI has it's place.. but for the world simulation, IE. people guarding, going to lunch, and sitting around chatting about mudcrabs. This is all doable in a believable way through a series of motivators and behaviours. Ok.. I know they didn't do it well in Oblivion.. but my point is I believe it when they say their primary problem was a configuration problem (Some of it was also in the animation department, where they didn't show the internal actions that were happening).

Yes.. it is hard to test because you may have behaviours and switches that either react to each other in an unexpected ways or can be misconfigured. Quite frankly the feeling I get with oblivion is that they had a VERY complex series of behaviours before they were getting close to a hard ship date. I think they were having problems with people reacting wierdly and they ended up just dialing all the stuff back and removing alot of behaviour because they didn't have the time to get it to behave realisticly. So what shipped was a minimal set of behaviors.. this is also why when they had promo videos before launch, it looked more impressive.

I am hoping that they just took the time since they didn't have to rebuild the system to get the all the motivators and behaviors to mesh better so they don't have to dial back the AI.

EDIT:

I think this points to one lesson I feel they have learned this time. Oblivion was rushed. They didn't have time to make the systems work as they were intended. I do think they could have handled this better with real post-lauch development work (I haven't played any of the expansions so I don't know if they fixed much). But at least now with Fallout the impression is that they are taking a longer window to polish and ensure that the underlying systems are working better rather than rush it out in august.
 
Reminds me of "Trapper Dan" (or whatever his name was) from Fallout 2 - the guy with the key to the rat-infested part of Trapper Town. In order to get the key from him, you could (off the top of my head):

1) kill him and take it.
2) pick his pocket.
3) simply ask for it, if you had the right stats.
4) buy it from him.
5) sleep with him to get it, if you were female.

There were other ways too, that I can't specifically remember at the moment. You could also bypass him by picking the lock to the door (or doing any of a number of other things that resulted in the door being open), or obtaining a key in a completely different way.

A game that has over ten years of technological advantage over Fallout 2 and purports to be a vast improvement over it ought to increase the complexity of encounters by a substantial amount. I doubt it will, though.
 
UniversalWolf said:
A game that has over ten years of technological advantage over Fallout 2 and purports to be a vast improvement over it ought to increase the complexity of encounters by a substantial amount. I doubt it will, though.
Click to expand...

It's fucking sad when the best that can be said is "yeah we learned our lesson and have (allegedly) fixed X from Oblivion" yet still the game is a piss poor RPG from the standpoint of having a solid bedrock of RPG gameplay fundamentals. And we are talking about 10 years here. They're regressing .

It's clear where Beth's priorities lie and it's with gimmicky visual bling (see VATS) rather than the meat and potatoes of role-playing design.

Don't let TES' pedigree fool you because for all intents and purposes this company stopped making RPGs with the departure of guys like LeFay and Rolston, and the ascention of the Toddler.
 
Ashmo said:
The technology doesn't matter, the results do.

Beth likes hype and hype lives from technology buzzwords, so it's obvious why they keep on re-inventing the wheel square when plain old scripting could go a long way and yield better results.
Click to expand...
My thoughts exactly.
Jimmeh said:
What might work are network of agents with knowledge database, but i guess it would costs too much, seeing todays budgets.
Click to expand...
I'll just point to Outcast, in case people have forgotten.
 
1.) Go play the game, then. You're missing out on one of the best action/adventure games made.
2.) It did work well, yes.
 
UniversalWolf said:
Reminds me of "Trapper Dan" (or whatever his name was) from Fallout 2 - the guy with the key to the rat-infested part of Trapper Town. In order to get the key from him, you could (off the top of my head):

1) kill him and take it.
2) pick his pocket.
3) simply ask for it, if you had the right stats.
4) buy it from him.
5) sleep with him to get it, if you were female.
Click to expand...

This illustrates my point in a way. Even though we aren't talking about AI, of these options to get the key, only 2 of them are "scripted", the other 3 are simply the application of global rules (steal, kill, barter). The same thing applies to AI and behaviour.. the more you have the global behaviours and systems for each person(robot, mutant), the more it opens your options to have situations react more realisticly and also open up new avenues of gameplay.

EDIT: I'll have to try that outcast game.
 
"yeah we learned our lesson and have (allegedly) fixed X from Oblivion"
Click to expand...

Speaking of which.. this has bothered me for quite a while. Time so spill some blood.

May I ask what is it exactly that Bethesda is still learning? Aren't they supposed to be a professional company? What are they now learning how to make games and we need to hold their hands through this?

How come Interplay didn't need to learn when they released Fallout which was the first game they worked on for many people there? And before some retard goes "oh, loak you think Falluot is prefect, lol youre fanboi!!111", I'm gonna say this: No, you fucking moron, I don't think it's perfect, but it certainly is close to that and even more so when compared with Bethesda's games. And don't even start about the bugs, whoever judges a game based on bugs should be shot. Plus, Bethesda's games have always been full of bugs too.

Anyway, I can understand when a small independent company like Basilisk games says they are learning and they'll improve their next game based on the feedback they received for their first one. But Bethesda... they have quite a few games under their belt, don't they? So when exactly is Bethesda gonna sell a game and not only an attempt?

Of course, the new Bethesda has only two games (soon three), but for each they said they learned from the previous one and they'll improve on that, And yet, each time what we saw were more mistakes, more failed attempts.

No, Bethesda doesn't learn, they don't want to learn and they don't need to learn.

Honestly, I don't really understand how Bethesda has come to get away with these poor excuses. But then again I don't understand how the postal service gets away with their excuses, neither do I understand Keanu Reeves.

But they get away with it... they get away with selling half-games and it's probably because of the hype they generate and the modders. Which is why I hope they don't release an editor for Fallout 3. Maybe then people who stuff the games they buy with dozens of mods from the first playthrough will finally realize that Bethesda isn't able to create games. But I'm not getting my hopes too way up... stupidity may be just too big.
And Pete Hines can take so much advantage of that. Seriously, I believe this made should have the biggest payment at Bethesda because he's selling the games.

Oh, and let's not forget the media. Oh, the awesome media. You know, I still remember when journalists used to write their own articles and not simply copy/paste what the developers say. Because really that's what most articles are today, seasoned with bits of wisdom like "WOW, I CAN'T WAY TO PLAY THAT, IT WILL BE AWESOME".I remember when the journalists weren't all some fucking hippie-don't criticize-see the bright side drooling morons. Then you could have a hearty laugh when a poor game was trashed in a magazine, but now you could hurt the feelings of the developers who worked so hard on the game.

Yeah, Bethesda, keep telling people how hard you work while fat fucks like Ashley Cheng who don't do a thing all day (that's what producer means, really) spews random idiocies on the internet like "I WANNA STARCRAFT FPS".

Good one, Bethesda. Good one.
 
Cutting edge (or next-gen in this case) often means hasn't been tested enough and probably won't work the way it's supposed to in practice.
 
Well, AI is still not a solved problem, I'll allow Bethesda to feel they are still learning in this area. I don't know any games company which has developed a strong AI solution.

AI is a really CPU intensive task, so I'm pretty interested in this AI co-processor someone mentioned. I may investigate it. Be nice if it caught on.

Anyway nuff waffle, while AI is one of the hardest tasks in game dev, given Oblivion, I hold no special hopes for the AI in FO3.
 
You must log in or register to reply here.
Back
Top