NPC AI (ken levine polygon article)

[chosenpredator]

Imagine an RPG. You wander into a village and strike up a conversation with an orc blacksmith. You understand that he has three main passions, which include a devotion to the old gods, a romantic interest in another NPC and an utter loathing of elves.

From this point on, everything you do that addresses the orc's passions affects how he feels about you, and the ways he will interact with you. If you decide to go on an elf murder spree, he may give you bonuses. If you marry an elf, he may try to kill you.

Other orcs in the village, known as "stars" in Levine's demonstration, behave in much the same way. Each has his or her own passions. They are not always aligned.

Essentially, this game is all about worrying over other people's feelings. The way the stars view you shapes how lesser NPCs, like village grunts and guards, view you. They create a kind of macro village feeling about you.

Naturally, there are other villages and other NPCs, so the strategy in this world is going to be all about playing characters off against one another, and picking friends and enemies.

Levine said that this can get really interesting when the player is given emotionally fraught decisions. Perhaps you have the option of marrying a character who has access to great riches, but is kind of a jerk. There is another suitor who is lovely, but offers little. The choice you make lies somewhere between emotional fulfillment and strategic thinking. Of course, you can go back and choose differently in later play sessions.

http://www.polygon.com/2014/3/21/5534258/ken-levine-next-project

Do you think we will see something like this for Sui Generis? I am already greatly impressed by the combat in this game and the similarities with Ultima but i really like what ken said here it's something i really feel more rpg need.

 

[lvk]

Little is known about the implementation of the AI system, but much about the plans for it. I would suggest starting here for a good idea on the plans.

 

[InfamousPotato]

Little is known about the implementation of the AI system, but much about the plans for it. I would suggest starting here for a good idea on the plans.

Huh, after reading that over again I can't help but wonder: What happens when we run out of events? As far as I know, in Sui Generis events happen whether or not you are there or interact with them (which is fantastic). However, if events are always happening independent of you, at some point won't the amount of events (like the one described in the kickstarter post) run out? I get that some things will still happen because the NPCs have wants and needs, but after a certain amount of in-game time (a months, a year), will all events run out, including events/quests in areas that you haven't yet visited?

 

[lvk]

@InfamousPotato I haven't been able to find anything on the longevity of the planned system. I doubt they can give much of it away right now though, since it might all be affected by the main plot which they've been intentionally secretive over. Exploration is an important asset of the game, and if the main plot eventually affects the world at a large scale we'll just have to wait and see.

However, since NPCs will be driven by all kinds of external motivations, I'm guessing new things will happen as the player affects the world.

The instance of the world carries over through multiple characters (unless you start a new one intentionally), so it's likely there are plans in place to keep it going for a while. Due to every player interacting with NPCs in a different way, and your actions having consequences, every player would experience something unique. Maybe your decision to barter rather than shank someone for their goods will affect something way later, and that'd make it very hard for the game to run out of events.

 

[InfamousPotato]

@lvk

You're probably right. We won't really know more until much later when it's revealed. I still wonder how much we will miss because our characters aren't in the right place. Still, having the system that they described definitely makes it worth it.

 

[Oona]

Sounds great!

We'll get a sample of SG's AI on May, I guess, when the Prelude will be released.

 

[Crayfish]

It sounds like Ken Levine is trying to design a procedurally generated soap opera. It's a refreshing idea and is something that is probably inevitable in the long run. I expect that the first few attempts will be severely flawed however, just like the first attempts at interactive movies were flawed. We learn as much from our failures as we do our successes. We need people like Ken Levine who are prepared to attempt things that have a high chance of failure. Otherwise we will end up with endless iterations of the same games we've always played.

 

[lvk]

@Crayfish Are you referring to Ken Levine's game when addressing flaws in first iteration(s), or SG? It's worth nothing that many RPGs have already tried for dynamic NPCs in the past, and both Ken Levine's and SG's systems have plenty of past failures to refer to and build upon. We've already had first attempts at this for a while.

 

[Crayfish]

@Crayfish Are you referring to Ken Levine's game when addressing flaws in first iteration(s), or SG?

It was more of a general statement but mainly addressing Ken Levine's new game and his ambition to procedurally generate emergent stories. As a games designer I've thought hard on many occasions about the problem of whether it is possible to procedurally generate a good story. I've been fascinated by the idea since reading Asimov's short story "Someday" which had a profound effect on the way I thought about computers http://en.wikipedia.org/wiki/Someday_(short_story)

The computer in that story is a child's toy called "Bard" that simply created randomly generated fairy tales. Ever since then I've thought that as an excellent and potentially achievable goal that would tie in with my passion for games programming. It turns out that there's a lot of complexity and "world understanding" needed to randomly generate even a simple story. I think we will get there some day but the first few attempts will almost certainly fall into the "uncanny valley" of having jarring inconsistencies, juxtapositions that a human would spot immediately.

 

[lvk]

I think we will get there some day but the first few attempts will almost certainly fall into the "uncanny valley" of having jarring inconsistencies, juxtapositions that a human would spot immediately.

I'm thinking we're going to get pretty close with SG. I've been slightly skeptical in the past, because every update sounded too good to be true, but BM has delivered on everything they promised so far - they did create very accurate physics and on-the-fly collision models from scratch, the slotless inventory system was implemented as imagined, and we've seen plenty of other cool things outside of the alpha.

When they mention what they have planned for the NPC systems, I'm inclined to believe them. I'm fully aware of how daunting a task it is, as I regularly program myself and wouldn't even think of attempting something like this, but some pretty amazing things have happened here so far that I've yet to see anywhere else.