Vapourware The Future of Modern Game Development Is Procedural And AI-Driven | Musing of a Chopped Lamb

[a cut of domestic sheep prime]

Somewhat related video that got me thinking about this again:


http://forum.outerra.com/index.php?topic=637.0
(link in case you want this hiking simulator to end all hiking simulators)

So for a while now, we've been near the point now to where we can generate the entire planet and many existing structures inside a video game using publicly available data - Google Earth, for example, though that's the tip of the iceburg. There is enough data on us trees, animals, plants etc to populate every species of bird in its correct location and with accurate migration patterns. You could populate redwoods where redwoods grow, firs where firs grow, an accurate map of US highways, you could auto-generate NPCs with correct population density and accurate demographics.

The main difficulties are related to content. Structures at this point for example: Lots of data stored in different formats and to different quality standards. You can generate an entire city relatively easily, but making it look good and have things like buildings, homes, highways and overpasses show up correctly is another matter. Then there's the insides of those structures.

How can you create a high quality model/mesh for so many objects? We don't have the manpower/money to continue the development balloon that has seen game companies go from 12 man teams creating sprawling games for 3 million dollars to 500 person teams creating smaller-scoped games for 50 million+ dollars. We need machines to fill in the gaps. How/when that is going to be accomplished is anyone's guess though and I haven't been keeping track of it.

*ahem* all this to say that the most complex game could become a reality soon: ie, the entire planet earth with virtually every city explorable. [Vapourware]


It'd actually be best used as a platform for future games - create the engine and setting and then license it out to developers to make their games in. Imagine a superhero game where the Hulk can smash a real New York or a GTA-style game that could span the entire to-scale map of the US.

Smaller-scale games could also benefit from from such a game base. Game creators could focus on telling their stories rather than securing some astronomical budget for simply creating the world.

Anyway, just wanted to get this idea out. I know it's nothing revolutionary, but I'd like to know what you men know about his sort of thing. Also, if I created an indiegogo/kickstarter scam for this, would you gives monies to it? Lemmme know.

 

[Leechmonger]

you could auto-generate NPCs with correct population density and accurate demographics.

You should brace yourself for accusations of racism if you do.

 

[a cut of domestic sheep prime]

you could auto-generate NPCs with correct population density and accurate demographics.

You should brace yourself for accusations of racism if you do.

"Oh, Diversity, no! Let me just change the numbers a bit...aaaand all the whites are gone."

 

[Makabb]

This would be ideal for a flight sim

 

[PorkyThePaladin]

I have felt for a long time that the true potential of gaming is in procedural generation and content (which includes AI). That approach brings certain advantages that cannot be matched by static development. It can generate endless content, it can dramatically improve the depth of gameplay, it can change the very quality of gameplay (for example by allowing perma-death and realistic player behavior due to that).

However, I think you are looking at the wrong thing (that outerra project). Dwarf Fortress is a much better realistic example of what you are talking about. Generating a huge model of real Earth is nice and all, but miles of mostly empty land and some plants/animals aren't enough to carry a game. I am a huge defender of realism in games, but that's mostly because they are so unrealistic right now. I think it would be a bad idea to go for full realism. The real world is mostly empty. Even with 7 billion people, if you ever drive outside the cities, it's mostly just empty land and boring things. To support an entertaining game experience, you need to hit the sweet-spot between 1-to-1 scale with real world and bethesda's/bioware's run-into-something-cool-every-2-steps. And you need to figure out how to create interesting procedural content (quests/interactions/events/NPCs) that will actually carry the game. This is what games like Dwarf Fortress and Ultima Ratio Regum are doing right now, but because they have simple graphics, people tend to overlook them. But in reality, they are doing truly revolutionary stuff when it comes to procedural gaming.

 

[NeoKino]

Yah! more generic open world sandbox games without any depth!

 

[a cut of domestic sheep prime]

Yah! more generic open world sandbox games without any depth!

No, the idea is to take the work of creating such sandbox games away and let developers focus on other things. Ever since games turned into voice-acted pseudo-movies with more and more detailed graphics being the top priority. Or would you rather developers spend all their time designing lame hubcities that have to be small because designing a GTA city would eat up too much of their budget?

"AI" (or whatver that passes for AI that can do the job) and procedural generation solve that. Daggerfall dwarfed Skyrim and did so with a team/budget a fraction of the size. It was also filled with generic crap copied everywhere. But with a mix of procedural generation and AI design, this problem could be overcome.

And there's already been a lot of progress. Back then every NPC was one of a handful of hand-drawn portraits copied over and over. Then we moved on to randomized (but ugly) model morphs in the Sims and Oblivion. Now you can auto-generate 1000s of different NPC with different faces and constrain the variables in such a way as to not make them look hideous.

The important things are what I want devs to focus on - story, scripting, characters. I want whatever technology that can let them focus on that instead of making a graphical world that could really be ascii for all I care, but that normies want to be at least Witcher 3 quality.

However, I think you are looking at the wrong thing (that outerra project). Dwarf Fortress is a much better realistic example of what you are talking about.

It is if you want a comparatively tiny world that isn't modeled after Earth. Again though, this was just a video that got me thinking about it again. Their planet was generated via real, publicly available elevation data.

it's mostly just empty land and boring things

You could say that about areas of Skyrim too - albeit on a smaller scale. And? Then the player can build on it or wander through it if they want. Or ignore it. Or maybe a developer will want to put something there or stage a battle there. Maybe a modder will.

It won't matter because an inordinate amount of time wasn't spent on making it. It won't be a handcrafted area that someone spent a lot of time on, yet value can still possibly be derived from it, should the player or the developer want, depending on the type of game.

With current technology and development methods, if a developer wants to create a spy thriller spanning multiple cities, they need to spend a huge amount of their budget on graphics and area design and you can bet that those cities aren't going to be anywhere near the size of a GTA city - let alone a real one.

Now imagine that they didn't need to do that. Imagine all the cities were done, there was already a functional platform for people to drive in or do whatever. All the developer has to do is drop in the elements of their game into the existing city. No handcrafting of hub cities like in Deus Ex: MD is needed.

If you want a scene where you meet outside the Whitehouse, it's done. Just write it, script it and voice act it. You don't need to pay a team of designers to model it - unless you want it to be a "custom metal whitehouse in the future" or a whitehouse that's been hit by a nuke or something.

To support an entertaining game experience, you need to hit the sweet-spot between 1-to-1 scale with real world and bethesda's/bioware's run-into-something-cool-every-2-steps.

I disagree. All you have to do is place content in such a way that that happens anyway.

You can still have a game that takes place mostly in the total area of a few city blocks if you want. I don't want people to have to really drive for 30 minutes or take a 23 hour flight to China in real-time. But all you need to do is implement fast travel.

Space games like Frontier did this. You had a massive galaxy with millions of stars, but the actual game was confined to a few hundred systems and usually on and near planets and not in the depths of space. The majority of Dwarf Fortress takes place in generated dungeons in adventure mode or else player created dungeons dug out into random areas of a massive procedural generated world. The world is still there for those that want it, but no players are going to explore every inch of it.

All that said, the same technology that would let you generate a planet full of cities could let you generate one city, or a custom sized planet. So even if you still disagree that a 1:1 planet would be a cool platform for games to be made on, any company that developed these technologies could conceivably be custom world designers, procedural generating areas of any size for game developers so that they don't need to spend the time and money doing it. The advantage to a real earth is that much of the data to make it exists an is publicly available already or will be soon.

 

[Rabbid]

This would be ideal for a flight sim

Google earth had this feature (probably still has). You could fly a plane anywhere you wanted. It was like watching paint dry, but was entertaining in some weird way.

Check out this "game" as well: http://www.gefs-online.com/