(Some say that the amount of detail needed in game environments is getting bigger, and in the future it will be too time consuming or expensive for designers to make it all, and they should find ways for the software to generate some of the content). One way to create some of the content in RPG / Simulations would be to have the AI characters select it. I haven''t played The Sims, but I reckon that it would be possible to use AI to choose the decor and content of their houses based on various variables. For example, their taste in colours , and these could be modified (as the characters get older) by randomly chosen "fashion".

I''m working on a system whereby layout and decoration style of a certain type of building can be encoded in a 64-bit (say) value. A city of house exteriors can be generated at load time by associating certain bit combinations with architectural models.

When you go in the house it generates the interior based on this value and a fixed value (such as the house''s position). The idea is by modelling a few hundred chairs, roofs, doors, plant pots etc millions of different houses can be generated, even the people inside shouldn''t be too much of an extension (once the AI is in place) by sparing a few bits for appearance (clothes and family size etc would be related to the house style).

********


I am not the Iraqi information minister.

I think the problem is to generate top quality content that can compete against human created content ... plus there are of course some technicall dificulties, but each time you get pass one of those you will almost surely have to lower quality ...

I think that this types of things is good in the technology sense, since we would be needing that someday, if someone wants to put tons and tons of work in something that will work as a reference now, then go ahead ... I wouldnt expect getting but loses in money by doing so (or time) ... however once you have something, if it is well done, it can be improved ... but we are talking about years .... edit: and by that time someone will probably had done some ...

[edited by - eglasius on July 1, 2003 11:28:46 PM]

I think for small developers it''s very interesting to be able to create levels,... by ai.
It''s not a new idea -> think back to the game "elite" - Thousands and millions of planets, solar systems, ... on a computer with 30k memory.

look also @
http://www.amazon.de/exec/obidos/ASIN/1584500581/qid=1057150028/sr=2-2/ref=sr_aps_prod_2_1/302-9682121-5312061

and

http://www.amazon.de/exec/obidos/ASIN/1584502134/qid=1057150069/sr=1-2/ref=sr_1_10_2/302-9682121-5312061

McMc

Elite wasn''t done by AI though, it was done by procedural generation from seed values.

I think there are 2 distinct content generation technologies, with a lot of crossover, that are going to be essential in the next 10 years.

1) Procedural - where a predictable number sequence generator (eg. random number generator) can be used to deterministically create whole worlds (or whatever) from a single number. With multiple players in a single world, CPU power is not so much of a problem as the terrain generation can be performed on the fly on each client. All you need to send is the seed value and the algorithm.

2) Agent-based - where non-player characters powered by some degree of artificial intelligence will form the world based on their own individual rules, and will interact in ways that make this generate interesting results. Given a basic terrain layout, you could generate some primitive tribes which attempt to find a place to live, will establish trade links, generate roads, build housing, and so on.

Now what you could have is to generate the basics on a massive scale using the procedural method, and then use an agent-based method on the results to get the level of believable detail that you require.

[ MSVC Fixes | STL Docs | SDL | Game AI | Sockets | C++ Faq Lite | Boost
Asking Questions | Organising code files | My stuff | Tiny XML | STLPort]