Weekends at AiGameDev.com are dedicated to rounding up smart links from the web relating to artificial intelligence and game development. This week, we have 2 Turing Test competitions, a flash spider and some thoughts on ‘cheating AIs’; also, as always, there are some good articles and blog posts for you to read. Remember, there’s also lots of great content to be found in the forums here! (All you have to do is introduce yourself.) Also don’t forget the Twitter account for random thoughts!
This post is brought to you mostly by Marcos Novacovsky (aka “Novack”). If you have any news or tips for next week, be sure to email them in to editors at AiGameDev.com. Remember there’s a mini-blog over at news.AiGameDev.com (RSS) with game AI news from the web as it happens.
Another approach to a definition of AI
The site Nintendo wii, xbox, psp Video Games, gadgets published an article entitled A Preface for Understanding Artificial Intelligence where -besides other interesting comments-, stats another approach to a definition of AI:
“Although there is no standard definition of Artificial Intelligence and many experts have suggested different definitions relevant to their work on it.We can say that Artificial Intelligence is the study of intelligence that can be generated in the entities that do not inherit it.”
Impressions: “Oblivion” AI (and Modding)
The Niss LiveJournal published an analisys of the Oblivion artificial intelligence.
So, under what circumstances would it be worth making smarter game characters? Some straightforward tweaks like better pathfinding would help, but a lot of what could be called “real AI” has such a high research and hardware-power cost compared to its gameplay value that it’s unlikely to be seen much in games. AI middleware will help with the research-cost problem, but introduce new problems like proprietary licensing fees and not address the other problems. Getting “real” AI in games seems to be most likely to appear in some kind of software toy that’s not even as much a game as Spore — something not motivated by the desire for an action-packed player experience.
Xaitments on Serious Games
The blog Future-Making Serious Games posted a review on the Xaitment’s AI middleware solutions.
AI is one of the most important aspects of any game; yet it often falls short of how humans and objects really act. Humans and objects can learn from their mistakes. They sometimes act irrationally. They forget. Artificial Intelligence in games rarely takes these realistic characteristics into consideration. That’s where xaitment’s technology stands out.
The (Lack of) Human Behavior in Video Games
An excellent article by Christopher Espinal taglined “Introducing economics to video game developers”, was published in the site Nolan Chart, exposing some parallelisms between economics and game’s npc behaviors and reactions.
All of this follows from the idea that economics aims to model human behavior, and software developers ought to understand that. Just as humans respond to incentives, the video gaming characters must do the same. Aiming to defeat a computerized opponent is comparable to the card game of I Declare War, but playing against another human seems closer to a game of Chess - because both are responding to incentives. It is the theoretical mission of video gaming developers to bring the game as close to chess as scientifically possible.
Media comments on Spore
Here there are a couple of transcriptions on the AI side of the media reaction to the recently released EA title Spore.
SPORE Review in A New Way To Play
The Creature Creator is a major breakthrough for artificial intelligence, especially insofar as it is available to everyday users, not just game designers or Pixar animators. It casts the user as the designer — which makes the game virtually boundless. You can’t yet directly play against other people (AI code, not human beings, controls the the monsters in the neighboring village), but you can download other users’ creatures into your game-world. So Spore is the sum of the imaginations of everyone who plays. And as that sum expands, the game will only get better.
SPORE in Webescape’s Weblog
In Space stage you can explore other planets, some of which are created by other users, interact with their creatures and lean more about them. The space phase is vast and you can explore the 500000+ planets in the game’s galaxy. A user-created planet and its AI borrows traits off its creator s style of play: if the creator is aggressive, the AI will be aggressive. You can make allies like in the tribal phase or launch an inter-galactic war. The Space Phase is never ending; you can keep exploring deeper and deeper, and keep interacting with the ever increasing world of Spore.
The blog Predator Art posted an article where its author rants against a lecture he attended at (game design) college where apparently the expositor claimed that “true AI is impossible to create and emotions have no place in games”.
Additionally he mentions that Artificial Intelligence is now referred to as Apparent Intelligence because nobody in 30 years has been able to succesfully simulate all the nuances of a human conciousness. It may seem like we have when you play F.E.A.R. or something, but that AI is still based on a set of rules. What he is talking about is true AI - a perfect simlation of the human brain. And it’s true that that is really damn hard to simulate, but that’s no reason not to try!
AI difficulty and replayability
Some days ago, I found myself playing LotR BfME with some friends again (we used to play against the AI years ago), but we quickly found the original showstopper: the difficulty level is too low. The three of us, playing against five AI armies on “Hard” (top dificulty), becames a sure win after just a couple of games, beeing passed two years after our last.
For a game of about 5 years ago, is notorious that are not the graphics or the general visual impact what turns it limited for replayability, but its AI.
That made me view the following articles in a different context, where on one side a user found some resources of his own to solve the situation, and in the other the developer raises the difficulty level, at risk of beeing accused of “cheating”. After all, we shall really reconsider if the old “cheating AI” should not be incorporated as an option for the player; I’ve never complained about it as long as the game in question remained fun.
How to improve the MOO2 AI
Matt LaCrosse in his blog The Last Orion posted an article about the tricks he use to raise the AI difficulty (aka usefulness) of the Master of Orion II, to make the game more challenging when not in multiplayer mode.
Another way to do this would be to save your first turn from a multiplayer game then load that with AI opponents. The idea is that with stronger race picks the AI will have less of a disadvantage. On Impossible, with good race picks, the AI plays about as good as the average new player. They still make lots of mistakes but it’s better than just using their default race picks.
Top Spin 3
ITP.net published a review by Jason Saundalkar, on the third edition of the tennis game, where he commented about the game’s AI difficulty level:
Straightaway then, you’ll need to know what you’re doing right from the off. If this is your first Top Spin experience, we highly recommend trying the tutorials first and then playing on easy mode because even on medium difficulty, the AI can be quite brutal. (Forget about the higher difficulty levels until you’ve had at least a month of game-time.)
Loebner Prize Chatbots
Although controversial, year after year the competition renews the challenge. This year’s contest will be held at Reading University in the United Kingdom on October 12, 2008.
The Loebner prize in Artificial Intelligence is the first formal implementation of a Turing test with a substantial prize. The ultimate winner of this contest will receive US $100,000 and each year, approximately US $3,000 is given to the creator of the computer program that converses in the most human-like manner.
Following the pattern of the Loebner Prize on Chatbots (above article), B2K donates the prizes for a similar structured competition but aimed at the game AI development: Botprize.
$10,000 value prize! Create the first bot that plays like a human. Top Game Development Studio is providing A$7,000 cash plus a trip to their studio in Canberra for anyone who can create a game-playing bot that passes the “Turing Test for Bots”: to convince a panel of judges that it is actually a human player. The 2008 2K Bot Prize Competition will be part of the, to take place in Perth between 15th and 18th December 2008. Competition entrants will attempt to create a bot that can play like a human player. Details can be found on.
Herman Hults Talks Killzone 2
Killzone 2 is probably one of the most anticipated PS3 games at this moment. Here it is a trascription of what Guerilla’s Games Director Herman Hults said about AI, interviewed by dasgamer.com.
You mentioned improved AI, what have you been able to do differently this time around?
Herman Hulst: AI is an area that we’ve invested heavily in from the first game on. The AI is very clever now. There’s no sort of sitting back, because they’ll pop a grenade at you. They’ll come around. They’ll chase you down. It depends on the enemy type as well, so if you get a commando, for instance, he’ll charge through walls. Each enemy classes has what we call their signature moves. They’re a lot smarter than they used to be.
Gamefest 2008 presentation
Jare our fellow community member commented on the last Roundup about a Gamefest 2008 presentation, Adaptive Tactical Team-AI for Next Generation Bots (thanks Jare). Here it is:
Adaptive Tactical Team-AI for Next Generation Bots Xaitment Presentation from GameFest 2008 Download Page
This talk shows how different state-of-the-art game techniques, such as probabilistic target tracking and dynamic tactical position evaluation, can be easily combined with hierarchic, probabilistic finite state machines to create more realistic team-AI. We also discuss the use of so called “in-game-learning” algorithms and how they can be used to create bots that can adapt to the player, resulting in a more engaging game experience.
Learning Bos Wars game AI
The video shows the work done by people of the Faculty of Science at the Utrecht University, applied to Bos Wars, a futuristic RTS game. Although the learning computer players shown, are not in the regular Bos Wars development branch, is interesting to see the results in the video.
Shows an example of a Bos Wars RTS game (somewhat accelerated) that uses a learning game AI technique. Rather than statically scripted, the red player is controlled by 2apl. It is autonomous in it’s decisions and can learn from past games using reinforcement learning techniques.
You can see it rush-building an army of soldiers. This behavior wasn’t pre-programmed in any way, but learned through trial and error over a couple of games. Actually, it found this specific counter-tactic after 11 played games.
Developed by Eric Kok at the Utrecht University, The Netherlands. More info at http://ekok.nl/tech/bos22apl/
Clint Hocking on Far Cry 2
Rock, Paper, Shotgun published an interview to Clint Hocking, creative director at Ubisoft Montreal, and the lead brain behind Far Cry 2. Here is the transcription of what he said about non-linears story and world.
RPS: Can you tell us a little about how you’ve approached telling a story in a non-linear game world? What kind of problems did you encounter with this during development?
Hocking: To tell a meaningful and engaging story in an open, non-linear world, we decided we needed to tell an open, non-linear story – a story that adapts itself to the player’s actions the same way the non-linear gameplay experience adapts to the player actions. In order to do it, we more or less invented the field of narrative design. Patrick Redding – our narrative designer – is tasked specifically with making sure the story adapts to the player’s actions and decisions. That means that major characters, such as the Warlords, Captains, Lieutenants, and the Buddy characters can live or die based on the actions the player takes and the story will dynamically change based on who lives, who dies, and who the players friends are at any given moment. It was an immensely difficult creative, logistical and technical challenge that required among other things hiring the best writers in the industry to ‘fill in the blanks’ of a truly massive script. We hired Susan O’Connor – one of the writers of Bioshock – to help us bring all these characters to life in the nearly infinite number of permutations possible in the script.
Play With Spider
On his site OneMotion.com, Tomas Eriksson developed a flash game where a customizable real-world looking spider is simulated. Creepy!
An experimental project to make a natural spider in Flash, combining math and graphics. You are welcome to send feedback.
Stay tuned next week for more smart links from around the web!