Game AI Roundup Week #9 2008: 4 Stories, 6 Quotes, 2 Videos

Novack on March 1, 2008

Saturdays at are dedicated to rounding up Smart Links from around the web. This week, there are many reports and summaries of the AI sessions at GDC 2008, including some interesting interviews.

This roundup is brought to you by Novack and Alex Champandard. If you have any news or tips for next week, be sure to email them in to editors at

Remember there’s a mini-blog over at (RSS) with game AI news from the web as it happens.

Artificial Intelligence Roundtable

The roundtables are often the highlight of the conference for most AI developers. This year’s AI roundtable covered a host of topics related to the development and application of artificial intelligence techniques in video games. Luckily, for those of you who weren’t there, Dave Mark recorded the session and made his notes available online.

Building a Better Battle: HALO 3 AI Objectives

Damian Isla (AI Lead, Bungie Studios) released a powerpoint 2007 version of his slides. (See the forum thread for a download of other formats.) Here’s a description of the session:

“The HALO 3 AI Objectives System turns the techniques used in previous HALO games on their head: rather than a designer ordering groups of NPCs to certain physical areas and then transitioning them to other areas as the battle progresses, H3O allows the designer to specify a set of tasks in the environment and then comes up with an optimal distribution of squads across those tasks.”

Open World Pathfinding

Quinn Dunki, Senior Engineer at Pandemic Studios, also released slides from her presentation on pathfinding in large streaming worlds.

Interview with John Abercrombie on BioShock

You’ll probably hear about this again in the coming days on, but Dave Mark (who had an amazingly productive GDC) put up an audio recording of an interview with John Abercrombie, discussing the AI in Bioshock.

Crowds in Assassin’s Creed

Robert Fermier posted his notes of the Ubisoft session entitled Taming the Mob: Creating Believable Crowds in Assassin’s Creed. James Therien & Sylvain Bernard, Ubisoft. The topic of big teams, mentioned in the Q&A part, was discussed in Tuesday’s discussion also.

Creating a Character in Uncharted

Tara Teich, Lead AI/Gameplay Programmer at LucasArts, also posted her notes about the session entitled Christian Gyrling’s talk on the AI/animation system used for Uncharted.

Call For Papers: ECAI ‘08 Workshop

The Workshop on AI in Games AIG-08 is targeted to both academic researchers and practitioners that actively try to use new technologies in game development. The event continues previous efforts to bridge the gap between the two communities.

From an academic perspective, games are an excellent testbed for many AI fields. Puzzles and board games are typical applications of areas such as heuristic search and learning. Video games contain a large diversity of domains for both single-agent and multi-agent problems. They may or may not include features such as uncertainty, dynamic environments, and hidden information. New ideas in fields such as planning, search and learning can be developed and tested using a game environment as a testbed.

On the practical side, games are a booming, multi-billion dollar industry. Recent years have seen an increasing focus on making the inhabitants of a game world smarter. The quality of a game is directly impacted by character skills such as navigating on a map, acting according to a meaningful plan, and learning from previous experience, which need to be implemented using AI algorithms.

Crytek revisits Crysis

Echoing his lecture at GDC Lyon in December, Cevat Yerli described some of Crytek’s goals in creating the AI for Crysis.

“The challenge, Yerli noted, was achieving human AI and not perfect AI. Perfect AI, anyone can do. Human is a more difficult challenge, because you have to replicate humanity, he said. To illustrate how Crytek overcame this obstacle, he showed a video of North Korean soldiers pouring over a stone fence, with most landing on their feet, but a few tripping and falling.”

An Audience with Soren Johnson

“I think when talking about games in general, it’s important to differentiate between when you’re talking about an abstraction game or a real world game. Because the two get mixed up a lot. When you’re talking about a real world game where you’re moving around some kind of environment, you have an expectation that things will function.

Straight human behavior stuff. You can get a long way with convincing animations and having the guys say smart things at the right time, but for the most part, these are all deeply asymmetrical games (ie, games where the AI isn’t bound by the same rules as you - Ed).

This is why no one is interested in playing symmetrical shooters that aren’t multiplayer. But when you’re talking about these real world games, if it’s an asymmetrical situation… the issue with AI isn’t one of challenge. It’s one of: do they act like people? Do they run away if they’re scared? Do they get stuck on things?”

Narrative Design in Far Cry 2

Robert Fermier again wrote notes about the design of the AI and story in Far Cry 2, in particular this nugget:

“Can’t create levels like in Rainbow 6. Needed a more stochastic way to deploying enemy AI in the game world. One approach: Sims like method, primitive set of needs and desires that the AI can address through affordances scattered in the game world. I need to go on patrol, there’s a patrol route. I need a rest, there’s a break spot to smoke a cigarette.”

The Making Of Dwarf Fortress

Gamasutra posted a great interview about the creation of this ASCII simulation. There aren’t many details about the AI, except on the 8th page:

“The dwarves themselves mostly move around with A*, with the regular old street-distance heuristic. The tricky part is that it can’t really call A* if they don’t know they can get there in advance, or it’ll end up flooding the map and killing the processor.”

A Gentle Introduction to AI

For the beginners among you, Kannan Balakrishnan wrote a introductory article on how to approach AI, also discussing its possible applications like game playing:

“Artificial Intelligence (AI) is a branch of computer science, which tries to give “intelligence to machines”. But the concept of intelligence itself is debatable, and making these machines without life “intelligent” is something near to impossible. But we can say safely that AI aims at producing intelligent behavior from machines. What is the difference between having intelligence and having intelligent behavior? You can exhibit intelligent behavior in a narrow field for some time without really being intelligent. For example a computer playing chess at the master level does not even know that it is playing chess. But for the outsider the view is that it is intelligent like a master. Also we need only this intelligent behavior for many practical purposes.”

Conversational Intelligence

Leslie Spring discusses SILVIA: A Practical Hybrid Approach for Conversational Intelligence. This talk describes a top-down, product development oriented approach to architecting a conversationally intelligent software system. Cognitive Code’s SILVIA platform is used as an example of a system that is representative of this approach.

The SILVIA platform employs a number of known methods within its implementation. The SILVIA platform also incorporates a number of new, unique methods, as well as new applications of known methods. However, independent of the platform’s algorithmic components, the overall SILVIA architecture and implementation was developed with an eye toward rapid product prototyping and development.

As such a system, SILVIA derives many of its elements from disparate development disciplines such as game engines and content management systems. It is proposed that a more general hybrid system not only creates an environment conducive to rapid development and deployment across multiple markets, but also may exhibit the desirable property of emergent behavior via the interaction of its integrated components - Association for Computing Machinery (ACM)

Numenta Game Challenge Winner?

A few months ago, Jeff Hawkins’ new company, Numenta, who specializes in machine learning technology inspired by statistical and connectivist approaches, released on Open Challenge to create a game based on their HTM system. Sadly, the winning game is nothing more than their tech-demo for image recognition, applied on the web, and a timer…

Tabula Rasa Creature AI Changes Video

“One of the more intriguing elements of any massively multiplayer online game is the developer’s ability to change the level of AI within the game to adapt to current trends by the player base. After only a few months on the market, the Tabula Rasa team opted to make several of these changes to the behaviors of common monsters. In the latest update, both the Thrax Pistol Soldier and the Juggernaut (along with several other enemy and ally NPCs) have received substantial updates to their AI.”

Stay tuned next week for more smart links from around the web!

Discussion 0 Comments

If you'd like to add a comment or question on this page, simply log-in to the site. You can create an account from the sign-up page if necessary... It takes less than a minute!