Preface

On June 17, 2002, our committee on standards for artificial intelligence interfaces (AIISC) was officially launched within the International Game Developers Association (IGDA). The committee's goals are to provide and promote interfaces for basic AI functionality, enabling code recycling and outsourcing thereby, and freeing programmers from low-level AI programming as to assign more resources to sophisticated AI. Standards in this area may also lay grounds for AI hardware components in the long run.

Especially because many institutions from the simulation & military sector are very interested in our initiative, I want to stress right at the beginning that the initiative is directed toward computer games (on PCs as well as console platforms), i.e., our goal is not to provide interfaces for general artificial intelligence applications. While we hope that our work will also be useful for other domains, computer games pose very specific requirements on issues like available computing resources and goals, and we will primarily direct our design decisions on optimizing the interfaces for the computer-game domain.

Preparations for the committee started more or less with a roundtable at GDC 2002 to check the interest in such an initiative. The roundtable was very successful, and I started organizing the establishment of the committee soon after. It turned out to be an unexpectedly huge undertaking; my related e-mail folders contain far over 2000 individual e-mails until now.

The committee has about 65 members and is composed of multiple working groups, which currently work on the following topics: World Interfacing, Steering, Pathfinding, Finite State Machines, Rule-based Systems and Goal-oriented Action Planning. An additional group on Decision Trees has recently been put on hold until we have a stronger force to tackle this topic. Furthermore, we have a support team, mostly composed of students, which do a great job in supporting the working groups with summaries, documentation and so on. Overviews of the work of all these units during the last year can be found in the following sections.

Every working group has a coordinator (and potentially also a co-coordinator) that manages the group and makes sure that the work goes on. The coordinators and myself represent the committee's steering committee, which focuses on organizational and political work, may set up new or cancel existing working groups, and decides on issues like membership approval.

Our work is coordinated via a project page at SourceForge. We mostly use the forum features for discussions of the single working groups as well as the CVS file system for documentation. This style of operation is far from perfect, and SourceForge's forum features did not turn out to be the most convenient option. Despite these problems, the discussions become more intense and useful every day. Not even considering group-internal e-mail traffic, 1777 posts have been posted to our forums as of now, with large increases every month.

Especially in the beginning, we faced many problems with respect to member activity. Many people were very interested in participating but obviously did not evaluate their available temporal resources. In consequence, we already had to replace about one third of the initial members. This issue is still not fully overcome, and we continue the process of replacing inactive members.

Pursuing a standardization goal, one is immediately confronted with fears of impeding innovation, needless bureaucratic overhead and possible market-power related goals. We are well aware of the potential drawbacks that can result from bad interface standards, and also implemented measures to counter potential influence from institutions with their own agenda, e.g., not allowing leading positions of the committee to be filled with middleware representatives. It is to say that we seem to be lucky until now, and we mostly have a very harmonious work atmosphere. Besides pushing the standards issue, the committee discussions provide a great learning experience for everyone.

In comparison to other standards in the area of AI, our approach is clearly oriented toward a specific application. It seems to me that for many other AI standards initiatives, the goal was not to enhance specific applications but to standardize specific techniques that might be of use... somehow. I am honestly not sure whether those approaches are useful at all, and whether many AI techniques today are at a level of general applicability - without a relation to a specific application domain. In contrast, application-oriented standardization initiatives, like OpenGL or DirectX in the graphics domain, have been a huge success and led to a tremendous progress in that area. Our "hands-on" application-oriented approach is also reflected in the committee composition: Nearly half of the committee consists of game developers, the other half being equally distributed between middleware representatives and academics.

Creating a report like this is always a trade-off. We certainly want to keep the public informed of our work and progress. On the other hand, our resources are limited, and we would of course like to direct as much work as possible to the creation of the interface standards. We have thus kept the report relatively short, and would like to invite you to also have a look at the online information of our 2003 GDC roundtable on AI Interface Standards: The Road Ahead. More than a hundred slides and many comments on the discussions are available there. Our next larger presentation will be at GDC 2004, where we hope to present and discuss some concrete drafts of the standards.

Finally, if you are interested in joining the committee, we would be happy to receive your application. Please follow our membership application information on the internet. Unluckily, the page structure of the IGDA's internet presentation is changed now and then. The safest way to locate our pages is thus to use the IGDA's main page at http://www.igda.org as entry point.

I would like to thank everyone who contributed to our committee's progress and this report. The committee is based on voluntary work and we would be nothing without the altruistic support of our members and coordinators. Given the outrageous regular workload in jobs of the game industry, this can hardly be appreciated enough. Special thanks go to Jason Della Rocca, the program director of the IGDA, who probably already has nightmares about e-mails from me coming in faster than he can reply!

I am convinced that - if we really should succeed - our interfaces will represent a giant step for game AI development and the field of artificial intelligence in general. It is certainly not an easy task, but with our great team and excellent support by the IGDA, our chances seem to be pretty good! See you at GDC 2004 for the presentation of our first interface drafts!

Alexander Nareyek
(Committee Chairman)

June 17, 2003
Pittsburgh, PA, USA