Generally, Artificial Intelligence (AI) may be defined as that intelligence exhibited by machines. In the research field of computer science, it is defined as the study of ‘intelligent agents’; namely, any device that perceives its environment and takes actions that maximises its chance of success at a particular goal.
By Mitchell Duncan, Technical Analyst
In more general terms those outside academia might define AI as when a machine exhibits mimics ‘cognitive’ functions that humans associate with other human minds, such as ‘learning’ and ‘problem solving’.
However, in the gaming world, the concept of AI is looked upon with some ambivalence, since to some developers, AI in gaming has little to nothing to do with the advanced AI techniques that come out of academic and industrial organisations. It is because game AI has wholly different challenges to those of AI at large.
Outside the games world, advanced AI development isn’t about creating AI that is “barely” intelligent, nor is it typically hinged on authoring an entertainment experience for an intended user. Some branches of AI attempt to create systems that closely mimic true biological intelligence, the focus being on computers systems which learn to perform tasks without being explicitly programmed to perform those tasks.
For game developers it is all about the player experience. AI needs to augment the game and operate in such a way as to convey an experience that provides for fun, drama or some form of enjoyment. Frequently this translates to the game following a tightly-controlled, authored set of rules and a script, with a few selected ‘random’ events injected for simulated actions. This means that most game AI is designed to be exactly as intelligent as it needs to be in order to give the player a good experience, which means the AI should behave in predictable and barely intelligent ways.
AI is only slowly being adopted in to the gaming world, the main reason is that implementing concepts such as machine learning into games isn’t simple, and is often not desirable from game the production and game design perspective. Game developers are able to get good behaviour from machine learning techniques, but they not be the behaviours wanted for the game experience they seek. Basically, no-one wants to be crushed by a smart game all the time !
Some in the gaming industry see AI, and machine learning in particular, as the next frontier in gaming. They see it not as ‘man vs machine’ but more like ‘man and machine vs machine’, where the human players get their own robot adviser to boost their own playing skillset.
In an AI augmented game, the AI agents could collect player data dynamically, then autonomously create new content or alter the gaming environment to fit the individual or player group’s play style. This approach would significantly change the players’ experience, while keeping a game fresh, and alleviate the developer’s pressure to constantly create all-new content.
So, the future scenario for gaming might be that each player (or player’s team) may be able to use an AI agent as an ‘adviser’ to prepare a better strategy within the game. Of course other teams may do the same! Further, the game itself may learn from all the players so that the game may then ‘evolve’ to a create elements that can be used elsewhere within the game. This could result in the AI agent creating scenarios and artifacts not previously imagined by the developers.
The result might be a game that even learns to beat its own makers !!
