2019 Workgroup Topic Proposals

Machine Learning and Adventures in Infinite Space

Two wildly different genres — AAA open world games and mixed reality content games — are circling around different design problems that may share a common solution.

As open worlds get bigger and bigger, they demand huge amounts of time and resources dedicated to level design, art production, narrative production, and testing. Designers restricted to working with traditional tools can only affect small portions of the game, while the sheer number of people (and budgets!) involved present huge risks to studios and publishers. If only there was a way designers to could “duplicate” themselves to affect more of the virtual real estate…

On the other hand, mixed reality games have no control of the space in which they happen (unlike VR or location based entertainment). But they still have to work wherever they are booted up, from the giant living rooms of the American midwest to the tiny studio apartments of New York. If only there was a way to ship a designer with every copy of the game, so they could take maximum advantage of the environment…

One possible answer is to use machine learning to augment the tools and abilities of content creators. Given a system with awareness of the world (e.g. this is a rock, this is a tree, this is a coffee cup, this is a window), it is feasible to train AI so that it develops its own heuristics of content creation based on observing the work of human designers. (this essentially inverts the normal relationship, where programmers and designers create rulesets to generate content)

(for more information on how this might work, please see this excellent 20 minute talk by Sam-Snider Held of Media Monks called Enhancing Human Creativity Through Machine Learning and VR)

(you can also reference this snippet from EA’s Andrew Wilson about using AI to generate stories)

(I would also love to know if anybody is aware of what happened to that initiative)

This is an incredibly complex issue with a potentially terrifying scope, so the group’s first order of business might be to restrict this problem set to a specific medium, genre, or content type. The alternative is to “go wide” and try to establish some sort of creative pillars that developers can look to if they seek to empower digital creatives. It’s also super deluxe exciting and very much The Future.

As an aside, the title of this proposal was certainly inspired by but bears no relation to Strange Adventures in Infinite Space, the delicious lunchbreak roguelike of 2002. Give it a try, if you’re into that sort of thing!