Welcome to our new regular discussion on games AI, written by Dr Tommy Thompson. He’s the creator of the Patreon-funded ‘AI and Games’ series on YouTube and a freelance AI consultant to the games industry. Thompson is an active game AI researcher, senior lecturer in computer science and an independent video game developer. He founded his own company, table flip games, with his first game Sure Footing out on Steam and coming to consoles in 2019.
It’s hard to go a day in the tech-world nowadays without reading a headline about artificial intelligence, with every new breakthrough accompanied by a fresh take on the robot-driven apocalypse it will bring forth. Video games are increasingly prominent in AI news, mostly because they’re learning to play them! Machine learning systems taking on the world’s best at the likes of Dota 2 and StarCraft II isn’t the end of the world as we know it, it’s just the next step of a long and storied relationship between the two industries.
Artificial intelligence is all about building software that is autonomous (thinks for itself) and rational (does the best it can with what it knows) and while in most sectors AI is sought for optimisation or efficiency; the games industry provides one of the most novel applications: entertainment. This ranges from non-player characters in Call of Duty to squads in Ghost Recon: Wildlands, procedurally building worlds in No Man’s Sky or characters in Sea of Thieves. In many of these instances, the `best’ solution is either too subjective to quantify, or just not any fun to play against. In providing an enemy for players, AI is sought more to provide theatre rather than challenge: given it’s no fun if opponents are landing pistol headshots with perfect accuracy.
Over the last 15-20 years `game AI’ has carved its own niche that bridges the two disciplines. This resulted in innovations for combat in games such as the NPCs in F.E.A.R. and the director AI hunting players in Left 4 Dead, all the way to more contemporary releases such as herds of robots in Horizon Zero Dawn and the nemesis system of Middle-Earth: Shadow of War where Uruk’s develop a history with the player. In many cases, these games utilise existing academic research in new and interesting ways.
However, many of these achievements are through ‘classic’ AI tools and methodologies. But as games become bigger, new challenges faced in development are beyond the status quo given their scale and complexity, but they’re problems that machine learning algorithms are well-suited to tackle, given they can learn the solution rather have it programmed by a developer. The boom we’re seeing in the media is evidence of this shift happening in AI research and there are plenty of opportunities where games could benefit from this, be it in character control, procedural tools and content, player analytics and modelling, and in-game balancing – where all those self-learning AI players come in real handy!
Fortunately, the UK continues to be prominent in AI research and development, with a lot of work happening in the gaming sector. Studios such as Horsham-based Creative Assembly invest heavily in AI tools and teams to build award winning titles such as Alien: Isolation and the Total War franchise, while London’s own Bossa Studios are currently exploring AI-driven storytelling. Outside of the final games, there is significant innovation in the underlying tech, with continued development of the TruSkill ranking system for Xbox Live and Drivatar player modelling systems at Microsoft Research in Cambridge. Meanwhile AI and cloud-powered testing helps Rare and Playground Games run large-scale automated tests for the likes of Sea of Thieves and Forza: Horizon. Outside of industry, many UK universities publish at top-ranked journals and conferences worldwide, using games as a testbed for AI research and often bringing innovations to commercial games through collaboration with AAA studios.
AI is now a huge industry in and of itself and its impact on the games industry will become even more significant in the coming years and we’re going to help you get up to speed! We’ll look at the big trends that took shape in the games industry over 10 years ago, the academic research communities that arose in tandem, the innovations happening the industry now and where it might all wound up in the coming years.