Romain Trachel and Alexandre Peyrot, machine-learning specialists at Eidos-Sherbrooke, demonstrated the game I just described at Unreal Fest 2022. It combines machine learning with an Unreal Engine feature called the Environment Query System (EQS), which lets developers use spatial data to inform AI decisions.
Normally, this is handled through behavior trees that layer variables and branching possibilities. But in this demo, the AI behavior is driven by a machine-learning model. Unreal EQS acts as the AI’s eyes and ears, providing information about its environment, while the machine-learning model becomes its brain and decides how it should respond.
The game is not as frightful as I made it sound, mostly because of its top-down presentation and placeholder visuals, but its gameplay is a classic cat-and-mouse chase that tasks players with collecting orbs strewn across a map. It’s Pac-Man, basically—but the ghost’s behaviors are no longer scripted.
Machine learning can be used to create a brutal foe. IBM’s Deep Blue and Google’s DeepMind AlphaStar have proven that. However, that isn’t always the desirable—not only because it raises the difficulty, but also because the AI’s specific tactics may run counter to enjoyable gameplay.
Trachel and Peyrot tried using AI for several game modes, including a “multi-output model” that learned to predict the player’s score (earned by collecting orbs) and cut them off.
Matthew SmithBut in this game mode, the enemy tended to camp on the orbs’ positions. It wasn’t fun and engaging to play against, so we didn’t show these results.
The topic of employing AI to improve the gaming experience in single-player mode comes up regularly in discussions related to Civilization, each time with the same conclusion: a full machine learning model would likely make AI opponents much harder to defeat. But this in turn would make the game less enjoyable for the majority of casual players who don’t go up to the highest difficulties – I myself have never gone beyond Emperor difficulty in Civ 6. The benefits would thus be limited to the small set of players who seek to challenge themselves with a Deity level capable of out-strategizing humans – with the considerable downsides of vastly increased computation costs to train and deploy the machine learning model. Another case of ‘not quite there yet’ for the adoption of AI.
Post a Comment