Arc Raiders’ New Machine-Learning Enemy Looks Absolutely Terrifying
Embark Studios' Arc Raiders is receiving its largest update to date, Frozen Trail, which launches on 8 October and introduces a new machine-learning-powered enemy called the Bully. This follows the game's existing ML-enabled enemy, the Leaper, and represents the studio's ongoing experiment with machine learning as a game design tool to create dynamic, unpredictable combat encounters. The Bully, a ground-based adversary, will relentlessly pursue players whilst adapting to environmental challenges in real-time.
The Bully uses reinforcement learning to navigate terrain and respond to player interactions, including physics-based manipulation through a new tether system. Embark Studios describes machine learning as an emerging technology still in development, noting that the unpredictability it introduces is both a feature and a challenge—the studio cannot fully control the behaviour and must iteratively update models when unexpected issues arise. The studio plans to expand machine-learning implementation across future enemies as the technology matures, with legs-based enemies proving most effective for learning adaptive movement. Frozen Trail will release as a free update across PlayStation 5, Windows PC, and Xbox Series X.
- Arc Raiders' Frozen Trail update adds ML-powered Bully enemy on 8 October
- Machine-learning enemies adapt and stumble unpredictably, creating dynamic gameplay
- Embark Studios considers ML an emerging technology with expanding future potential