AI is only getting better and better in games as time goes on, and EA has shown off the results of some of its growing technology. The company just released a video showing off what AI can do for skill-based play, using its popular shooter Battlefield 1 as a vehicle for self-teaching neural network AI.
The video is a collaboration between developer DICE and EA’s Search for Extraordinary Experiences Division, or SEED. The video shows the current progress of a self-learning AI as it has been teaching itself how to play Battlefield 1. The AI model is based on learning via trial and error, and started by watching human players run games, then moved on to honing its skills in bot-versus-bot matches.
Not only were the bots able to navigate the maps and engage with other bots, but also do things like run for cover, prioritize objectives, and even run for health and ammo pickups when needed. But there are still milestones to be made, as we also see the bots get locked into running circles around each other.
While the bots are currently used as extra tools to help in the quality assurance phases at DICE, the hope is to develop technology like this to add further complexity to gameplay, such as with NPCs capable of more complex tasks.