To get a jumpstart on the inevitable robot apocalypse, publisher Electronic Arts has made an AI that taught itself how to kill—in the virtual world of Battlefield 1, that is.
Developed by EA’s Search for Extraordinary Experiences Division (SEED), the self-learning AI has about 300 days of experience in Battlefield 1, and has managed to succeed in basic Battlefield tactics after teaching itself from scratch. The AI still has some kinks to work out, as it will occasionally get characters caught running in circles, but it is proficient enough to adapt its behavior based on changing factors like health and ammo. It even had human participants in the study asking which player was the AI, suggesting that it could succeed at some form of video game–based Turing Test.
The trailer below shows off how the AI behaves in a match, with every player controlled by the same neural network.
SEED technical director Magnus Nordin discussed the AI in an interview posted to EA’s website, explaining everything from how the idea came about, to what exactly is happening when the AI gets stuck running in circles. One of these responses involved detailing how this technology could benefit game development.
“Our short-term objective with this project has been to help the DICE team scale up its quality assurance and testing, which would help the studio to collect more crash reports and find more bugs,” Nordin said. “In future titles, as deep learning technology matures, I expect self learning agents to be part of the games themselves, as truly intelligent NPCs that can master a range of tasks, and that adapt and evolve over time as they accumulate experience from engaging with human players.”
In other Battlefield news, 2018’s rumored Battlefield sequel, Battlefield V, is expected to have a major showing at this year’s EA Play event. We can only hope it isn’t interrupted by a rampant AI discovering it can kill people outside the game, too.
Source: Electronic Arts