I'm not sure if this has been discussed already. Engadget reported over a week ago that at GDC EA announced that it was working to create an AI using machine learning and neural networking. Though EA admitted to having problems with the AI, it performed well enough to compete with a human a on a very basic level and had plans to pursue teamwork and bigger and tougher conditions

www.engadget.com/2018/03/22/ea-teaching-ai-troops-battlefield-1/

Most people who responded to this were filled with worries over Skynet or perhaps more realistically using the AI's developing mannerisms that adapt to human behavior as a template to better analyze gamers in order to condition them to using paid content. 

However it seems to be the opposite as Engadget has reported a couple of days ago that a source in GDC was playing against an more recently updated AI with a few others with problematic results for EA. When prompts came up for Battle packs, the AI gave very erratic inputs. EA eventually had to shut down the session following an much more predictable behavior the AI. Since the AI uses the a sound adaptive system, Eurogamer cites that, "[the source] understands that the Ai was able to [repeat toxicities]" before it was hastily shut down. It appears now EA is rethinking its future.

www.engadget.com/2018/03/28/ea-shuts-down-battlefield-1-ai/

Again, I don't know if it's been discussed, but personally, I think this is a hilarious learning experience for those that read both the original article and then the later article not to jump to conclusions. The Ai thing seems interesting, so I hope it succeeds regardless of the issues it has currently.

Probably hit too many friendlies with tanks and Jeeps.