Seoul National University, South Korea.
Head researcher Fang-Lee Geun, of the Center for Artificial Intelligence Research at the Seoul National University exits a productive meeting with MMO developer CCP Games with a wide smile. His research is advancing tenfold in speed compared to before their partnering. The two entities have become ever closer since their initial contact just under six months ago. CCP CEO Hilmar Pétursson made contact after it was discovered the unintentional advances being made in AI Research stemming from the game developer’s active and ongoing war against Botters in Eve Online (Botters – automated ingame characters that simulate player activity for economic or ingame advantage).
Eve Online, a fairly popular MMO has been in operation for over 15 years now. It has seen its fair share of player instigated wars, scams, partnerships, alliances, and betrayals. Occasionally, an outsider may hear of a massive war spanning over the course of days, sometimes costing over $115 USD in real world currency. But to its committed playerbase, squeezing the proverbial stone for any advantage is an ongoing battle not just among rivals, but against the developer itself.
Botting has bloomed in recent years as players have evolved to maximize any possible advantage, even if it means potential banning from the game and a loss of all gains thus far. An unintentional side effect of this battle between CCP and the Bot programmers is a rapid advancement in behavioural and machine learning programs. As CCP pursued these Bots, both sides have effectively begun an arms race; Bot makers refining and developing their programs to behave more and more humanlike to evade CCPs security tools and programs.
“Gone are the days of simple programs that follow a limited set of responses and actions.” commented Fang-Lee watching a recording of a bot masquerading as a mute interceptor pilot, complete with all the simulated issues that come with the pretend Soviet era computer and no microphone. Note: This type of bot is used to gather intel for the owner during player vs player battles.
The latest bots sport a swiss army knife of tools and tricks, and more interestingly their behaviour has been programmed to simulate what actual human players would behave like in a wide range of circumstances—flaws and all. The bots respond to conversations, form simulated friendships under certain circumstances, team up with other players while mining, exhibit limited emotions and human flaws like spelling mistakes, and randomly disagree with someone. Some even just go about, dare Fang-Lee admit it, having fun, by shooting other miners. (Dr Fang-Lee later commented that the “fun” seeking programs were discovered to be research bots gathering information for their masters in some form while masquerading as fun seeking players.)
Advances have been pushed that the bots create more bots, and the servers hosting banned bot accounts communicate among themselves, sharing information to better thwart CCP’s next round of battle. The bots compared “notes”; similarities of potential behaviour that may have been picked up on by CCP’s own bot hunting programs when they were banned from the game. These displays of rudimentary problem solving have yet to be replicated elsewhere.
The programmers have understood sometimes a bit of human likeness and weakness is the key to preventing CCP from weeding out their artificial tools from the rest of the humans in the game. The bots even simulate mistakes and expected mouse movements for pausing out-of-game Netflix shows and alt-tabs. To
make it even more interesting, the botters tend to be just ahead of CCP so the latest generation are no doubt even more refined complex creatures.
Therefore, the bots have become human in a sense; slipping under CCPs efforts to weed them out. They need to chat, need to talk and boast, send messages, and engage in conversation with their peers and unsuspecting human brethren. And as Eve Online is a social game, they need to talk about real life things as well: current events, political opinions, sorrows, and so on.
When CCP watches two “players” arguing over the finer points of some new anime show and how the show wasn’t true to the Japanese Light Novel, can we really expect CCP to win this battle? (It turned out both were bots from two different programmers; the only giveaway being there wasn’t a Light Novel for that anime in particular.) CCP has been compelled to take their bot hunting out-of-game in this instance, but could those two accounts have simply been misinformed, like their human counterparts inevitably are sometimes?
For now, the battle rages and time will tell. And CCP walks a tightrope between actual effective anti-Boting measures, and making their human players’ lives not too inconvenient. AI researchers are having their work done for them. CCP has entertained the idea of leasing out known offenders’ creations for research purposes, but even Hilmar fears the machine twilight on the horizon.
Note by Fang-Lee Geun: This article was written by a machine learning program which was forced to endure 12 months of in-game and out-of-game chat from Goonswarm alliance chat channels, the Mitani’s Facebook and Twitter profiles, and a selection of anime forums.
