Angry Robot

The A.I. Game (The Beast)

This was originally published on Joystick 101 but has since gone offline.

Like thousands of others, I’ve gotten hooked lately on the bizarre web promotion for Spielberg’s new film A.I. Essentially, it’s a web-based multiplayer game. Such a simple description, however, belies many of this game’s interesting characteristics. So, I thought I’d try to spell out what makes this game unlike any other I’ve played.

Now, I’d like to apologize in advance for analyzing this game before it’s finished. It feels a bit like doing an autopsy on someone who ain’t dead yet. There is still, I assume, a month or so before this game will end, but that’s in fact why I wanted to write about it now. If you haven’t heard about it yet, and it sounds interesting to you, there’s still time to join in and play.

Okay, begin at the beginning: it seems that some observant folks noticed some odd details in the AI trailer. For example, one of the credits was “Sentient Machine Therapist: Jeanine Salla.” A google search for Jeanine Salla pulled up some websites that were from the year 2142, such as a homepage for the Chan family, and the site of the university that Salla apparently works for. Exploring the websites that are linked to from these sites expands the story a bit at a time. Puzzles of various sorts are encountered; when solved, they generally grant access to a previously unseen page or site, which in turn continues the story. The story concerns the murder of a man called Evan Chan, and it involves artificial intelligences of various forms. The main theme appears to be slavery—robots and AIs are still considered property; there’s an ‘underground railroad’ that frees escaped robots. At this point in the game, it seems that the AIs and their supporters are gaining the upper hand in this futuristic civil war. As one of them says, “Moore’s Law will put us up against the wall first.”

That said, I’d like to focus on the formal elements of this game, and specifically those that distinguish it from other games. I have a feeling we’ll see more games like this (Majestic?). It’s also interesting to take a game such as this one, which feels so different from what we tend to mean when we say ‘game,’ as it helps us forget about FPSs and RTSs and RPGs for a moment and realize how broad the concept of “game” can be.

1. web-based

Of course, there are many types of ‘web-based games’ (links to newgrounds, However, this game is much more fundamentally based on the web The story is told through the process of surfing. To qualify further:

1.1 content is fictional and dissimulative: Everything that is on the many sites created for this game is, of course, fiction. It is important to stress that the sites are also dissimulative, that is, feigning to be real sites. It’s almost like a parody without the intent to amuse. This is what distinguishes the game from some of the works of hypertext fiction that have gone before it. There are no pages announcing that “this is a work of fiction,” and in fact, some of the sites could easily be misconstrued as real at first glance (

1.2 Content is decentralized: There is no homepage or gateway to this story. The content is spread across many websites (link like crazy). Interestingly, the most effective entry point is a site made by the players themselves (

The end result is that playing this game feels like surfing the web, except with way more drama than usual. You’re still clicking links, checking out pages, going from site to site, but there’s a murder mystery to be solved this time.

2. Not just the web

The bulk of the game happens on the web, but the most dramatic events seem to happen offline. Email, faxes and phones all play a part in this game. For example, the A.I. trailer also included an encoded phone number, which when called, presented a mysterious message from ‘mother.’ Players have been able to enter characters’ passwords into fictional voicemail and uncover new information. A.I.s have even called players at home.

Most dramatic, however, were three real-life “rallies” held by the “Anti-Robot Militia” in New York, Chicago, and Los Angeles. Players were given a date and address, and attended what appear to have been clever theatre pieces. The ‘rallies’ themselves had puzzles, the solutions to which required real-time collaboration between players at the event and those at home in front of their computers.

3 Episodic release

The content is updated weekly. Some things are added, others are taken away. Emails are sent out; increasingly, sites are hacked by rampant AIs. With a complex, ongoing narrative like this one, the disadvantage of episodic release is that the later you start playing, the harder it is to follow and/or catch up. The advantage is a heightened sense of urgency, I think, since the game can’t be paused. Once again, the episodic release feels natural for a web-based game, since most real sites change with time as well.

4 “distributed biological problem-solving”

Many of the puzzles in this game are fiendishly difficult. For example, messages have been hidden in the html source of some pages. Anyone could uncover this, of course, but given that this game has so many websites, it’s unreasonable to expect solitary players to do so. In fact, it’s safe to say that it’s impossible for an isolated individual to play the game from start to finish. Thus, fan sites (club) have served as a meeting ground for game players, who collaborate by sharing new developments and results. Problem solving tasks are organized and shared out amongst the players. The whole scene reminds me of the eclectic mix of cryptanalysts assembled by the British government during WWII: in order to break the Nazi Enigma cipher, the talents of mathematicians, linguists, classicists and and chess masters were called upon. In this game, some puzzles require knowledge of code, others of poetry (one involved T.S. Eliot’s The Wasteland).

5 strong interaction between authors and players

Presumably, there is a set story arc to this game that will end with the release of the film. The weekly updates generally involve puzzles to be solved before new story content can be consumed. Since the puzzles require group collaboration, but the size and effectiveness of the group is an unpredictable variable, we can assume the work of the authors of the game is quite frequently based on how well the players perform. For example, if a puzzle turns out to be much too hard for the players, the authors must find an alternative means to provide the story update that the puzzle’s solution would have granted. Otherwise, there is a risk of the story never being completed. The experimental nature of this game adds to this as well; many of the techniques they are trying haven’t been done in this context before. At one of the ARM “rallies,” a piece of paper was found floating in the toilet. Unfortunately, it had soaked too long before being discovered, and the print on it was illegible. I would assume the authors had to find a different way of transmitting that information to the players. Thus, the game has an element of improv to it.

It is worth noting that the game authors have deliberately blurred the lines between themselves and the players. In few cases, game pages link to fan pages without breaking the dissimulation: Jeanine Salla’s paper on ‘Multi-person social problem-solving arrays considered as a form of “artificial intelligence“’ links to the cloudmakers page, the Center for Robotic Freedom urges you to help fight for A.I. rights by visiting the spherewatch page. In fact, the easiest way for game authors to control the story delivery would be for them to surreptitiously join the fan clubs, posting solutions to puzzles when the real players were having trouble.

It would be interesting to think about ways to fold some of these characteristics back into more traditional videogame models. For example, this sort of multiplayer game makes Quake 3 and even Tribes 2 look hopelessly primitive. What if multiplayer games were narrativized a little more? If their levels were more like single-player levels, that required a great degree of collaboration to conquer? Or, imagine a more open-ended, multiplayer world. Depending on choices made by the players, the designers “improvise” what part of the story or world is opened up next. And also, perhaps there is a small market for text-based games like this one. (consider the “play novels” of Japan.) If developers could spend less time worrying about camera movement and cinematics, and more time thinking about story, they might come up with something interesting. Finally, what if games had many more ‘modes’ than usual? What if they were as decentralized as this one? Part of a game could be a FPS, another part more puzzle-oriented, and yet another more strategy-based – depending on what the story called for.

That’s what I like about experiments like this campaign: it makes us all rethink what we mean by ‘game.’ This industry is obviously still very young, so it’s always a shame when we form preconceptions about what games can and can’t be. Even if the experiments don’t succeed, they have shown us new directions. This experiment, though, I would consider a success. I hope that when it’s all over, it’s creators step out of the shadows and take the credit they are due.