This was originally posted on tumblr on June 19th, 2014.
We all enjoyed Twitch Plays Pokemon whilst it lasted, and with good reason; it was hectic, ridiculous, entertaining and spawned its own equally strange metanarrative that acted as grounding for a sense of involved community feeling. I’m evoking this spectre to set it against ‘Automating Threes’ from Team Colorblind, in which a robot powered by an AI program plays successive games of Threes.
While TPP was about thousands of people interacting with a game at the same time – taking the concept of ‘single player’ and parsing it, dispersing it into a mad spectrum – Automating Threes is about the absence of people. Instead of people interacting with a system, we have one system interacting with another system. This changes our perception of Threes the game; it becomes an ersatz version of itself when viewed from a distance. The sound effects that play whenever sufficiently high-numbered tiles are paired together – small, cute greetings that reflect the light, casual atmosphere of a little handheld puzzle game – are no longer addressed to a player, but to another machine that cannot possibly answer them, react to them, or hear them at all. We can hear them as viewers, but the distance of the act of viewing means that those little greetings are not forus, the spectator. This uncanny quality is then driven home by the rotor-buzz of the robot arms that come down from on high to swipe the screen. We are made acutely aware of the machine-machine interaction, and the artifice of the personality of Threes is foregrounded to strange effect.
As I mentioned before, TPP had its own metanarrative; specifically, it had its own viewer-made religion based around the Helix Fossil (later Omanyte (later Omastar)) and the rest of the party acted as demigods-cum-prophets. Something similar happened in Automating Threes; the machine evolved into a symbol, its function into a kind of magic beyond human understanding, all because the audience placed that kind of hyperbolic signification onto it. The robot, named “Threesus” by the Twitch chatroom, was always making moves that didn’t make sense to the audience, who would then chastise the robot for not making the obvious pairing between two identical tiles, for instance. And then, a few turns later, the robot would make one pairing after another, a 12 became a 24 became a 48 became a 96 became a 192 became a 384. It was clear that the robot was thinking in the longer term – not just taking the present board into account but seeing some way into the future, into what the board would look like in several turns time.* Because none of the AI’s thought processes were explained until much later, this act of prediction felt more like prophecy, something not understood until it is fulfilled. Thus “Threesus”, thus the machine became a machine-god.
So Automating Threes is interesting because it tells us a lot about meaning and signification. On the one hand, the machine-machine interface is cold and distanced, and so it takes away the friendliness, the casual warmth that the game’s visuals and sounds are meant to convey.** It removes a layer of signification that masks the machine aspect of Threes, forcing us to look at it purely as a system that exists in and of itself and does not require us to be there to be. This is pretty much the opposite of the way we usually approach systems and machines – because we turn them off and on when we want to use them we get used to the idea of their presence being conditional on our need for them. When they seem to be acting of their own accord, it subverts our ideas about our relationship with them. Think about any time you’ve discovered that the television has been left on in the middle of the night – it’s eerie, right? It’s broadcasting to nobody because it doesn’t actually need anybody to broadcast to.
But the reaction of the audience was to find – or invent and impose – signification somewhere else. It would be bold to suggest that “Threesus” would not have been coined had TPP never existed or been such a potent, if brief, cultural force, but at the same time I don’t think it’s simple coincidence that the AI’s powers of foresight were translated into supernaturally-inspired prophecy. I’m not saying that “Threesus” came out of some crowd-wide existential angst, but it reflects our fascination with machine autonomy and machine intelligence that exists alongside a contrapuntal urge to translate what we can’t understand into what we can; essentially, the idea of a machine predicting the future state of a game board is still so awesome to people who aren’t well versed in AI engineering, whereas ideas of prophecy and deities are the stuff that the fabric of society is practically soaked in, and so we draw the connection between the thing we cannot understand and the thing that we can, the connection becomes transformative.***
* – Here is an excerpt from a blog entry by Walt Destler, who programmed the AI, which better explains the AI’s basic logic:
‘Fundamentally, Threesus works by looking 6 moves ahead into the future and then evaluating the potential future game board to determine its “quality” […] It does this look-ahead for every possible move in addition to many of the possible locations and new values of cards that might be added to the game board. Multiply all of these possibilities together, and Threesus will often examine millions of possible game states to decide just a single swipe. Once Threesus has calculated the quality of all of those game states, it simply swipes in the direction that is most likely to produce the highest-quality game state.’
** – This probably wasn’t helped by the looping Boards of Canada soundtrack.
*** – For a great example of the relationship between an omniscient machine and a society that doesn’t understand it, you should read Dystopolis by Chris Fraser. The relationship constitutes a gap between man and the machine, a void which drives the action.