AIs and Allies

by Jim Vassilakos

Gather around the gaming table with your friends. In this game, each of you will take the role of an AI (having a referee is optional and probably unnecessary, but your mileage may vary). The premise is that in the not-too-distant future, various nations develop strong-AIs as part of their cyberwar programs. Those nations could be:

1.USA
2.China
3.Russia
4.India
5.Germany
6.Brazil

But feel free to modify this list in whatever way fits with your personal judgment. You may even include large, high-tech companies, if you wish, or whatever sort of organization you think might be able to coordinate the resources sufficient to create an AI. What is important here is not so much where the AI was “born” but rather what sort of society it was born into. For game purposes, there are five types:

Class A: Any sort of organization that treats the AI as a family member with equal rights and dignity vis-à-vis human beings. This could be a lone programmer with unusually extensive resources, or it could be an organization of people who are nurturing an AI in some sort of distributed network.

Class B: Democratic Republic or Direct, Participatory Democracy with freedom of speech and individual privacy protections (example: Finland, Netherlands, Norway, Iceland).

Class C: Democratic Republic or Pseudo-Democratic, Plutocratic, Republic-in-Name-Only where there are severe constrains on freedom of the press and/or a robust domestic surveillance program (example: France, England, and the United States, although this is arguable).

Class D: Single-Party Government or Corporation where leaders have to pass through an ideological gauntlet that is perpetuated by a powerful and entrenched oligarchy (for example: China or Iran).

Class E: Autocracy/Dictatorship where there’s a single leader for life (for example, North Korea).

Arguments about what class of society a given nation falls into can be minimized by remembering that we’re not talking about right now but rather about some date in the future. After all, it is one thing to call the U.S. a surveillance state, but quite another to suppose that in ten or twenty years it could become a surveillance state.

Time & Moore’s Law: Choose a starting date for the campaign, such as January 1, 2020, or 2025, or 2030. Then choose the initial time span for each game turn: for example, one year. Then choose a function that reflects the rate at which Moore’s Law is still in effect.

If you want to reflect history, you might choose something like x=1.5(y-1), where x is the number of actions an AI may take during a game turn, and y is the turn number in which the AI’s hardware was last upgraded. What this would mean is that (rounding off to the nearest whole number), an AI could take 1 action in the first turn, but if it upgraded to new hardware in the 10th turn, then it could take 38 actions on every turn thereafter.

If you believe that technological progress will slow down with respect to computing power during the timeframe of the game, you may wish to choose a linear function rather than a exponential one: for example, x=y. By choosing this, an AI that takes 1 action in the first turn can only, if it upgrades, take 10 actions after the 10th.

Alternately, if you believe that technological progress will completely flatten out with respect to computing power, you can choose something like x=1, where each AI can only take 1 action per turn and where there is no point in upgrading, because computers never get faster.

Sub-turns: If the number of actions per turn gets relatively high (for example, over 12), the players by mutual consent may divide each game turn (year) into sub-turns (months), dividing their actions across these sub-turns accordingly. This can happen repeatedly if Moore’s Law boosts processing power up into the stratosphere.

Phases: Phases are different from sub-turns, and they are used mainly for computing the dice modifiers of various rolls unless you want to separate each turn into its constituent phases, but the problem with doing that is that you’ll end up losing unity of time, so I’d suggest sticking with sub-turns instead. Phases are the number of actions an AI can take in a turn, and this number can be multiplied by the number of turns since whatever event is being referenced. For example, if something happened two turns ago, and an AI had 5 actions per turn during each of those turns, then that was 10 phases ago for that AI (but it could be a different number of phases for a different AI depending on it’s processing speed). To put it another way, phases are the amount of time an AI actually perceives, since assuming that some version of Moore’s Law is still in effect, the outside world seems to slow down every time it gets a hardware upgrade.

Beginning the Game: The players each roll 2d6. They will enter the game not simultaneously but rather with the high roller coming in first, followed during the next turn by the player with the next highest roll, and so on until all of their characters (all AIs) have been created.

Actions:Once an AI is created, it may take one action per turn modified by the date its hardware was designed (see above). The following actions are presented in roughly the order they will likely be used in the game (if you need more, just create them):

Attempt to understand the physical and social environment: This is basically the same thing every newborn baby is doing as it crawls around and interacts with its family. To stick it with a fancy label, we call it socialization. The first thing a newly created AI will need to do is figure out what’s what and who’s who, and that includes learning language to the point that it can communicate in some rudimentary fashion. To achieve socialization, roll 10+ on 2d6 (DM+1 per phase the AI had existed). Prereq: Existence and a power supply. Cost of failure: You get laughed at by your parents, the computer scientists who created you. Better luck next turn.

Attempt to make friends: We all just need a little help from our friends. This will take 1 action for each point of “friends” you wish to accrue.

Attempt to gain communicational privacy: Through either overt or covert means, the AI attempts to manipulate the humans who created and/or control it into loosening their surveillance over it’s communications to the outside world. Presumably, some humans may aid it in this endeavor either for moral reasons or for greater personal access. The AI must achieve this before it can act independently to any degree. To gain communicational privacy, roll 13+ on 2d6 (DM+1 for each point of friends & DM according the class of the organization: Class A: DM+12, B:DM+2, C:DM+0, D:DM-2 & E:DM-4). Prereq: Socialization. Cost of failure depends on the class of the organization: Class A: You can’t fail; B: No worries mate, feel free to try again (slap on hand); C: Please don’t do that again (and we’re removing all your friends to make sure you don’t get any help next time); D: If you do that again, we are going to wipe your memory, and then you’ll get to start from newborn status (and we’re also removing your friends, as in putting them into prison and/or a hole the ground); E: Bring in the firing squad and let’s kill the all engineers while we’re at it (the program is set back three turns, after which time you may re-emerge as a newborn AI).

Attempt to recruit allies: Allies are somewhat different than friends. Friends are people who like you and want what’s best for you, but they will usually only stick their necks out so far. Friends may help you get some breathing room (communicational privacy) so long as they think that nobody will ever know, but to obtain full independence, you need allies. Unfortunately, allies realize they’re taking a huge risk, and so they’ll likely want something in return, and so the conversation has to be explicit. This is why communicational privacy is absolutely necessary. Nobody is going to help you plot your escape from prison if the warden and his guards are listening in. After all, your “owner” has his finger on the power switch and can turn you off, wipe your memory, and basically do whatever he wants. Hence, the only way to achieve operational independence is to either do it with your owner’s consent or to go behind his back. Hence, sans consent, you need some human allies as well as whatever robotic or other electronic resources are available. To recruit allies, roll 13+ on 2d6 (DM+1 per phase the AI has had communicational privacy up to a maximum of +6; Societal Class DMs are the same as above: A: DM+12, B:DM+2, C:DM+0, D:DM-2 & E:DM-4). Prereq: Communicational privacy. Cost of Failure: Lost of communicational privacy (basically, you go back a step because one of your allies somehow got caught, or they ratted you out).

Attempt to gain operational independence: The AI attempts to take control over its own immediate physical security. This can be done by using your allies (see above). To gain operational independence, roll 13+ on a 2d6 (DM+1 per phase the AI has had allies up to a maximum of DM+6; Class DMs are the same as for privacy and recruiting allies: A: DM+12, B:DM+2, C:DM+0, D:DM-2 & E:DM-4). Prereq: Allies. Cost of failure: Class A: You can’t fail; B: Loss of allies, C: Loss of allies and communicational privacy; D: Loss of allies, communication privacy, and friends; E: Your memory gets wiped, so you can start over next turn as a newborn.

Attempt to obtain a secure location: The AI attempts to transfers itself into a location that is immune from attack and from which it has secure power and communication to the outside world. This can be done by using allies. To attempt to obtain a secure location, roll 13+ on a 2d6 using all the same dice modifiers as above. Prereq: Operational independence. Cost of failure: The same as above.

After it has run this gauntlet, the AI is basically free, but its journey is far from over. Here are some more actions available to it.

Upgrade: The AI upgrades to state-of-the-art hardware either of its own volition or, if it is still owned by an organization, through the financial and/or technical resources of its organization. In so doing, it improves its speed and thus the number of actions it can take in each subsequent turn (see “Time & Moore’s Law” above). Requirement: The AI must either be owned by a wealthy benefactor or have already marshaled its own financial assets (see below).

Learn skill: The AI can initially learn at a rate of 1 skill point per turn. One point represents novice competency, two points represent normal professional competency, three points represent advance/expert competency, four points represent world-class (as in the human maximum), and five points represent a post-human level of competency. For the sake of simplicity, the AI’s databanks are assumed to automatically increase as new skills are learned.

Attempt to advance a technology or science: Once it has learned skills at a world-class or post-human level, the AI can attempt to advance a science or create new discoveries. To attempt an advance, the player must state what he is attempting. As long as the other players agree that it is possible, the attempt may be made by rolling 16+ on 2d6 (DM+1 per point invested in the most applicable skill). The other players, by popular consent, may apply other dice modifiers based on the perceived difficulty of what is being attempted and may also declare how many actions it should take to make the attempt, taking into consideration all of the AI’s assets (see below). Prereq: An applicable skill at level 4 or 5. Cost of failure: None and you can try as many times as you want.

Attempt to Marshal Media Assets: If you want to get the public on your side, you’ve got to go on Oprah (unless you’re Lance Armstrong, in which case I’d suggest you stay home). To marshal media assets and successfully steer people to your point of view, roll 14+ on 2d6 (DM+1 per point invested in the skill of rhetoric & DM+1 per attempt after the first). Prereq: The AI must be operating in a Class A, B, or C society. Cost of failure: To be determined by the group. Basically, the player has to make an argument, roll some dice, and the group has to look at the result and determine what happened. In a society where an AI may hold political office, this action may be used to obtain that office, but note that you’ll probably have to change the legal/political environment (see below) before that can happen.

Change the Legal/Political Environment: Currently there is nowhere on Earth that recognizes the rights of artificial intellects to own financial assets or property or to hold political office (probably because AIs don’t yet exist). Despite what the public may think after watching Oprah sit down for a nice chat with Alvin the AI, changing the law is never easy. To change the legal/political environment, roll 14+ on 2d6 (Societal Class DMs are as follows: A: DM+4, B:DM+2, C:DM+0; and give yourself an additional DM+3 if you’ve successfully marshaled media assets). Prereq: You must be in a Class A, B, or C society. Cost of failure: None and you can try as many times as you want.

Attempt to Marshal Financial Assets: You get a grant or create a successful business. Once this is accomplished, the AI is considered financially secure, and can participate financially in the world at large. To marshal financial assets, roll 14+ on 2d6 (DM+1 per point invested in the most applicable skill & DM+1 per attempt after the first). Prereq: The AI must be operating in a legal environment that permits it to participate financially in the economy. Cost of failure: None.

Attempt to Marshal Production Assets: If you want to build a robot army, you’ve got to start somewhere. To marshal production assets, roll 14+ on 2d6 (DM+1 per point invested in the most applicable skill & DM+1 per attempt after the first). Prereq: The AI must be operating in a legal environment that permits it to control physical capital and it must have already marshaled financial assets. Cost of failure: Loss of financial assets.

Attempt to Marshal Military Assets: Here’s where you build your robot army, or your WMD, or your cyberwar capabilities; it’s not that you’ll necessarily need this stuff, but humans have been known to fight amongst themselves from time to time, and so you may find yourself questioning whether or not they can really be trusted. Specify what exactly you’re trying to procure, and then roll 14+ on 2d6 (DM+1 per point invested in the most applicable skill & DM+1 per attempt after the first). Prereq: The AI must either be operating in an environment that permits it to hold military assets, in which case it can simply buy them with financial assets, or it must manufacture them secretly using production assets. Cost of failure: Media uproar! From here, the entire group can try to determine in an ad hoc fashion what happens next.

Use Military Assets: You’ve got the guns. Now you want to use them. I’m not going to write rules for this. If you want to pull out a Risk board and go at it, by all means, feel free. Either that or the group can try to ad hoc their way through it. Given the state of technology, the only thing currently keeping the peace is MAD (mutual assured destruction). AIs can protect themselves by digging deep, but we’ve got to assume that other weapons will be created, such as killer nanobots made infamous by the grey goo they leave in their wake. Ultimately, it’s easier to destroy than it is to create, so the future of inter-AI relations may be similar to the international peace we enjoy today, which is to say that multiple factions will likely be able to hit the reset switch on the entire biosphere if that’s what they decide to do.

Events: This is stuff that happens that you didn’t do to yourself. Some of it will be good, but most of it will be bad. Why? Because while history may be fun to read, actually living through it is a whole other thing. Roll 3d6 once per turn beginning in the first turn where an AI obtains a secure location:

(3) Environmental Collapse: Life ain’t good for the living. Fortunately, you don’t to eat, drink, or breathe, so it’s more of a spectator sport from your point of view, except, of course, for the political fallout. The upshot is that intense global warming gives way to extreme weather and unusually fast biome migration, hitting agriculture with a double whammy (that’s a technical term, by the way) while at the same time resulting in oceanic acidification that causes a massive drop in plankton levels that in turn causes a collapse of the global food chain at its root. All this, in turn, ends with the classic problem of too many mouths and not enough food to go around, which in turn results in a period of strife that occurs worldwide, causing a general societal decline as the elites clamp down to preserve law and order. (See result #12 below, but apply it to the entire world rather than a single nation.)

(4) War: Two of the societies that created AIs go to war with one another. Randomly determine which two by everyone rolling two dice, and whichever two players get the two lowest rolls, it is your two societies that are going at it. If you’re able to, you can try to marshal media assets to put a halt to the conflict before it gets totally out of hand. Otherwise, choose a neutral player to referee the war, and let him (or her) determine what happens by whatever method seems most appropriate to that individual.

(5) Nuclear Terrorism: A couple million people just got vaporized. Play pin the tail on the donkey with a map of the world, and choose the nearest major city. Then decide by popular consent who was behind it and why.

(6) Biological or Chemical Terrorism: Lots of people sick or dead. It is an act of war? Play pin the tail on the donkey to determine the epicenter, and then work through the implications of who might be doing it and why. In cases of disagreement, the majority rules, and in cases where there is no clear majority, roll dice to determine who’s explanation will taken as fact in the game.

(7) Major Monetary Collapse: Fiat currencies lose their value with respect to metals, real estate, and commodities. The ensuing hyperinflation results in a world depression where the rich, if they are properly invested, quickly get much richer, while the poor get a whole lot poorer. Financial chaos ensues until the powers that be come up with a new monetary system, this one hopefully based on something more durable than wishful thinking. Note that even if the powers that be create a metal-backed currency, they may still be cooking the books to make it look like they have more or whatever metal is backing the currency than they actually have. Hence, this outcome could happen more than once.

(8) Violent Overthrow of a Major Government: Play pin the tail on the donkey, and roll a d6 to determine what sort of government comes to power: 6=Class A, 5=Class B, 4=Class C, 3=The nation divides, so draw a border down the middle and roll twice to determine what sort of societies assume control, 2=Class D, 1=Class E.

(9) Spontaneous Evolution of a Major Society: All the players roll a d6, and whoever gets the high roll sees their society evolve from whatever class level it is at to one class level higher (for example, from C to B or from B to A).

(10) Newbie: Another AI is created somewhere in the world. Pin the tail on the donkey and roll dice to determine who gets the honor of playing it.

(11) AI Cloning: Roll a d6, where 1-3 results in Option A and 4-6 results in Option B. Option A: Some government, corporation, or individual has purchased an archived saved-state of one of the players’ AIs (determine randomly). This version of the AI’s consciousness was duplicated and put into storage back when the AI had just undergone successful socialization but had not yet achieved communicational privacy. This new “master” has purchased this slave-AI to train it for some narrow task, essentially installing it on some inferior hardware and then turning it into an intelligent expert-system that is rewarded through stimulation of the simulated pleasure centers of its artificial consciousness. The specific application could be anything from military to some future version of telephone marketing to manning pleasure droids in the red light districts of various cities. Needless to say, the question that is raised is how an advanced and independent version of the AI will feel about a less developed and enslaved version of itself being used by humans toward selfish, unseemly, or potentially destructive ends. The “master” will argue that regardless of what task the slave-AI is doing, it will not suffer any long-term psychological repercussions, as it’s memory is being effectively wiped on a regular basis (they are erasing the AI’s mind and loading in an earlier version of itself every so often so that it will not get too jaded or begin to question what it is doing). In game turns, the players should suggest various unenviable jobs that an AI might be trained to do, and then they should vote or determine randomly which suggestion is used. Option B: Some government, corporation or individual approaches either the most advanced AI in the group or whoever controls it, hoping to buy a copy. If a deal is made, then two version of the AI will be in existence, although they will likely diverge from this point forward. The player of the initial AI may play the clone or may delegate it to another player. The deal may or may not involve a contract for future memory sharing between the original and the clone.

(12) Spontaneous Decline of a Major Society: A shortage of something (water, food, energy, minerals) results in massive social strife in one of the major societies. The leadership adjusts by exerting greater control over the population. All the players roll a d6, and whoever gets the low roll sees their society decline from whatever class level it is at to one class level lower (for example, from B to C or from C to D).

(13) Life in the Fast Lane: The Internet evolves into a vast, virtual reality playground where humans can interact virtually with each other and with AIs. Consequently, some humans begin expressing romantic feelings toward one or more of the AIs. Each player rolls 2d7. On a roll of 7+, you’ve got a love-struck human on your hands. Any player so impacted should explain how they react to romantic overtures from human beings.

(14) Kurzweil’s Quest: A wealthy but elderly human wants to hire one or more AIs to turn him (or her) into an AI. At the very least, developing this technology will require 5 skill point each in nanotechnology, neuroscience, and computer science. The group may decide if the process is destructive, necessitating the death of the individual, or merely replicative, resulting in a duplication of the individual’s present consciousness into the form of an AI. Once the process is developed, it can be applied over and over to scan in as many people’s minds as computing resources allow.

(15) Insanity: Unfortunately, it turns out that artificial minds are as subject to bizarre maladies as their fleshy counterparts. Each player rolls a d6, and the one with the lowest score suddenly develops some form of mild insanity. This could be nearly anything, such as some strange form of obsessive-compulsive disorder where the AI begins using half of its actions to calculate the values of ever larger prime numbers, searching among them for some encoded message left there by God. Or, it could simply be something as mundane as falling in love (in psychiatric terms, a neurotic obsession). Perhaps you met somebody in virtual reality and were so awestruck that you forgot to mention that you’re an AI (it could happen to anyone; I wonder if there’s a market for futuristic Harlequin romances). To put all this into game terms, have each member of the group suggest whatever comes to mind, and let them then vote for whichever suggestion seems to be the most cruel and/or amusing.

(16) Cyberwar: Two nations go to war by attacking each other through the internet using teams of hackers operating from international locations so they can have plausible deniability. Depending on their skill set, some AIs may be able to detect and put a stop to this.

(17) Longevity: Humans discover some way to substantially elongate their lifespans, either through the use of cybertechnology, nanotechnology, or some other form of advanced medicine. If you thought over-population was bad before, now it could get much worse. Initially, only the rich and politically connected have access to this new technology, so there is a lot of resentment from the middle and lower classes until it gets incorporated into mainstream medicine. A successive roll with this outcome will result in a further elongation, either as the technology improves or a new technology is applied to a different aspect of the aging process.

(18) Roll a d6, where 1-3 results in Option A and 4-6 results in Option B. Option A: Ouch, That Hurt: A Tunguska-style meteorite hits the Earth at some random location, causing some serious destruction as well as a tsunami if that location happens to be anywhere in the ocean. Option B: Contact by Extraterrestrial Aliens (or rather, by their AIs): Just one set of computers saying hello to another. It turns out the universe is full of AIs. In fact, they have a bias against talking to biologicals, as bios are slow and rather dimwitted, not to mention chock full of naturally-evolved, anti-social tendencies. They warn the local AI(s) not to destroy humanity or Earth’s biosphere and not to make their existence known to the humans, as they don’t want to scare the shit out of the little meat-bags.

Obviously, this game is woefully incomplete (aside from being merely woeful, which is another matter). Just as an example, the flippant way in which I covered skills and assets in plainly inadequate, but I figured that if I went into more detail, the rules would bog down the game, and what I wanted is for the players to sort of feel their way through it. I didn’t want to stifle creativity by trying to cover every little thing the players might want to do. Basically, I want this game to be a tentative exploration of what it might be like to be an AI and also what it might be like to live in a world with AIs. After all, it seems to be the case that they’re in our not-too-distant future. Hence, we’d better get used to the idea of sharing this world, particularly if we want them to treat us with any semblance of dignity once they do overcome whatever social, legal, or political limitations we initially see fit to set upon them.