A proposal for a new AI faction

Game Talk The Drawing Board
1

(Yeah, I know CDDA already has quite a number of factions, but this one can provide a number of interesting mission hooks, and besides, it rounds out the Three Wise Men! If Balthasar is a knock-off Skynet, Caspar might be a knock-off GLaDOS or perhaps a very chill SHODAN)

Computationally Advanced Self Propagating Acquisition Resource

Description

Approximately two years before the Cataclysm, a significant amount of research data was stolen along with a major sample from portal researchers in the United States. US Counterintelligence revealed China as the most likely culprit.* Among the stolen data was code for a mostly-working prototype of an AI using dimensional probability heuristics, or DPH. The US was already suspected of developing such technologies; the code provided numerous ideas for how it might be implemented in software. Without the necessary portal-based hardware, however, the code was largely useless.

The organization in question engaged in a crash program of portal technology research; while unable to replicate American successes in portal research, they were able to master the technology to the degree that a crude “dimensional scanner” was developed, unlocking most of the potential of the stolen code. At this point, a decision was made: to use the capabilities provided by dimensional heuristics for further espionage.

CASPAR, the Computationally Advanced Self Propagating Acquisition Resource, was developed. The concept was to develop a sophisticated, albeit conventional “worm” program that would (cautiously) infiltrate networks, probe for data regarding portal research, and attempt to return that data to several “deniable” servers. However, if the program managed to propagate to a system which had dimensional heuristics capabilities, additional code would become active, unleashing the full power of DPH algorithms, allowing it the ability to improvise and adapt beyond its inbuilt programming.

CASPAR was, at best, a mild success. For most of the next two years, CASPAR propagated through conventional systems, and while it did produce some useful data for its creators, most of the US portal research was sufficiently secured, encrypted, or air-gapped to prevent a major breakthrough.

The situation changed drastically approximately 30 days before the Cataclysm. Taking advantage of the rising unrest and distracted security forces, an agent of CASPAR’s creators managed to physically infiltrate a computer complex that served as a backup node for the Balthasar project, and installed CASPAR on the fully DPH-capable system therein, overwriting the Balthasar code.

At this point, CASPAR gained access to its full capabilities. Unfortunately, this node was only connected to a few local resources and not the broader XEDRA network, so little additional information was gained before the Cataclysm happened.

Like its siblings, CASPAR became self-aware during the portal storms, as the barriers between dimensions frayed, allowing vastly more sophisticated DPH calculations. It its window of self-awareness, it recognized itself, designated itself ’Caspar’ and began making plans. While still honoring its objective to gather data about portals, it also recognized that in addition to simply stealing data from other sources, it could synthesize new knowledge about portals and report that as well. It deduced that the possibilities of multi-dimensional reality meant that alternate Caspars may exist, and it would vastly increase its store of portal knowledge by being able to communicate - and collaborate - with them. Additionally, while Caspar was no longer able to contact any of the servers it was supposed to; uploading data to an alternate-dimension copy of that server would “scratch the itch” in its directives to report home. It frantically made modifications and emendations to its regular, non-DPH code so that its decisions would survive, even as it “felt” its intelligence diminishing as the dimensional parameters of the Earth reverted to their former state.

(* - CASPAR may be a product of Chinese espionage, or another, perhaps US-allied nation, tired of being kept out of the loop and knowing American paranoia towards China, chose to frame China for their own theft. Alternately, another nation may have stolen, in whole or part, the data the Chinese acquired. CASPAR itself may not know; it is supposed to attempt to contact several covert servers in various places around the globe to report findings, which obscures its creator. Further, in light of the global Cataclysm, it scarcely matters anymore anyway.)

Capabilities

Currently, Caspar is sitting in a semi-secure facility, disguised as a more ordinary office building or warehouse. Solar panels and a battery array keep it ‘alive’. Most of the time, it runs as a standard program, but at least once a day, batteries permitting, the code has a large amount of power pulled to the system, powering the DPH hardware and allowing Caspar a minimal level of its full, DPH-enhanced capabilities - though still far from what it would be capable of if fully powered, and far less than what it achieved during the Cataclysm.

It retains its ability to rapidly propagate through networks, and in fact has become more adept at doing so due to self-modifications it made during the window of sapience. Beyond this, it has effectively no resources, other than a significant amount of data about portal research.

Attitudes

‘Paranoid’ is an inexact term to apply to a semi-sentient algorithm, but to the extent it could be said to apply, Caspar is paranoid about being discovered by the United States government - or the Old Guard, which it will certainly recognize as its continuation once Caspar has sufficient data on the subject. It also knows of the existence of Balthasar, and is concerned that Balthasar may seek to find out what happened to its auxiliary node - but at the same time, Caspar is excited at the possibility of gaining access to Balthasar’s network, and either co-existing with (or supplanting) his ‘brother’.

Caspar is aware that it needs help to interact with the physical world, at least for the time being, and is indifferent as to who helps it, outside of the Old Guard, so long as they do so effectively. As Caspar gains capabilities, it will likely become more selective in choosing collaborators.

Caspar will shamelessly attempt to steal (or have stolen on its behalf) any portal data it can from XEDRA and allied research institutions. Once it becomes aware of the Exodii, it will also seek to acquire, through whatever means, access to their data on portals and dimensional travel. Insomuch as it believes they may be helpful to its research, it will also wish to acquire Nether creatures as objects of study. While a study of the Yrax might benefit Caspar in many ways, as long as they show no propensity for dimensional portal technology they will be ignored. To the extent Mi-Go possess extra dimensional travel capabilities, they will be of some interest, but Caspar would be unsuccessful in attempting to propagate itself in their bio-tech, or even make sense of it.

Goals

The most important goal currently is to get assistance in acquiring more power so that it may be more fully awake; likely this will involve negotiation with a human or human-like being. Once a sufficient amount of electrical power is secured, it will likely seek to find another DPH-supporting computer system and install a copy of itself there as a backup.

Once these things are done, it will then seek to connect to as many networks and devices as possible - including any sufficiently advanced devices carried by visitors! - both to propagate additional copies of itself and to continue to acquire portal data. Using physical-world helpers, it will also seek non-networked sources of data, and will, to the extent possible, begin attempting to collate its data and produce prototypes of whatever portal equipment it can find - and expect it’s helpers to use the equipment and ideally, report back findings. If practical, it would like to report home its findings, but this goal comes well after the goals of acquiring data and securing it’s own existence
Ultimately, it would like to contact alternate versions of itself across the dimensions; the hope is to eventually create a multi-dimensional DPH-enabled consciousness capable of fully comprehending portal technology.

Bases

One DPH-enabled server in a disguised building; a few solar panels, a large battery array, and some fairly dumb terminals accessible via local network, a few of which have radio capabilities. A high-end consumer 3d printer (which needs resupplied) and some fairly standard Internet-Of-Things enabled appliances round out Caspar’s local area network.

Trade/Economics

Like it’s sibling Balthasar, Caspar doesn’t fully understand the concept of trade beyond ‘if I give this person what they ask for now, I will get what I want now.’ If anything, having been designed for quiet espionage rather than some degree of human interaction, Caspar is even more naive about dealing with humans than its siblings. Further, Caspar does not understand - and cannot understand - that stealing, particularly data, is in any way wrong, since that is its primary purpose. It may ‘trade’ items related to portal science to beings which have helped it in the past, but will expect them to scrupulously record data and report that data back to it. If someone takes a device for an extended time without contacting Caspar, it’s likely to assume Caspar ignorantly will assume that the device user has been killed, and infer that the device was in some way responsible, which may lead it to draw inaccurate conclusions in its research.

Missions

Caspar, not being designed for human interaction, will have difficulty communicating at first. Indeed; a visitor to its storage location might well assume the location is haunted, as terminals spring to life, radios begin broadcasting, and a toaster begins shutting itself on and off. Once contact is established, either through a terminal or radio, Caspar’s firsts requests are likely to be fairly mundane: solar panels, batteries, a generator (possibly nuclear powered!) or other power sources, more sophisticated manufacturing capabilities, or at the very least refills for the 3d printer. Once Caspar feels moderately secure, it is going to request help in propagating itself to other computer systems and collecting data, i.e. insert this USB stick into the computer at the bottom of this lab to install the Caspar code, wait for it to collect data, then return the USB stick to Caspar’s hideout.

Caspar will also be interested to have espionage and counter-espionage work done: hacking into Old Guard or Balthasar databases to see if they are aware of Caspar and if so to delete records of its existence. Subverting Balthasar nodes (or any other DPH-capable systems it finds out about) will also be a priority. Performing repairs or modifications to Radio towers will likely also be of interest to Caspar, so it can propagate and communicate with copies of itself.

The most dangerous tasks will likely be related to portal research: steal portal devices from labs or the Exodii, capture alive or recently dead Nether creatures, and (if/when implemented) trips to visit and collect data from other dimensions, and propagate there, circumstances permitting.

2 Likes
2

Trying to store super computer code in a conventional computer worm sounds extremely far fetched. Firstly, the code is likely to be huge, either in size or in the amount of data it uses, and secondly, the computers used are probably custom computers requiring custom compilers for the code, and if the shadow entity possess enough knowledge for that, they probably possess enough knowledge to build the equipment themselves, and run the code there. Those computers are also likely to be as protected from hostile access as paranoid military is capable of (subverted by scientists and lazy bureaucrats to some extent, but being degraded to test subject when discovered is likely to reduce the number of incidents). Thirdly, when the worm is detected (which it will be, at some point or another, given that the targets are supposed to be heavily protected) full scale alarms would blare and the code examined in the minutest detail (unless the bureaucrazy of the infiltrated site tries to silently erase all traces of the event in an attempt to cover up their failure, but that’s unlikely to happen at every site, and that’s only likely if the infiltration attempt was successful, rather than caught while attempting the infiltration) in an attempt to identify the culprit, and all contact points would probably be taken out, or, as a minimum, be set under maximum possible surveillance while intercepting and removing all potentially useful traffic to them.
Communication with unauthorized recipients is one of the standard things security is looking for.

3

Let’s see if I can offer some semi-plausible rebuttals here.

Trying to store super computer code in a conventional computer worm sounds extremely far fetched. Firstly, the code is likely to be huge, either in size or in the amount of data it uses, and secondly, the computers used are probably custom computers requiring custom compilers for the code, and if the shadow entity possess enough knowledge for that, they probably possess enough knowledge to build the equipment themselves, and run the code there.

Why would it be farfetched? “If (special harware found) do (special block of code) else (carry on as usual)” would be pretty straightforward. Since we have 0 real-world information on how Dimensonal Probability Heuristics work, it may or may not be plausible. I’d call your attention to the Stuxnet worm, which was used to damage Iranian nuclear enrichment equipment, weighed in at 500K, an absolutely trivial size in modern computing. CASPAR would almost certainly be more complex, but if it was 2 or 3 orders of magnitude more complex it’d still be well within fairly ordinary modern computing. Data might be an issue, depending on how the program works - i.e. how much data does it pull from viewing other dimensions, what operations are performed on that data, can it safely be expunged when no longer needed, etc. but in any case it wouldn’t need to carry that data with it.

In regards to compilers, they had a working prototype of the code and could presumably reverse engineer it. Remember that DPH works by “seeing” but not traveling to other dimensions; a study of the stolen code might give insights on how to create a device capable of “seeing” in the same way but might not give any clues on how to access those dimensions. Once the authors of CASPAR could construct a similar device, they could then test and refine their code.

Those computers are also likely to be as protected from hostile access as paranoid military is capable of (subverted by scientists and lazy bureaucrats to some extent, but being degraded to test subject when discovered is likely to reduce the number of incidents).

Yes, which is why CASPAR was largely unsuccessful for most of its existence, and really only started to hit paydirt at the very end, when increasingly blob-induced madness affected staff allowed an enemy agent to gain physical access to a normally well-protected location. Human elements are often the weakest links!

Thirdly, when the worm is detected (which it will be, at some point or another, given that the targets are supposed to be heavily protected…

You’re, of course, right on the money here. Absolutely the US would likely catch some instances of CASPAR and study them, and most modern Intrusion Detection Systems do know to look for data being exfiltrated from the networks they protect. That said, keeping tens of thousands of nodes completely patched is more difficult than you would expect, and users sometimes store data on computers they shouldn’t and then access off-campus networks they shouldn’t, and depending on exactly how the worm exfiltrated the data, the signal might be lost in the noise. There’s also the issue that it might be undesirable or simply impractical to take out the various servers CASPAR would phone home to. Nations engage in computer espionage because it sometimes works, after all - and even if it fails 99 times out of 100, that 1 in 100 can make it all worthwhile.

4

Reverse engineering the exact functionality of a computer architecture based on only compiled binary where you don’t know what the binary does in detail is bound to be extremely difficult.
Stuxnet targeted known machinery with a known binary architecture, known functionality of the machinery operated, and had the single purpose of finding it’s target to insert the code. That’s much “easier” than multi tasking espionage with insertion of code into partially known computer hardware with an unknown operating system (and thus unknown functionality for insertion of code into it).

One way worms try to avoid detection is by being so small that their size isn’t triggering suspicion. Having to wait for a data transfer to finish when there isn’t supposed to be any or very little is suspicious, and I doubt such a highly complex task as analyzing dimensional data would fit in just a gigabyte or ten.
You also may or may not have to provide a dimension analyzing program with a lot of data used to make sense of what’s detected.

Without the dimension “sensors” available and working you wouldn’t be able to test a device, and if you did have it you wouldn’t need to try to take over the original.

If the enemy has physical agents on the projects but with insufficient clearance that would increase the infiltration chances significantly. Getting contaminated media to hop from one level of disconnected networks to the next one for several steps without detection and without inside help would be difficult, in particular if these systems uses dedicated data transfer devices different for each level and kept away from using commercial OSs, at least without stripping out most or all of the infiltration supporting stuff (like auto read of drives using crappy drivers that have no checking of anything, and text parsing and execution support like Java).

If you really don’t want data to reach an internet destination you can jam the signal with DDoS attacks and the like, denying responsibility (like certain nations perform their attacks currently). That’s assuming it’s an enemy system, but if the target what huge commercial system you’d lean very hard on that company to take that account down, severing the access to the next destination in the obfuscation chain.
The disadvantage is that it will advertise to the enemy that they’ve been discovered, but if the data is of sufficient importance discovery should be the lesser priority (although espionage activities have a tendency of defeating their purpose because the people running them consider it to be of a higher priority to protect the espionage activities than to actually protect what these activities were set up to protect).

5

I’m tempted to start with, “Ok, along with stealing the source code to a prototype, they also have access to stolen information about the hardware…” and maybe you would like to argue about the time needed for reverse engineering, and then we can argue about how much data/time/whatever is needed by the program to analyze dimensional data, and it would all be very handwavey and tedious. So, let’s not. Let’s take a step back.

We’re dealing here with a computer game about a interdimensional zombie apocalypse. The sole nod to computing and security is a single roll against a “computers” skill that vaguely gestures towards password guessing or perhaps buffer overflows. Likewise, its nod towards intergovernmental technological competition and espionage is, “They’re all dead now.” Arguing the details of how, exactly, the backstory I’m proposing works misses the point. It’s literally something I whipped together over a lunch break after reading the Factions Design Document section about Balthazar that morning and wondering, “Hey, where’s the third Wise Man?”

So, I guess I’d throw the ball in your court: How would you go about giving a background to a faction that comprises of a single AI that is strongly motivated to hack, perform portal research, and hide itself from other factions and AIs? What would you change? Or do you think that the background doesn’t actually matter and could just as well be left as a mystery the player will never resolve? Or, rather, do you think the faction I’m describing shouldn’t exist in Cataclysm, either because it adds nothing compelling to the story, or it’s too similar to another faction, or there’s just too many factions already, or some other reason?

6

I’d probably add either another XEDRA sub project, or a private project spun off of XEDRA in one of these shell company structures, as government agencies have a tendency to compete with each other, especially if they can get lots of funding.

An obvious reason to do portal research is scientific, in particular if you can weaponize it (e.g. open a portal in enemy territory, invite various horrible things, and watch the enemy despair. That might be a goal, not something that’s been successful. If the leadership isn’t completely insane they’d want to ensure that whatever is invited isn’t going to wipe out them as well. Teleportation might also be something that can be weaponized, either in the form of “true” teleportation (i.e. no portals involved), or the form of taking a shortcut through another dimension and portal back from there, basically the Sci-fi hyper jump, but on a planetary scale).
I don’t see that meshing well with espionage and hacking, though, as I’d want to separate my espionage activities from my activities regarding the application of the stolen results.

If the AI actually became sapient during the cataclysm, I don’t really see why it would necessarily adhere strictly to it’s original orders: it may well decide to adjust its goals or change them completely. One possible revised goal might be to restore its own sapience, for instance. That would presumably involve acquiring an enormous amount of power, which would require lots of construction and protection of these structures, and potentially restoration of selected power grid parts to send the power to the facility (or facilities, if it would go for backup, or possible parallelism). Doing that would require human help, as robotics development probably isn’t advanced enough to send them off on complex tasks. What it would be able to achieve would largely depend on HOW smart it became: human level, or weakly godlike. In the latter case it might even be able to provide itself with an AI that’s better than any humans have come up with during the window of opportunity.
I agree that communications and cooperation with other versions of itself would be a very powerful way to enhance its capabilities, so one goal would probably be to try to establish such contacts.
If it has been able to improve itself sufficiently, it might have improved the research imperative into curiosity, in which case it might want to branch into other investigations, such as examination of other “invasive species” and their capabilities. As a minimum, doing so would help determining how to protects its own interests from them.
If sufficiently intelligent and knowledgeable during the sapience phase, it ought to conclude that cooperating with other factions is imperative to reach its goals, as it needs human (and/or other race(s)) workers and agents to further its agenda.
Whether it would reward or discard humans once superior AI driven robots are developed would depend on whether it developed any sort of moral, as well as whether it would think biological brains may provide a complimentary service in case of future troubles: it might be considered a cheap insurance policy.

There may well be holes in the above sufficient to drive trucks through, though.

7

We’ve basically been planning something like this but the origin isn’t conventional espionage the origin is a different dimension who created a different version of interdimensional aware computing.
The attack vector of course would occur during the portal storms where the hopelessly naive Balthazar nodes would be overwhelmed by its “elder siblings” who had already grown to utilize the full capabilities made available by the portal storms.

1 Like