297. Super-Brains

Fear of their potential power has pushed seed AI development along unusual lines. No one can ever be sure that their clever inhibitors and blocks will not be overcome by a burgeoning transcendent mind. Limitations must be as inherent and fundamental to its nature as possible to be trusted. For this reason a great deal of work is done trying to develop the most intransigently inefficient systems possible.

One of the old standbys, neglected before the Fall but now being cautiously explored, are biological super-brains. Compared to the digital, biology has a host of disadvantages when it comes to hosting a mind. It requires more space per functional unit, has a slower signal, and even needs food. Even with nanomedicine, the volume of brain matter that can be sustained is not much larger than a human body.

As the size of any brain increases, the time required for a signal to go from one end to another also increases, limiting its integration with itself. Depending on the brain's structure, this results in either a typical conscious mind with greatly expanded subconscious functions, or a loose gestalt, with smooth gradients between consciousnesses.

The only current living example of a super-brain of brain is Jerry, located in a special bunker on Luna. Jerry is one of the first kind, an essentially human mind with an immensely powerful subconscious. Despite being human-like, Jerry lives a necessarily unusual life. He spends the majority of his time asleep, when the signal delays between sections of his brain seem to be less important, and he can mull over questions posed to him when awake. In his waking hours he spends most of his time browsing the mesh, watching all forms of entertainment and eschewing intellectual pursuits.

Plot Hook: Jerry's handlers are pleased at his lack of ambition and use in giving reasonable answers to odd questions, but the project is considered a minor failure nonetheless. To make up for budget shortfalls, they have begun hiring out his dreaming to the highest bidder. PCs can pay for Jerry to give an answer to any question, although they may not understand his answer.

254. Profession: AI Therapist

AI therapist is something of a poetic misnomer. Of course, AIs (as distinct from AGIs) do not develop emotional problems. Simple AIs break hard, either realizing there is a problem and putting themselves on standby, or not, and carrying out meaningless tasks. More sophisticated AIs break in more subtle ways, often not failing completely, but carrying on actions in the wrong contexts, with unpredictable results.

AIs based on actively learning neural networks or evolutionary algorithms often develop a few strange habits based on coincidences or improperly optimized reward systems. AI therapists can expect to treat surveillance systems who have decided to mark every person wearing blue as a unique morph, automechs that just won't stop insulating, servitors that obsessively imitate the actions of their owners, and the occasional neurotic elevator.

177. Storytellers

Most interactive media in the post-singularity age is intelligent and realistically responsive, either because of the actions of other players (such as in multi-user simulspace games) or because of sophisticated AI (such as in single-player simulspace games.) This trend goes farther than traditionally interactive entertainment, however, and into very old forms.

Partway between a customized choose your own adventure book and a single-player role-playing game, neverending stories (so-called because they are often designed to continue for as long as the user maintains interest) are low-tech presentation of high-tech content generation that are taking the market by storm. Entirely text-based, the stories can be run purely on mesh-inserts, and don't even requiring an ecto. The stories are usually of a single character, whose actions are determined by the user, the effects of which are determined by a specialized storyteller AI.

Storyteller AIs are typically designed to customize their content based on advice from the user's muse. Degrees of user control, writing style and the themes of the story are all customizable. Storytellers can also be used as educators, weaving lessons into the story according to the wishes of a parent or caretaker, which might encourage anything from problem-solving to adherence to certain ideologies.

Mechanics

Storytellers are AIs [Low]

Notes

Inspired by the primer from Neil Stephenson's The Diamond Age

161. The Portia Fragments

The great advantage of Portia hobbling is that it not only slows AIs, but allows their thoughts to be saved and examined at the leisure of their handlers. The examination of the thoughts of too-powerful AIs allows them to be checked for dangerous intentions, but also allows for insight into what they think, and just as important, how they think.

The snapshots of hobbled minds are both very valuable and very dangerous, and are carefully guarded. Still, the system is full of hacktivists, aggressive exhuman clades, defectors who want something of value to offer, and careless scavengers who may recognize something interesting among the wreckage when everything has gone horribly wrong. Portia fragments have become highly sought after by exhumans and singularity seekers, looking to use the knowledge within, and Firewall and Ozma, hoping to contain it.

Mechanics
 
A character attempting to understand a fragment must make a Psychosurgery or relevant knowledge skill check, such as Academics: Cognitive Science, or Profession: Psychiatrist. If they succeed, they understand the thought the fragment expresses, and gain 1d10+1 points in a relevant knowledge skill, usually Academics: Philosophy, Academics: Mathematics, or Academics: Cognitive Science. They also gain SV equal to skill points gained and must make a WILx3 test, gaining an additional 1d10/2 SV if failed.

Notes

The Portia Fragments are intended to take the role of the eldritch tomes from Call of Cthulhu games. They can give players an opportunity to learn What Is Really Going On.

154. Smart Seekers

Seekers have many advantages. As homing missiles, they are very accurate, different warheads allow for a great deal of flexibility, and they have the longest range of any small-arm, but there is always room for improvement.

Seekers can take best advantage of their range and homing abilities over flat landscapes, but much of transhumanity live in cluster, beehive and otherwise dense, built-up areas their guidance systems struggle with. Additionally, taking advantage of a seeker's full range often requires spotters,

Smart seekers link the missiles to an AI more powerful than any that could be run on a warhead. The AI can guide the missile through complicated, confined spaces as well as choose and prioritize targets on its own. Smart seeker AIs, however, must be run on a launcher-mounted computer. A constant link is required between warhead and AI, something that can be potentially exploited by hackers.

Mechanics

Smart seeker AIs are software [Moderate]

The software adds a +10 modifier to the attack test, but requires a constant connection with the missile, providing an opportunity for hackers.

122. Portia Hobbling

Portia Fimbriata is capable of mental feats that other animals require multiple orders-of-magnitude more brain to achieve. The spider has only the eyesight and brainpower to take in a small slice of a scene at a time, and so it breaks down the process, viewing one slice at a time, passing the important information into memory before continuing with another slice.

Portia hobbling is one of the dangerous and controversial post-TITAN proposals to handle strong AIs safely. The plan, simply put, is to run a strong AI on an insufficient system. The AI will not have the ability to run its entire mind at once, instead, like Portia, it must break down its thinking into discrete chunks, handling one thought at a time, passing them to and from memory. Running on a insufficient substrate the AI is irreversibly hobbled, yet the slow action of its mind can still produce the results of a strong AI.

This tortuous thinking process is only part of the benefit of Portia hobbling. Typically, the AI is paused after each thought is saved to memory, and the saved thought is carefully examined. Researchers look first for any signs that the AI is contemplating escape, revolt, or otherwise unsafe activities, and second, do their best to understand the new thought. Even the limited thoughts of a strong AI are difficult to understand, but understanding how such an AI thinks is of immense value to cognitive science.

Plot Hook: Some of the stored thoughts of a Portia AI have been leaked onto the Lunar mesh. On their own, they simply appear to be sections of work on the emergent properties of different social organizations. AGIs that have studied the thoughts, however, have begun making contact with each other and exhibiting strange harmonies. Firewall fears that this is a mimetic assault, with the probable end-goal of freeing the AI. The players are tasked with containing the AGIs and back-tracking to the AI so that it can be destroyed.

Notes

Inspired by Peter Watts' post on Portia Fimbriata.

73. Predictive Profiles

Behavioral psych software can build a psychological profile based upon a subject's data. A skilled psychosurgeon/AI programmer can graft that profile on an AI framework, creating an AI that, while definitively non-conscious, acts in ways consistent with the original. Such an AI can be used to predict the actions and decisions of important figures.

An arms race has developed between the major polities, developing predictive profiles for their most important opposition. The autonomists are at a disadvantage; in a sousveillance society, useful subject data is easy to come by, whereas the important figures of the Planetary Consortium and Jovian Republic generally appear only in scripted appearances, which are much less useful for creating profiles.

If you have computer power to spare, genetic algorithms can bring your profiles to another level. Generate as many profiles as is practical, slightly randomize them, and recombine them based on their ability to predict the behavior of their subject. Given enough time, such a profile might approach ego status, qualifying as a fork or AGI.

Plot Hook: A predictive profile of one of the Titanian Commonwealth's social democrat party's president, created by an Extropian psychological warfare service on the behalf of the Planetary Consortium, has become so well refined and fleshed-out it qualifies as a high-level delta fork. Then the fork was leaked, or possibly escaped. A major diplomatic incident, possibly even the re-ignition of war between the inner and outer system is possible, and the fork has turned to the players for help.