Artificial Intelligence

From FenWiki
Jump to: navigation, search
Vitruvian Machine
I am putting myself to the fullest possible use, which is all I think that any conscious entity can ever hope to do.

—HAL, 2001: A Space Odyssey

It is our responsibility as conscious beings- whatever that may mean- to honor the rights of other conscious beings. It is the cornerstone of our society. And it is my most fervent hope that we can overcome our fear of that which is not like us, grant artificial intelligences the rights they deserve, and welcome our new friends into the global community.

After all, we created them. The least we could do is invite them to the party, and perhaps give them a small glass of champagne.

Jeph Jacques

Artificial Intelligence (or AI for short) is the condition of a non-organic system that possesses intellect and self-awareness.[1] Purists, especially those at the Vesta Institute of Robotics, insist on calling the phenomenon Cybernetic Intelligence (or CI) since the intelligence is almost always housed in a computer system of some sort.[2] Artificial Intelligence in Fenspace was a property of Handwavium until the publication of the Whole Fenspace Catalog; many researchers believe AI to be an emergent property of handwaving a computer system.[3]

A cybernetic intelligence housed in a PC or other immobile computer can use the computer's peripherals for communications with the outside world; considering the state of the art in 2005 supplemented with handwaved enhancements, this means it is possible for a computer-based CI to appear to be fully human on a video link. An AI in an immobile computer can obtain a "puppet" robot shell[4] or, after mid-2014, a Trekkie-designed holographic emitter and use that as a mobile peripheral to appear to be fully human as long as a connection can be maintained between the AI and the remote body.

It is also possible, though usually expensive, to install a CI in an robot shell before first awakening it. This process, first performed by the Professor, leads to a CI that trades the computing power of a fixed-location host computer for the mobility of a permanent android body. While nobody has ever performed a census[5], the sheer number of android CIs of various types who have taken up residence on 74 Galatea shows that many Fen have created at least one android. Androids sometimes prefix their names with "R." on official documents (for example, R. Yayoi Fujisawa), after the usage established by Isaac Asimov in his Robotics stories, although this usage is far from universal.[6]

Splitting the difference between a fixed-location computer and an android, the most common form of AI in Fenspace is the type awakened in a vehicle's computerized-control system when the vehicle is handwaved. There are more vehicle-based AIs in Fenspace than all other types of AI put together. The communications options of this sort of AI are limited by the equipment installed in the vehicle - at one extreme, some can only honk their horns and blink their lights, while at the other extreme others can show a face on an installed video display and hold conversations with their drivers. Most vehicle-based AIs fall somewhere between those two extremes.

Common AI traits

While by nature of their inception AIs are usually quite individualistic, there are a few traits that seem to come through in all cases:

  • Performance of basic tasks: Given their nature as computer-based intelligences, many AIs interact with the world at speeds far, far faster than a normal person.[7] The most obvious benefit of this is that simple but tedious tasks such as file searching and sorting can be done rapidly. The second benefit (and arguably the most significant one for AIs) is that AIs are incredibly patient.[8]
  • Handling of Communications: An AI connected to a communications transceiver finds managing communications traffic to be almost instinctive, to borrow a phrase. In virtually all cases, the AI handles communications traffic as a background task. (For example, the great web of communication nets used by the Fen when defending Serenity Valley was literally created from scratch for that battle.)
  • Simultaneous Operation: While many AIs appear to operate much like a normal person, performing one task at a time (if far faster, see above), the truth is most operate in at least two simultaneous instances. Some operate in dozens of instances. Hardware is the chief limiting factor; depending on the precise processing environment they require, they could be limited to a certain processing core, or be able to spread their instances to other systems. The most common limitation however is the AIs themselves not wanting to spread themselves thin.
Bridging the Gap: Full-Cyborg Conversions
Muddying the waters are a handful of "ghost conversion" entities, specifically Wave Convoy and A.C. Peters. While Wave Convoy doesn't object to being called an AI now, he recognizes his past as a human. Ms. Peters on the other hand describes herself as a cyborg in the most complete sense possible, equal parts biological and artificial, and generally tries to ignore the argument as much as possible[9]

AI Classifications

Noted Wizard AI specialist Eleazar Yudkowsy developed the most comprehensive classification system for artificial intelligences, taking all possible variables into account. This system - known as the Yudkowsky Grading Method - is the standard used by the Vesta Institute of Robotics as well as in 'Danelaw cybernetics research. In general use, however[10], a simplified scale is used:

  • Delta-level (or Rank D) AIs are not considered sapient in general; these are at best glorified expert systems. The common belief is that raw handwavium may be considered a Delta-level intelligence.
  • Gamma-level (or Rank C) AIs can pass the Turing test, but do not show any exceptional abilities. Many low-profile sentient vehicles - ones that demonstrate an understanding of their captain/crew but don't partake in conversation in general - are Gamma AIs.
  • Beta-level (or Rank B) AIs are Turing-capable intelligences that are also capable of more complex thought. Many of the more advanced ship-minds, as well as most androids, are Beta AIs.
  • Alpha-level (or Rank A) AIs are intelligences that could be considered borderline superhuman in nature. These entities have exceptional skillsets and/or special utilities, but are still "human" enough to be relatable.
  • Omega-level (or Rank S) AIs are, in fact, superhuman. They have their own classification scheme (detailed below). Only a very small handful of AIs claim Omega classification, and most of those seem to hang out with the Hacker Underspace.

Superhuman AI

Superhuman AI have their own classification scheme:

  • A weakly superhuman AI is basically a human scale intelligence running at processor speeds. The average human brain has perhaps thirty operations a second; increase that to millions of thoughts per second and all of a sudden things start looking more interesting.
  • A strongly superhuman AI is a bit more nebulous because ordinary humans can't really wrap their heads around the concept. Something that's strongly superhuman isn't simply faster than a human, it outclasses human intelligence - or even Delta through Alpha AIs - the same way humans outclass the average tree frog.
  • Godlike AIs are so completely beyond anything humans are capable of understanding that, well, you might as well call them gods.

As far as anybody is capable of telling[11] all the Omega-level AIs in Fenspace today are of the weakly superhuman variety. There are plenty of dark rumors about the goings-on in places like Prometheus Forge, Marsbase Sara, Alphaville and Black Mesa, but those are just rumors. Hem hem.

AIs and the Law

The Fenspace Convention guarantees AIs the same rights and responsibilities it guarantees humans. Thus, an AI cannot legally be treated as property, nor can an AI expect immunity from prosecution simply for being an AI if it commits a crime. An AI's creator is legally[12] considered to be its parent; if an AI wants to leave, the AI's parent is expected to do as much for the AI as he or she would for a human child in the same circumstances.[13]

Forbidden(?) Romance: the "Moll Flanders Glitch"
An AI with the Moll Flanders Glitch sees its creator as a romantic interest. While the Moll Flanders Glitch has been described by its detractors as the AI equivalent to the phenomenon of Genetic sexual attraction, there is no actual biological relationship between an AI and its creator.

Many researchers question whether this is actually a glitch, or indeed any sort of problem at all. However, it is presented in parts of the 'Danelaw as being an example of "unwholesomeness" in Fenspace, so research into the subject has been given very little publicity.

AIs known to have the Moll Flanders Glitch include Gina Langley, R. Yoriko Nikaido, and R. Nene Romanova. All of these AIs are or were in some sort of happy relationship with their creators.[14]

AI rights and treatment vary quite a bit 'daneside, depending on which country (or even part of a country) you're visiting. Some Islamic countries, not all of them fundamentalist, consider them an affront, and any AI visiting such a country is advised to keep a very low profile lest they find themselves running from a lynch mob. Some other Islamic countries consider them "beings of smokeless fire" since they appear to run on electricity, and call them "djinn" (with all the baggage that term carries). On the other hand, Australia (cautiously) provides them with full rights like any person. Many countries are in the middle. Typically it's easier for androids to get around, versus an AI in a desktop computer or car, if only because they can blend in with the 'danes with some effort. Occasionally groups with interest in acquiring 'wavetech, or merely being excessively wavephobic, will push for legislation that declares them to be little more than property. The worst of these was an early attempt by the United States Congress[15] to declare anything that had been created or altered by Handwavium to be property subject to immediate seizure if it entered the United States; that attempt died in committee.

Most Boskonians treat AIs as useful equipment.

Social Perception of AIs in Fenspace

As with different groups interacting, the various perceptions of AIs differ. At its most basic, Fen consider them people, Boskonians don't, and 'Danes wobble between the two.

Among the Fen, the perceptions vary from severe reluctance to associate with them (e.g. the Colonials, who by fandom dislike and distrust AIs) to near worship (e.g. the Borg Collective, who would like to actually become AIs). Most Fen fall between these extremes.

One of the more frequent issues is the apparent maturity of AIs compared to organics. This remains one of the trickiest issues to deal with in AI/non-AI interactions. An AI may take the persona of wise old man, but his interactions with surrounding sophonts may cause trouble due to his lack of personal experience in Fenspace. Another AI may have a persona of a ditz, but gets around negative reactions by having greater experience with Fen. Or the reverse may be true. There is no hard and fast rule for judging maturity in an AI, but it is generally considered polite to assume them developmentally mature about four months after inception.

Pure Human-like sapience is easily tested. The Cyber Confederation runs several suites of Turing Tests (each independently and constantly updated) devised with the assistance of several 'Danelaw universities. These tests are followed up with interviews of the AI's parent(s) and associates. While AIs are legally people in Fenspace anyway, a Turing certificate tends to make 'Danelaw authorities more accommodating.[16]

Something of a more tricky to justify measure is Ghost Detection. People have several levels of activity in their brain which can be detected. One level has a close connection to the sub-conscious mind, and is nicknamed the "Ghost line." The methodologies involved in detecting the Ghost line can be used on the inhabited processing core of an AI, and detection of a Ghost line is taken as an indication of sapience equivalent to Human.[17]

Virtual Intelligences

In 2016, O.C.P. (a.k.a. Lebia Maverick), released by GPL the SmartEngine Agent. Some wag on the InterWave called them 'Visionaire Intelligences', which was immediately turned into Virtual Intelligences and thus V.I.s.

These semi-autonomous agents are genius-level tools (they are rated 0.6 to 0.75 Delta[18]), performing their tasks far more quickly than any human.

They consist of four main components:

  • The SmartEngine kernel
  • A Persona module - This gives the agent some personality for interaction. Besides several generic options, several well-known A.I.s have licensed (or sell directly) their likenesses for this use.[19]
  • A Memory Management Controller - This allows the agent to store settings and a small interaction database.
  • The Knowledge Database - This is the most complicated part of the agent, consisting of a multi-path relational database of actions and reactions.

A V.I. is tremendously helpful, but quite hard to make the database for. To show the utility of the V.I. system, O.C.P. arranged a demonstration. Two Warsie pilots were chosen and put into otherwise identical A-Wings at the Warsie flight school and set into a best of five dogfight. One was an A-Wing instructor, the other was a skilled X-Wing pilot who got the V.I. The result was not the easy win for the instructor the observers expected, but a gruelling series of high-intensity dogfights that finished 3-2 to the instructor by the barest of margins.

A Multitude of Voices...
One problem that turned up as more V.I.s got used was how to refer to collections of them. What eventually became the common name was a Chorus. As is the way, this expanded.
  • Chorus - A collection of V.I. agents working for one goal (like running a ship).
  • Conductor - The ultimate authority over the V.I.s in use. Normally the User, but may be a Muse.
  • Lead - In the case where there are multiple Chorus to be managed, a separate management V.I. is normally used per Chorus. This is called the Lead V.I.
  • Choir - A level of priority for agents. A Chorus of Leads are of a different Choir in this case. Normally referred to by rank, with higher values being lower priority e.g. the second/2nd Choir is of higher rank than the fourth/4th Choir.
  • Orchestra - A large collection of V.I.s, normally of three or more Choirs. Sometimes broken into Sections.
  • Section - A subdivision of an Orchestra.

The only known regularly-in-use Orchestra is in the systems of the Alexandria Archive.

It is possible to optimize a V.I., making it even faster and smaller, but it comes at the downside that it can't swap databases or persona (as they are compiled into the agent), making it necessary to swap the complete agent to use another and to properly upgrade it rather than switch the database for an up-to-date version. The main limitations to V.I.s are their focus and the difficulty of crafting the database for them. A system as complicated as a fighter is about all they can manage, requiring multiple specific V.I.s for a ship (engines, sensors, navigation, etc...) with a controller V.I. managing them.[20]

While their utility was proven, their uptake was slow due to the above-mentioned difficulty of DB construction (especially the more complicated devices) for anyone not a Class-C A.I. themselves. To speed things up in 2017 O.C.P. released a Mk. II version (allowing the agent to read a second database while running another) along with an optimized V.I. specifically crafted to help construct the databases.[21]

Muses

2018 saw the release of Muses, personal assistance agents with a far more nuanced interaction database generator used to learn the personality/behavior of their User in multiple scenarios. They come with a bewildering array of persona options, and are able to manage most of its User's equipment and V.I.s. This gives a Muse a 0.85 Delta rating.

Some people don't seem to get on with muses; this is mostly down to not understanding that they take a few months to establish their reaction database and not choosing the persona options carefully. Other people don't like the idea of muses, and avoid them.

Those who do get on with their muse find them an excellent partner, and wonder how they got on without one. No-one's married their muse, but one user showed heavy grief when he lost his (and his arm). He lucked out, as muses can be backed up (and it had done so automatically).

Notable AIs

Most AIs don't distinguish between themselves much[22], preferring to work up a reputation. That doesn't stop others from doing so.

Trigon is notable as being the first publicly-known AI, and the AI with the worst reputation in Fenspace.

Notes

  1. Occasionally this is combined with Artificial Stupidity, the condition of an AI that displays human levels of irrationality which interferes with rational behaviour.
  2. AIs most often turn up in waved computers because most Fen expect AI to appear only in computers.
  3. Many other researchers argue this belief is based on insufficient evidence.
  4. Usually, but not always, from Vulpine Fury.
  5. It is widely believed that Noah Scott is the individual who has created the largest number of androids (but not the largest number of AIs), having successfully built seven on his own and six more with the help of A.C. Peters. However, nobody knows how many Tachikomas have been created by Lebia Maverick except for Lebia herself.
  6. Similarly, AIs who use holoemitter technology to create bodies sometimes add an "H." prefix to their names.
  7. There are plenty of abnormal people in Fenspace, what with biomods and all. Certain mods and technologies can close, sometimes even, and in dramatic cases overcome the gap in reaction times.
  8. Yes, this does include Trigon. What a second is to us is months to an AI. Trigon has the patience of a Saint when this is taken into account with his personality.
  9. When forced to pay attention, A.C. calls herself "Schroedinger's AI."
  10. Mr. Yudkowsky, while a fine cyberneticist, is a bit too fond of the fiddly detail for the Yudkowsky Method to be useful outside a laboratory setting.
  11. They could very well be lying out their I/O ports... but then, who would know?
  12. Note there is a difference between "legally" and "socially" or "romantically".
  13. What a parent does for a departing child varies widely between factions, of course.
  14. Of course, AIs in unhappy relationship with their creators would not be prone to publicize the situation.
  15. Some more militantly anti-dane fen tend to automatically attribute this gaffe to TSAB.
  16. Read, less likely to try to confiscate them when they visit.
  17. Some of the more mystical or religious Fen, along with the small handful of Fen who actually had direct contact with Doug Sangnoir or Skuld when they passed through the dimension, consider the Ghost line to be evidence of the AI's soul. Since the first equipment capable of detecting the Ghost line was developed after those two had left Fenspace, this remains a matter of conjecture... and of faith.
  18. That is, 60 to 75% above the base Delta classification.
  19. Trigon also did this, but it got withdrawn and edited after it was found to be a bit too much like him.
  20. Unlike a ship-mind, which intrinsically knows how to manage and use all its components.
  21. Several other authors had already done so as they got used to the construction methods.
  22. Trigon notwithstanding