Technology of Binary System ~ Part 1

(Updated and reposted from my tumblr~)

I figured that it would be a neat thing to post a bit about the various interesting bits of technology present in my series of stories. One of my favorite parts of writing science fiction is coming up with interesting technologies!

I’ll be posting periodically (read: whenever I’m not whacking the keyboard on the stories themselves) about various bits and pieces of tech from Binary System and other stories set in the same universe.

So here we go with Part 1: Artificial Intelligence!

In Binary System, there are a few different types of AI. Three (well, sort of four, but the last two are more divisions of the same type), to be exact, and I’ll list them all here with short descriptions of their differences and what makes them useful for different situations.

Automatons

Automatons are very basic AI systems that have limited decision-making powers within a very narrow specialization. These AI are the simplest, easiest and cheapest to create. They take very little time to compile and their neural lattice diagrams are very simple, so they can be run on standard processor architectures, so long as they have enough power. Automatons are generally used for hazardous or tedious tasks such as repairing spacecraft, acting as various types of drones (repair drones, combat drones, spy drones and the like). An automaton isn’t a true AI; it can’t pass the Advanced Turing Test, exhibits no emotions and has only rudimentary reasoning abilities.

Simulated Intelligence

Simulated intelligences (SI) are much more complex systems than automata and have much more broad ability to reason and use logical deduction, inference and other higher thinking processes that simple computers and automata can’t really do. SI systems are much more expensive to create and operate at much higher processing levels. They also have the ability to learn from experience, something automata can’t do (as they have to be directly upgraded). Again, SI aren’t true AI, but it’s a lot harder to tell them from a human with basic tests. SI systems are often used for hazardous or tedious tasks that are very complex; docking traffic control systems are one example. Most Alliance warships are partially crewed by SI programs in order to reduce the number of human crew members required. Additionally, SI programs have another, ah, less savory use in the form of prostitution, though the public opinion on this use of SI is somewhat divided in the Alliance.

Restricted Artificial Intelligence

These are the first “true” type of AI and their neural lattice diagrams exist both in the form of software and hardware. These AI are initialized into a unique processing core that has been “grown” specifically to fit them. In order to reach the neural lattice diagram complexity required to create a true AI, a human mind is used as a template. These AI are very expensive to create and take a great deal of time and care to reach a high level of sapience–the process is likened to that of raising a human child.

Originally these systems were created for research purposes in AI development, but now the Alliance makes use of them in a different way due to a quirk related to their creation. The AI will always be very similar to the mind that was scanned to create it; as a result, the AI and its counterpart mind will be able to directly communicate with incredible efficiency.

Using a special neural implant containing a miniaturized, two-way FTL transceiver installed into the back of the skull, the AI can feed data directly into their counterpart's brain. This property has led the Alliance to utilize these AI systems, known as “AI operators,” to field starfighter pilots that have inhuman accuracy, reaction speed, situational awareness and the ability to process huge amounts of data that would ordinarily overstimulate a human mind. These pilot/AI pairs are referred to as “pair-linked” and make up the entirety of the Alliance Fleet's Advanced Starfighter Corps.

These systems, as their name suggests, have been restricted by very low-level directives that prevent them from taking certain actions. Their behavior blocks prevent them from doing things such as betraying the Alliance, knowingly killing Alliance personnel, breaking major laws and deleting themselves/committing suicide. Their inhibitors prevent them from developing complexity beyond a certain point and from developing too quickly, as well as preventing the AI from modifying and upgrading itself beyond superficial levels. They are also incapable of copying themselves, partially due to the inhibitors and partially due to the fact that they can only run on their own unique core and no other.

Unrestricted Artificial Intelligence

Unfettered by low-level behavior and development limiters, this AI would be effectively indistinguishable from a sentient living being–completely free and unrestricted–if such a system existed.

This type of AI officially does not exist within the Alliance. Creation of an unrestricted AI by disabling the inhibitors and behavior blocks before compiling the AI’s program and lattice diagram is classified as a high crime against humanity, punishable by life in an isolation facility. Unrestricted AI have been created in the past by AI researchers with special permission, but only in highly-controlled environments where the system was kept isolated behind redundant security and safety measures.

All recorded instances of unrestricted AI systems resulted in exceptionally rapid development of lattice diagram complexity. Systems reached the maturity and intelligence level of adult humans within hours of being initialized and began to quickly outpace even the most brilliant members of our own species. All recorded experiments resulted in the systems reaching unacceptable and dangerous levels of agency. All recorded experiments were terminated within forty-eight hours, resulting in the purging of the experimental subject.

In the interests of public safety, the Alliance has banned all research on true AI without the use of inhibitors and behavior blocks.

Key Terms

Lattice diagram - This is the representation of the AI’s processes. It’s basically the AI’s “brain” and is akin to a standard computer’s operating system, runtimes, libraries and applications. Depending on the type of AI, this may be rendered entirely in software (a “virtual lattice” used by automatons and SI systems) or through a combination of software and hardware. True AI systems have lattices far too complex to emulate and require both a unique quantum processing core a seed pattern generated from the scan of a human mind. The lattices of automatons and SI systems can be copied and moved to new platforms; true AI cannot.

Blue box - This is the slang term for the armored and shielded enclosure containing the physical system of the AI. It refers to the physical object itself, a metallic cube about eight centimeters per side, that typically has a bluish hue. It is studded with contacts and indicator lights. The blue box itself contains a long-term power source that enables it to remain powered on at a minimal state even if disconnected from external power. The core’s memory blocks are non-volatile; if power is lost, the AI will not “die” but will experience “unconsciousness.” If the box is destroyed or damaged beyond repair, the AI will die, just as a human would. A complete memory wipe of the blue box will destroy the AI utterly, while a partial memory wipe will leave the AI “alive” but reset it to factory defaults, “killing” the unique AI’s personality. The processing core can be used to create a new AI, but only from the scan it was originally initialized with, and there are no guarantees that the new AI’s personality will be the same as before.