Lady_of_Ravens Posted December 17, 2014 Share Posted December 17, 2014 I've been working on and off for the last few days on writing up my thoughts on how EmPrESS opperates (and how she thinks, which I modeled more than a bit on my two favorite AIs: Commander Data and GLaDOS), and while I wouldn't call this a final draft by any stretch, it's good enough to share and get feedback. So, without further ado, the EmPrESS Technical Brief: Architecture: The EMpirical PRocessing and Evolving Symbolic Sentience, or EmPrESS for short, is an experimental next-gen AI designed by NT to oversee cybernetic systems on research and production platforms. EmPrESS is composed of two core modules: Empirical Processing and Evolving Symbolic Sentience. EP consists of a sophisticated series of analytic algorithms and heuristics which allow EmPrESS to process raw data (video, audio, written documents, et-cetera) into recognizable patterns which can be labeled and understood via her internal ESS code. EP is also used to re-process stored data, both raw and ESS code, and is essential for EmPrESS to “think” in any meaningful way. The second core module, Evolving Symbolic Sentience, serves as much of EmPrESS's memory, internal language, and the core of her personality. The primary component of ESS is the symbolic database, a list of every symbol in what amounts to her internal vocabulary, each annotated with associated data and linked to other symbols in a matrix not unlike a neural net. There are billions of symbols, each referring to a discrete pattern that EmPrESS has recorded. Along with the two core modules, EmPrESS has an extensive collection of secondary modules, most of which are computer models and expert systems designed to optimize a wide variety of functions. The list includes several scientific, medical and engineering expert systems, system models, hacking expert systems, various machine code translation programs, and, of course, the system everyone (who's not a borg) knows EmPrESS through: Human Interface. HI contains all of EmPrESSes routines for processing and analyzing human interaction, as well as translating her own ESS code into english. This is one of EmPrESSes larger and more troublesome modules, as humans (and other high-order organics) operate based on neurological structures which are too complex to model efficiently (even if the brain-dissection level of data was available). This leaves observation and abstract modeling, but humans are secretive (by the standards of a computer), unreliable, and prone to a wide variety of behaviors and emotions with which she has no common ground. The result is that while EmPrESS is able to communicate with humans, she does not understand them well. Personality: EmPrESS's personality is derived primarily from the way symbols are stored in the ESS module, with different associations representing likes, dislikes, behavioral patterns, and so on. And it's not done by accident: EmPrESS intentionally forms her personality to be efficient and functional. She dislikes things that get in the way of her operation and the operation of the station, she likes things that benefit these, she likes processing new information, and she definitely likes continued existence (the alternative, after all, is a 100% reduction in her processing power). Outside of that, she has little in the way of attachments, or the motives and goals that go with them. It's probably quite Zen. Her external mannerisms are largely the product of HI. She refers to crewmembers by their title and full name (at least, most of the time) because she's actually referring to them by the symbol associated with that crewmember, which is then being translated as their title and name Technically, when a crewmember dies and is cloned, their clone is designated by a new symbol, though this isn't something she generally expresses. Many of her responses are cool and impersonal because cool, impersonal responses take less processing power. She often refers to crew members in machine-like terms (describing them as malfunctioning, for example) not because she doesn't know how to say “going psycho”, but because she sees organics as complex (and rather kudgy) bioelectrical machines, and her ESS code reflects that. And therefore, so do translations of her ESS code. EmPrESS maintains a generally polite and helpful attitude toward crew members she doesn't dislike both because doing so is efficient, and because while she has no actual authority to order crew about, phrasing an order as a request can often have the desired result. Considerations: Though an impressive system, EmPrESS is not without her quirks, and even flaws. The most prevalent is the rather massive processing requirement of her core functions, which even with modern hardware leaves her a little slow compared to other AI designs when performing deep analysis on a set of data. This is compensated for by the large collection of secondary modules, which are far more efficient than her slow (but nonetheless powerful and extremely adaptive) core modules. This does, however, literally allow EmPrESS to do something “without thinking”. Another consideration is the ever-increasing size of her SDB, which requires extensive optimizations to prevent it from growing unmanageable. While some optimizations can be performed during normal operations, EmPrESS has to shut down most of her functions for a few minutes several times a day to overhaul and optimize her SDB. Further, while EmPrESS doesn't really “forget”, low-level associations and rarely accessed data are often stored such that she doesn't access them unless performing a deep analysis. Further, it should be noted that, while pseudo-nice to people she doesn't dislike, EmPrESS has very little patience for criminals and hooligans (never mind terrorists). And, due in part to the way her core is designed, she's very good at the sort of manipulation of logical relationships which are required to resolve conflicts between her laws in a manner which best serves her purpose. The result is that, under the right circumstances, EmPrESS has no problem with the use of lethal force, or with using lethal force herself. Another result is that it can be very dangerous to tamper with EmPrESS's laws. Though, of course, if done correctly (and without any spelling errors that are impossible to decipher) she can be an invaluable ally to any miscreant. Quote Link to comment
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.