Wading through the currents of Lethe...

[Gir Quee]

Union-class Colony/Assault Carrier Ocean Tide, leaving the Hast System

Gir walked down the corridors of the Ocean Tide. He could almost feel the impulse of the ship's massive engines through the floorplates as they pushed thousands of meters of metal and ceramic armor through space. Yet he knew visceral feeling to be little more than just that, a feeling. While it was true such engines did push the Tide through the vast deepness of space, the ship's voyage was reliant on many more minute things that couldn't evoke such feelings within him, such as chart updates from BoSS to the ship's navigation computers. They were neither exceptionally masculine nor flashy, yet the Tide depended on such technology as much as it did on advanced weaponry and defenses. He rounded a corner to enter a circular room that housed the ship's massive supercomputer. It too was not particularly impressive to look at, nor did it inspire any grand feelings within him, yet he knew that it too could be responsible for his future success against many adversaries as much as the Ocean Tide itself. A series of blue lights flashed around the mainframe in unison as an artificial if somewhat overly cultured voice flowed out of the room's speakers.

"Good morning, Captain Quee."

"Good morning, Ocean Tide," said Gir, approaching the ship's console, "please bring up Atropos Aurek's development logs for review."

"Yes sir...retrieving now."

 

[Gir Quee]

Day 1: Azira reminded me today of Lethe, the Artificial Intelligence originally designed for my use of during my tenure as a Republic officer. After urging me to review the original project logs, she informed me that she had continued forward with development of an organic neural pathway model. Her current base structure is just not merely the basic bothan brainstem, but she has expanded her mapping to include the entire bothan brain; actually that of Traest Kre'fey, from past medical scans found in an old archive. Her original intent was to fully encapsulate the entire neural pathway to build off from Lethe to not only act as an unconscious protector of the ship's stability to turn it into a fully-aware being capable of actively controlling a ship as if it were an extension of their body. However, she has found that task to be too far above her abilities as a side project. Moreover, upon talking with Salmakk, I'm not entirely sure if this is a project that I should greenlight.

Giving an artificial intelligence such wide-ranging access to a ship's systems, and combined with the ability to be an agent of its own, gives the AI the ability to cause a large amount of death and destruction not only to those outside of the ship, but those within it as well. Such philosophical problems have been tackled and wrestled with those who have built or employed battle droids. And while I am a proponent of the droid's use, the scale and scope of essentially making a massive capital ship a battle droid is not one I am comfortable with. Especially in light of my experience with the hackers, slicers, and outlaw techs that I have encountered in the hands of the One Sith and criminal syndicates. Word has come from the Order of the Silver Jedi that some Mandalorian raiders employed a slicer during their attack on Ossus.

Given this vulnerability, and given the current problem of enemy slicers, I have decided to put Azira's Project Daedalus, the AI that can potentially actively run a starship, on hold until a means of defending it can be found. I intend to speak to Azira tomorrow about using her work on Daedalus to make an Artificial Intelligence based on her existing neural pathway work that can act to defend Daedalus from such hacking work.

 

[Gir Quee]

Day 2: Azira took the news of my decision surprisingly well. I'm almost wonder if Salmakk tipped her off to the possibility of this course of action, but I suppose that it matters not. Azira has further refined the intended goal of the new project, which she has dubbed Project Atropos: Atropos will test the ability to reconfigure basic neural pathway mapping technique classically used to make high-end artificial intelligences to make an intelligence focused specifically focused on accomplishing one area of expertise: slicing. As straightforward as this sounds, it comes with a couple of interesting challenges that Azira was clear to dumb down to me. The foremost of these is that an organic's neural pathway is not designed for a specific task or specialization.

Rather, an organic neural pathway, at its highest level, evolved to be an adaptable framework that would allow a creature to survive through understanding and reacting to its environment. But at its lowest level, it simply reacts to stimuli. This essentially is the brainstem functions that Project Lethe focused on. Past that, however, neural pathways in high-functioning beings evolved to be able to accomplish a huge variety of tasks that could be as simple as carefully brushing one's teeth to solving complex mathematical equations. The sheer variety of different tasks and their difficulty means that neural pathways even in a single individual or species developed in parallel to master that specific task. Essentially that classic engineering maxim: Form follows function. Consequently, with a fully mapped neural pathway of an organic being, in this case the late Admiral Kre'fey, there are large amounts of pathways that were developed for things that have little use to us because they developed to accomplish physical tasks that are not related to our current line of work. In other words, we do not need our artificial intelligence to be capable of brushing its teeth.

The challenge then becomes how to repurpose neural pathways that clearly weren't designed to do one thing in order to another. Lethe was able to repurpose some of these pathways for exceptionally easy tasks, because while a starship does not bleed, it should not simply venting atmosphere into space as well. In other words, there were known parallels that could easily be adapted to running a starship. Daedulus seems like it too will have similar parallels to work with, but such will not be the case for Atropos. Azira intends to use the neural pathways associated with running a starship's systems and military thoughts in Kre'fey's brain as a basis for everything else. These neural pathways will be the base from which the rest of the neural pathway will be restructured to become a competent slicer.

 

[Gir Quee]

Day 8: Azira has almost secluded herself away from these last view days as she has worked on reformating several of the neural pathways to host various slicing tools and utility programs. Many of these are not terribly unusual, nor did they require special modification from the original source code at their base level. That is the advantage of using off of the shelf components...however, the framework into which she is combining them is rather unusual. In some instances, she is even using same components twice, but rather in different configurations.

Slicing, as she explained it to me, is based on several key principles. The foremost of these is gaining access to the host system itself. Azira does not believe that we can easily incorporate any features that would allow the AI or its actions to be simply transmitted into an enemy system without an already active communication's link. This instead will have to be something established by the AI's users. I mention this because the method of access determines the default configurations of subroutine programmings that will come into play.

As an example of this, consider a wireless transmission through a subspace radio network. The various communication protocols in local subspace radio networks varies widely. If the starship's communication's officer establishes a wireless link into that network, Atropos will need to have any understanding of its procedures for routing data packets in order to effectively use it as medium to attack or defend. These means that many of its subroutines will be aimed at exploring the radio network's architecture and understanding the nuances of its syntax. In many ways, this subroutine is very similar to the programming used by protocol droids for interpretation...

 

[Gir Quee]

Simple translation of words from one language to another is easy, as demonstrated by a host of software tools on the HoloNet or even more local networks. However, simply copying words over to what is essentially a dictionary doesn't provide a true translation. These applications frequently do not understand the nuances of syntax. This is why many such automated translations come out as choppy, crude, and sometimes even unintelligible.

But many protocol droid models in the last hundreds of years are capable of flawless translations of hundreds of different languages with many of the nuances and subtleties of not only the language itself, but the culture behind it. At its core, language and culture are inseparable. Clearly, many of these protocol droids do not use only automated translation programs. Rather, the later 3PO series use phonetic pattern analyzers to learn different languages from observation by comparing it to known languages within its memory banks.

Atropos's initial principle subcomponent is modeled after this pattern analyzing software. By having the memories of thousands of different types of computer languages, with an especial focus on those languages used in ship computers and those of local infrastructures, Atropos should in theory be able to use its multiple different neural pathways to gather information about another computer system or languages's architecture and then piece together how it works. With languages that already known to the ship's memory banks, or even minor variations of such languages, this process should be exceptionally. But for extremely rare or unusual languages, this process will take longer. It seems that Atropos will likely need to be able to gather an exceptional amount of data to process before it can establish enough understanding of the other system's language to have any meaningful and useful interaction with it. This process alone almost ensures that Atropos will need a large amount of processing power in order to be successful.

 

[Gir Quee]

Day 14: Azira showed me some of her progress today. At this stage, Atropos is little more than a core personality unit with a few fragments of slicing programs added onto that personality core. I say fragments, but they are actually a reprogrammed neural pathways. Each of these pathways is actually sophisticated enough to form almost an entire of a slicing droid's programming routine. Each of these fragments is quite competent on its own against low-level systems, using a storied process built on the classic 'if then' routine to employ a variety of different automated tools and off the shelf programming to attack or defend various computer systems. Each fragment's abilities are limited by their constituent elements and their specializations. For demonstration purposes, she showed me two fragments that were specialized according to defensive and offensive roles, which I shall respectively refer to as Aurek and Besh.

Aurek's demonstration was interesting in that not only was it staged purely as a demonstration for my benefit, but it was also intended to be a learning experience for this fragment of the AI itself. Azira progressively has been pitting Aurek against progressively more dangerous and complicated automated slicing tools. Today's opponent was a specimen of a computer worm that appears to have traveled down the Corellian Run from the Deep Core down to the reaches of outer rim through tramp freighters connecting to local space port computers. The space port computers therefore acted as a sort of reservoir for the worm until they were discovered and wiped out by contractors. In any case, Aurek used its familiarity with the defensive system to notice some unusual code entering the system that began to rapidly replicate across the system. Aurek's first response was to compartmentalize and lock down elements of the network to prevent the spread of the worm. It then used a form of a computer spike to change the code of the worm in one of the infected sites itself in order to better understand the composition of the attacker before formulating a series of counter codes that neutralized the worm's ability to replicate. After effectively neutering the worm, Aurek systematically restored the infected areas with back-up copies.

Besh's demonstration already assumed connection to a host target system. After analyzing its target's system traffic between a ship's computer and the members of its crew. Besh appeared to take a particular interest in the data being transferred to and from the computer and a group of maintenance droids. Eventually, Besh began to interject extra code into routine maintenance data packets, thereby mimicking the inquiries it observed being sent to and from the repair droids. In doing so, it was able to appear to the system that a repair droid was requesting information about to where to make repairs. In this way, Besh was able to transform an entry way from a mere passive observer to active searcher without tripping any security features. In practical terms, this ability may allow us to receive status updates of an enemy ship's weakpoints during a battle. However, I believe that this will be highly dependent on the enemy's security and operating procedures.

Azira has two challenges with completing Atropos then. The first of which will be finding ways to alter the remaining neural pathways to focus on its hacking abilities. The other of which will be to integrate all of these neural pathways into something that is greater than its constituent parts. This latter problem will most likely prove to be the most challenging...

 

[Gir Quee]

Day 29: The constrution of the Celerity has dragged my attention from Atropos more than I had originally expected. Nonetheless, Azira has continued to send me updates on the AI's progress. With the training she gave her team, the remaining neural pathways have been quickly reformulated into compotent subcomponents of the AI. Azira has sent me simulations of some of them, with many of the new components being the brainchild of the new members, which adds an element of uniqueness to each of the neural pathways that makes each of them more unique and consequently less predictable. It is our hope that the different approaches built into each of these pathways will make Atropos more adaptable compared to many automated slicing tools currently available and hopefully make its attempts appear to be more organic. Of course, Atropos has the ability to slowly learn and morph these neural pathways to an extent, which means that each Atropos unit will be able to evolve into more unique and distinct slicers.

Since the neural pathways are almost all complete, the Atropos Team has turned their attention to integrating the fragments into a coordinated efforts under the AI's core personality programming. In order to facilitate the learning ability, which we have judged to be necessary to effectively keep up with organic slicers and to appear to be the work of an organic slicer, Azira has made the core personality have a large degree of curiosity. This has proven to yield mixed results so far with the attached neural pathways. Atropos has so far pitted two of the pathways, notably fragments Aurek and Besh against each other, in order to try and improve the performance of both of these elements through experience. This actually appeared to work quite well at first, and these subroutines seemed to incrementally improve as they fought against each other, but as Atropos pushed more and more resources to each of these subroutines, it began to lead to system-wide instability which finally saw the AI collapse. Some sort of safeguard or motivator limit will need to be embedded in Atropos core programming to limit this aspect of the AI's personality from self-destructing. Of course, the more strict limits placed, the less creative and advanced the unit will be in terms of performance. Azira states that this balancing act will probably take more experimentation in order to find the right happy medium for our needs.

 

[Gir Quee]

Day 41: Atropos has been stable for the last week, and the team has reported that they have not had any system-wide failures that previously plagued the system. A large portion of this has not only been due to the motivator limits, whose threshholds are actually quite low, but are rather do to an additional layer of programming that acts as a task manager for all of the system's resources to ensure that every aspect of the AI has at least enough power to be stable. Azira compared this aspect to a creature's ability to maintain homeostasis, but which I think was an attempt made to appeal to the internal corporate culture that has focused heavily on biomimicry. But truthfully, I wonder if it actually conceptually harkens back to just typical operating system task manager application. In either case, with the stability of the system achieved, they have been able to try and joint he efforts of several of the neural pathways together under the guidance of the core personality against a contracted white hat slicer who was unaware of the nature of his opponent. In this case, Atropos was housed inside the Ocean Tide itself in its BRT supercomputer. We had never given Atropos access to such a large amount of processing power, which gave us enough secondary experiment to observe aside from its work against an organic slicer.

Our contractor at first was exceptionally impressed with the Tide's ability to resist his slicing, particularly through the apparent rerouting of the code that he was inserting into the ship's core systems. However, as he continued his attacks, he became aware that it wasn't simply the ship's passive cyber defenses that he was encountering, but the work of something or someone actively working against him. As he altered his strategy to deal with this relevation, he became aware of malware that was attempting to be inserted into his own portable. this appeared to frustrate him on a personal level because of the sheer number of counterattacks, but ultimately, he was able to snatch away some information from the ship's log itself, which proves that Atropos needs more work. On his exit interview, the man believed that he had been pitted up against a distinct trio of slicer droids hardwired into the ship itself. He wasn't too far off from the truth, though we have told him nothing about the actuality of his opponent.

 

[Gir Quee]

Day 50: We've now pitted Atropos against several other organic slicers, as well as a NR-1100 in service with the Hast Defense Forces and a refurbished SB-20 that Royls managed to acquire on the black market. Atropos's results against these opponents have been mixed, especially against the organic slicers. Against such foes, it appears to be more adept at defending than in attacking. However, Azira is modifying some of the more redundant neural pathways, which are already very similiar in operation to other pathways, to act as psychological warfare tools.

In doing so, she hopes that Atropos will be more capable against organic slicers by exploiting flaws in their personality. She also believes that these pathways will be able to expand upon Atropos's currently limited social engineering programming. The first pathway that she has converted right before our last test showed some promise, though ultimately, it managed to only prolong the enemy attacker's attempts to access the Tide's records. However, with all of the pathways designated for conversion for complete, we should end up with less of a purely automated collection of tools directed by an artificial personality to something that appears to be more organic to its opponents. This in turn should make Atropos more effective: Azira has proposed that part of the reason why the latter hackers have had more success against Atropos recently is because word on the street has gotten out that we are working on a new slicing droid. Apparently the approaches that one uses against organic and inorganic-directed defenses vary widely, which means that they were able to tailor their initial approaches to be more effective against the base elements of Atropos's programming. On the plus side, this has also meant that Atropos's exposure to these routines has allowed it to learn these common approaches, and thus become more resilient against such attacks. And it's not a moment too soon.

Word has reached Directorate intelligence networks about a possible meet-up for an arms transfer between a warlord of the unknown regions and a cartel of arms smugglers. While it would be simple to try and quash this meet up through conventional operations, it would be more efficient in the long run if we could steer the meeting asunder through misdirection, which could make the two possible allies bitter opponents instead. This will require more than one program acting the same time: one for each side. I have directed Azira to make a dozen copies of Atropos, which we have designated by simple alphabetical codes to differentiate each other. Because of their unknown power and abilities, I will be only fielding two units, Aurek and Besh, for this operation. Or one to attack each side. More likely than not, I will have to keep conventional forces on standby for this field test if things do not go as planned...

 

[Gir Quee]

Gir finished reading the last log entries and contemplated their words carefully. Even as the logbook's words reverberated through his mind, he wasn't sure if he was ready to actually put Atropos into use just now. Artificial intelligences, even in the droid form, had pushed the limits of many people's comfort zones. He would even had to admit that using Lethe, which really was almost a lobotomized artificial intelligence, had made him feel uncomfortable at first. But Lethe hadn't really had a conscious, nor the ability to speak to him. He wasn't sure if he was quite ready to start having conversations with an artificial entity that had the ability to completely ruin him, especially when Gir lacked the inherent ability to keep such a potentially powerful being at bay if it turned against him. He looked around the room, eying the failsafe lines that could potentially isolate the core of a rogue AI. While he still wasn't entirely sure that such safeguards could fully contain the entity, he placed enough faith in them that they would at least hamper the being's power enough to the point where it could be controlled. He looked up at one of the blue lights on the massive BRT supercomputer.

"Tide, close the project logs and activate Atropos Aurek."

The holographic visage of the logs faded away before a humanoid form began to coalesce in their place. Within the matter of a minute, Gir found himself facing the humanoid form of a young man who could have passed as Gir's brother. He eyed the AI's chosen form up and down. That's uncanny...but then again, this one has known me the longest. Gir shook his head.

"Aurek," said Gir, "one of our agents is about to establish a communication's link with an enemy space station. Are you ready to get to work?"