The Charybdis Project

[Sere Reene]

The Charybdis Project
Aboard the OS Stardestroyer Tantalus
Engineering Section

Sere’s foray into AI design had not been an easy one, for one, she didn’t have a strong enough personality to imprint anything into the AI’s, or care about doing so. Most AI’s to her were ship computers you regularly saw aboard starships, functioning tools, so that’s how our AI began its life as a functioning tool!

#Journal Entry 1, Charybdis AI, Eye of the Behemoth.
Day 10

Neurolinks established. Secondary buffers have been rerouted to filter incoming tactical feed from the cerebral cortex functions, usually assigned to process that task. Done to simulate better data handling, and designed for multitasking many smaller systems at once. Lack of spice or stimulants allowed aboard OS ships, makes for much slower progress than usual for our Engineers.

#Journal Entry 2, Beginnings of Functional Design
Day 12

AI Proving Resistant to expected Rogue Squadrons tactical data, seemingly too erratic to accurately model. Have suggested diminishing focus on one specific opponent to allow baseline routines to be established first, and also our technicians be allowed their performance enhancers, to unhinge their creative freedom. OS officers proving reluctant to allow drugs aboard!

 

[Sere Reene]

Material came and went, manpower changed.

#Journal Entry 3, Teething Troubles.
Day 25

AI progress halted on memory recall errors. Storage of multitasked data seen as a problem to accurately arrange which petabytes are critical to its operation, and which can be seen as secondary or tertiary to be processed as time allows. I have assigned an engineering team on filtration of junk data, to ease the use of system resources across such a large body of potential targets, as has been requested in the design brief.

After more negotiation limited spice and stimulant use was granted, technicians working at twice their previous capacity. Officers are pleased with the results, and have agreed to allow twice our usual shipment. I have not told them this won’t get us 400% out of our staff, but will keep us in high spirits, a few trips to sickbay from overused expected.

 

[Sere Reene]

#Journal Entry 4, Checkmates.
Day 32

Filtering out much of the junk visual data, has allowed the AI to function over a large surface area, while still retaining a centralized intelligence. Work has begun to interlace functioning visual sensors, to act as sensory organs where required, acting as secondary processing cores and reducing load on the main cortex.

Only one accident reported so far, a rodian technician tried to initiate a chess routine with the AI, and was shot as a counter move to his checkmate.


#Journal Entry 5, Human Imprints.
Day 36

Trial and error vs virtual or drone based personalities. Decision was made that to predict a pilot, the project required a pilots personality to best execute the task. This was anticipated as a possibility, and our choice of vessel has allowed us abundant choice of material to choose from. Two were chosen, one for her awareness, one for his experience.

The pilots are proving capable but easily bored. Additionally the AI has had its game subroutines put on hold for the foreseeable future.

 

[Ecarht Arak]

#Journal Entry 6, Yeah, I didn't expect to be here either.
Day 41
**akking thing on yet?....

Had us drilling the durahead today. Slow as a pup. Had to get to my bunk. Wouldn’t pass flight screening as it is now. They want me to teach it to fry alliance aces. Uh. Chows up in 5. Told us we’d be eating something other than dry protein packs, better be right. Try again tomorrow. Takki had more luck with his slides, he’s got the patience for this, frak knows I don’t.

#Journal Entry 7, Round and Round we go.

Day 42

Can follow a line, understand what i'm saying to it. What else can it do. Nothing. No sense, no gut, nothing. Frak, what do they want from me.

Nothing else to say.

 

[Sere Reene]

#Journal Entry 8, Non Traditional Teaching Methods.
Day 50

Memory recall, and processing speed have reached an optimum balance over this larger area of sensory feedback. Thinking back to the Rodian incident. We put a gun to the AI’s head and said it could choose life or death. After 2 days of processing it chose life, because death wasn’t a choice it could recalculate, whereas it explained life was. Only able to communicate in binary, the processing speeds and depth of the calculation were higher than expected.

Using this philosophy we installed the AI aboard a test vessel, and explained death was the result of not adapting to the targets preferences, or coordinating the resources at its disposal. While it has no survival mechanism, it does now comprehend that life is the state of actively processing data, and death blocks this. Performance tests were satisfactory as a result. Our Ace Pilots have been tasked with explaining their combat experience and awareness along these new lines of logic and reasoning.

 

[Sere Reene]

#Journal Entry 9, Major Breakthrough!
Day 55

Simulated areas of the frontal lobe have seen activity, showing higher conscious thoughts! Told to put in safeguards for one sith forces, and build in that these individuals are vital to the units function and processing of data, thus its life. It seems to have had the desired effect on its simulated target selection. We have scheduled another live test of its weapon systems with OS personnel in the room for 7 days time. Junior officers that have been selected are apparently of little worth, aging and not due for promotion, we offered them spice for any pain, but they refused.

As a secondary test, we have been asked to bring prisoners from the brig and put them in the room with a gun in their hands, not loaded of course. The results should be interesting to fully test the friend or fire distinctions...

 

[Ecarht Arak]

#Journal Entry 10, Not Funny
Day 61

Are you kidding me??? Durabrain tried to be funny. Locked us in there with it. Sith’sake’s. Told us it had learned situational awareness. It understood death was stopping data. So it stopped me in the room and said I was dead! Lucky I didn’t have my sidearm.

Do what the AI says they say, great idea geniuses. Orders. Its getting smarter, beat Takki at his own sims, never letting him live that one down.

#Journal Entry 11, From the Brig
Day 62

In the brig. Wouldn't stand in the room while its gun was switched back on after yesterday. Would you whoever reads this?
Never really liked the quarters anyway. Durabrain has learned that the crew were there to help it, so not to waste them, but how long is that going to last?

Me and my mouth.

 

[Sere Reene]

#Journal Entry 12, Partial Unity
Day 70

Partial unity was achieved today across the sensor net. Guns are divided up into clusters, like synapse links accessing memory clusters inside the human mind. Each small group of activated turrets form a cluster of memory inside the AI’s brain! Individuals are not calculated, instead group effort of the subsequent parts is done naturally, when new data associated with that memory cluster is accessed.

Much simpler than we thought, making the turrets integral to the AI’s own mind memory recollection! The utility is this could be apply to any sensor clusters which can be recognized as operating together. Better, as the memories of these clusters and their targets evolves, so will the detail naturally be part of the weapons response, taking Charybdis now further than a standard targeting computer AI as was the goal! These turrets are its brain.

 

[Sere Reene]

#Journal Entry 13, Protection of the AI and Operating Efficiency
Day 80

Partial Unity brought the problem of relying on that unity, and when it was no longer present, suffering system breakdown. Higher functioning ceased to operate, simulating a shock to the mind, as if a human mind lost access to several areas of its brain in one traumatic experience. Especially showing vulnerability to ECM or even conventional weapons to break up the brain, by simply taking apart its turrets and memory clusters. The more turrets that were switched off, the less intelligence the unit showed, suggestions were made for a simulated subconscious mind to store vaguer important references in a central databank, which the AI could continually draw from even if operating on a substandard amount of turrets, sensors and memory clusters. This idea is still in its infancy.

As a side note, we've run out of spice, stims and coffee.

 

[Sere Reene]

#Journal Entry 14, Tying the Simulated Subconscious into the Friend or Foe System
Day 95

Subconscious definitions of friend and foe have been rooted in place, almost impossible to override only add to. One Sith forces are the defining factor to protect, and those threatening the One Sith have been deemed the AI’s targets even beyond its direct programming. Any threat to their safety of highest importance.

What this means is, even if all its sensors, their related memories in every turret is gone, it’ll still operate on a basic friend or foe basis, knowing to cooperate with other aspects of itself or the One Sith to box in and neutralise threats in a group fashion.

Not only in the directives and programming on the surface but inside its simulated subconscious it has taken on the best traits of both pilots used, one for his experience and understanding of starfighter combat, and the other for awareness and tactical flexibility in situations themselves.

 

[Sere Reene]

Prescriptive Thesis:
Should an AI be able to control weapon systems over a large area?

Answer:

• Yes if the role is limited and well defined.
• Yes if it has thorough tactical preparation.
• Yes if it is able to operate those weapon systems as effectively as a human gunner.
• Yes if it has safeguards to prevent itself from becoming confused and firing on its own fleet.
• Yes if it is sufficiently motivated but has no illusion of being anything other than what it is.