Kerrus jerked awake, blinking at the officer leaning over his desk. "Yes... Mitchells, isn't it?"
Flight Officer Mitchells nodded. "You look worn out, Professor. How long were you working on it?"
The Professor sighed, and gestured at the pile of datapads and schematics. "Not long enough. I still can't fathom what went wrong with the Sixteenth models. Fifteenth seemed so close, and then the next generation were a complete failure."
Mitchells nodded sympathetically. "Listen, Professor, LSF high command is getting twitchy. Eternal Guardian drains the budget like no tomorrow. They want to know if you think it's viable at all. They want to pull the plug."
"Of course it's viable! I've devoted sixteen years of my life to this!" Kerrus rubbed his eyes, glaring at a circuit diagram as if it had personally insulted him. "Sixteen years," he repeated. "In that time, we've made more advances in computing than had been made in the previous fifty. The First Generation were so primitive as to be laughable, barely capable of doing anything except monitor a power flow."
"It could be argued that you've got no closer to the goal of the project in that time." Captain Namarra leaned into the small office, taking a look around the room. At the dead circuit boards from failed models, the massive posters filled with microscopically complex logic paths, the checklists of goals achieved.
Namarra commanded the base. He wasn't technologically gifted, but he didn't need to be. The team of scientists and technicians, the best in Sirius, even including a few ex-Lane Hackers, was all that was needed. If they couldn't get Eternal Guardian working, no one could.
"I beg to differ!" Kerrus snapped. "They're fully capable of flying anything we have data for. They can shoot with a high level of accuracy. They're tactically brilliant. They're -"
"- not capable of doing all those things in the same model. And they're still failures." Namarra gave Kerrus a cool look. They'd been friends to begin with, but despite the fact that he was only responsible for the base, Namarra took the failures of the team personally. "We're supposed to be creating Nomad killers. The ultimate defenders of Sirius. Uncorruptable, lethally dangerous, unmatchable."
"And the XCAI-17 models will be closer to that. I promise you, Captain."
Namarra snorted, and stalked out without another word. Mitchells nodded, and followed, leaving Kerrus alone.
He groaned, and leaned back in the chair. "They're right. This is getting nowhere."
Code:
I don't know. I've monitored the progress of your work, there is a definite progression.
He glanced at the computer screen, and smiled faintly. "Perhaps. But it's not fast enough. Not for the LSF. They're right, this is an immensely expensive endeavour."
XCAI-14-1 made a buzzing sound that was vaguely like a sigh. The Fourteenth models were military failures, but were the first real successes in terms of personality programming so far. So when the prototypes had been commissioned for spare parts, Kerrus had rescued "Onesy" as an 'intellectual assistant' to bounce ideas off.
Code:
It would be a pity if the project were abandoned at this point. As you said, Professor, sixteen years is a lot of time to devote to something.
Kerrus leaned forward, looking at the reports again.
XCAI-16 Testing Report Wrote:...while the Sixteenth series of Experimental Combat AIs certainly shows an impressive level of tactical acumen, responding adequately to all the simulations, and has shown impressive marksmanship with the laser batteries, they demonstrate great weakness at adaptation.
When run through the "Second Battle of Sigma 13" simulation, the Rheinland forces under their control were initally extremely effective. However, the simulated GMG counterattack on Frankfurt completely went unanswered, resulting in annihilation of both the defending and attacking Rheinland forces in a catastrophic defeat that no human would have managed. In contrast, the Fifteenth series managed similar results to the historical battle, with a stalemate...
...despite the apparent advances made with the 16s, it feels like a step backwards rather than forwards.
Code:
It is very peculiar. Perhaps there is some form of conflict in the code resulting in these failures?
Kerrus nodded slowly. "Maybe." He looked at the logic chain posters thoughtfully. "Maybe less is more..."
Kerrus looked thoughtfully at the three squat boxes that were lined up at the side of the briefing room.
The technology to build the XCAIs had been in place since the fifth year of Project Eternal Guardian. Before that, the units had been too big, too power-hungry, and too prone to overload. Even now, every now and again one of the processor units burned out and had to be replaced at great cost. They recycled as much as possible, and apart from the hardwired algorithms that had been deemed useless, the three units there were more or less the XCAI-16s.
He glanced around. The team was mostly assembled, so he clapped his hands for quiet. The usual buzz of conversation died away, and everyone looked at him expectantly.
"We're going to do something radical," Kerrus said. "We're going to delete most of the logic tree."
Suppressed shock seemed to be the reaction, and there were a few murmurs. "I know what you're thinking," Kerrus continued. "But it's what makes sense. The 15s were close to what we were after, they were just slow as hell, and drained a lot of power, almost as much as the old 6s. The 16s couldn't handle the integration code, and just gave up on most of the more complex features. Now, I've gone through the overall logic tree we've been using to program them, and I've picked out a lot we can delete. It should streamline things to the point that they run at an acceptable speed."
"What are we getting rid of?"
"Sections Gamma 15A to 17B -"
Uproar. Kerrus made a sharp gesture, and silence returned, though it had a mutinous edge. "Sections Gamma 15A to 17B, all of Lambda, and Omicron 3 to 9G."
"What do those consist of?" Namarra usually remained quiet in these briefings, but he always insisted on sitting in on them.
"Gamma is the sequences that prevent code from intermingling," began Kerrus, but he was interrupted by a thick-set ex-Hacker.
"If the code is allowed to run in parallel that freely, it'll begin corrupting. We've gone way beyond the normal limits of programming in producing the self-awareness and learning, if we don't have barriers -"
"It's the only way we can streamline it! Captain Namarra, unless the LSF is happy with units simply too large for fighter-level deployment, we have to take chances."
Everyone started talking at once, and Namarra bellowed for order. When silence again reigned, he looked hard at Kerrus. "Professor, what would happen if the code gets corrupted?"
Kerrus shrugged. "Logic glitches. Loss of combat effectiveness as the ethics programming destroys the tactical subroutines. Worst case scenario is that the hardware burns out. But this is all ifs, not definites. There's no guaruntee that we need the blocker code. Artificial Intelligence programming is a new science."
"Then do it. It's about time we got some real progress here." Namarra turned and stalked out, leaving the muttering scientists.
Sixteen years. Sixteen years since the Nomads had been repelled. Sixteen years since the LSF had called for an uncorruptable defence force to be produced.
Sixteen years of babysitting failure after failure.