The next screen was Prioritize. He checked the box labeled Very High, this being the highest level of priority available.
The man reviewed all his settings in a hurry, looking over his shoulder every minute or so. It was getting late, but he was almost done. He clicked on Save and Exit. The system prompted him for the access username and password required to save any changes in the self-guiding software. He took a piece of paper out of his pocket, and started inputting the data in the required fields.
Username: DSheppard
Clearance: Top Secret / L1
Password: EJgF236jhg@k34g5
When he finished entering the sixteen-character password, he clicked Confirm. The screen returned an acceptance message and an option button labeled Distribute. He clicked it, watched a status bar for a few seconds, and then saw the message he was waiting for: Field Distribution Complete. Satisfied, he removed his flash drive, turned off the PC, and left hurriedly.
…60
With a sigh of relief, Alex walked through the conference room door. She had made it on time. The usual players were already gathered around the table. Janet was there, but Alex decided not to acknowledge her specifically, or divulge the fact that they'd met in private. She greeted the entire group with a brisk "good morning," then took her seat.
John Dunwood, VP of manufacturing, looked even paler, older, and more wrinkled than he did at the last meeting. He was biting his fingernails. Alex wondered what supreme effort of will, or what desperate situation could make a man come through that door again, after enduring what he had endured. What could be the reason he was still here?
Robin Maxwell, from HR, was fresh and relaxed, as only an uncaring professional can be in a dysfunctional corporation. Alex could barely refrain from grunting at her. She still remembered Robin enabling Walker in his continued abuse. Ugh…
Miles Putnam, director of R&D, the guy who told Alex that pushing people the way Walker did generates innovation and drives performance, looked a bit worried, sifting through his pile of paperwork and reports. From the peek into the meeting's agenda, R&D results were on the table for discussion, so he was going to be in the hot seat.
Finally, Peter Wilson from Six Sigma was looking down, avoiding eye contact with everyone. At the last operations meeting, he had been tasked to look into the reasons behind the failure rates of the new designs. He could also end up in the hot seat, despite all his previous enabling. Alex almost smiled.
Walker slammed the door behind him, startling everyone.
"Good morning, I guess," he said. "What do you have for me today? Six Sigma, Mr. Wilson, have you finished the root cause analysis on R&D's new model failures?"
"We have some preliminary findings," Peter said, shuffling his papers. "We've looked at failure rates for all models currently in a design phase and compared these failure rates with the models already in production. A few of the in-design models have higher-than-average defect rates. The RX5, the reconnaissance drone, has a defect rate of 5.65 percent. This is the worst. Then the CX12, the UCAV, failed at a rate of 5.12 percent, also extremely high."
"Remind me, please, what is the target rate for the quality of in-design drones?" Walker asked.
"They shouldn't be failing in more than 1 percent of cases, cumulative, regardless of point of testing. Lab testing and field-testing results are compounded to create this failure rate. Combined, it should not exceed 1 percent," Peter clarified.
"That is fantastic," Walker responded, in his sarcasm-filled voice. "We're only missing the target by some 500–600 percent! Any insights about what's causing these high failure rates?"
"From what my team could assess, electronic components fail at an accelerated rate. Microchips failure rates are quite high, even after installation. However, we have also encountered software failures. The CX12 and CX15 combat drones and the RX series are all self-guided prototypes."
"What's failing on the software side?" Walker asked. "I thought we had the software piece figured out. If I recall correctly, we've already been shipping self-guided drones to the clients, right?"
"That is correct, sir," Peter resumed his explanation. "We've been shipping self-guiding enabled drones to the clients for almost eight months now, for their testing phase. The drones are still to be flown with remote guidance, rather than self-guided. But the Army, Navy, Coast Guard, and Marines, all wanted the opportunity to test the self-guidance module and prepare procedures for the next phase of deployment. They're not supposed to deploy them into active operations as self-guided — not yet — at least not until we've cleared them, and they have finalized all testing on their end."
"When's that supposed to happen?" Walker asked. "When can we clear them for self-guided operations?"
"Not for months, sir, not at these failure rates."
"We're shipping drones with defective software to our clients for testing. These high failure rates must be costing me a fortune. Hmmm…" Walker rubbed his chin angrily. "Someone will pay for all this waste. Let me reassure you, someone will pay." He wasn't yelling. His voice had that bone-chilling calm that preceded his anger-filled incursions into the land of psychological abuse. "What's failing on the self-guiding software?"
"The drones we tested have failed in orientation, navigation, and target acquisition," Peter answered. "They get lost somehow. They're supposed to navigate themselves via GPS and establish their patrol perimeter or attack, coordinated by either the landmark or target imaging that is preloaded in their systems, or by pre-entered geo-coordinates. In the case of coordinates, the drone's performance is satisfactory; defect rate is well under the 1 percent limit. In the case of landmark and target-image recognition, they fail quite significantly. They exceed their patrol area and have to be manually overridden and brought home. They fail to recognize a target and they just miss it. In a couple of cases, there were issues with correctly recognizing ground captured imagery; the drone in question opened fire on a civilian landmark, one that should have been ignored."
"Can someone please refresh my memory on how this works?" Robin asked.
"Yes," Peter answered, "quite simple. We load the drone's memory with two sets of images. One set is landmarks, items that are easy to recognize and should be used for guidance only. The drone should patrol, for example, from the Statue of Liberty, all the way to the Empire State Building and back. While flying, the drone captures ground imagery with its cameras and runs image recognition software, enabling it to recognize the landmark for what it is. The moment it recognizes Lady Liberty, it's supposed to execute a turn and head the other way, scanning the ground for the next landmark in its programming. So, quite logically, if a drone fails to recognize a landmark, it will not correctly execute the patrol or the attack plan. On combat drones, the situation can be even trickier if the drones fail to recognize correctly the imagery that was picked while traveling. It can generate a drone run into enemy territory without any targets being taken out. It can trigger attacks on different targets than intended."