Phoenix Sub Zero

The lowest elements of the artificial intelligence were the sensors—sonar hydrophones, electronic countermeasure antennae, as well as valve position indicators, pump running contacts and the nuclear reactor’s neutron flux detectors— functioning as nerve cells. The monitoring and control of the ship’s basic functions—reactor control, atmospheric purification and life-support, depth-control, weapons-control— were all done by the distributed process-control modules in a function much like the human brain stem, controllers of the human heartbeat and respiration. Surrounding the distributed process-control modules were the higher level functions related to sonar and electronic monitoring. Sonar and EM to the ship were its way of sensing the environment outside, much as vision and touch are to the living organism. And just as a human brain has whole lobe regions dedicated to vision and touch, the Second Captain had a layered modular neural network that had separate nodes associated with the reception, recognition and interpretation of sonar and electronic data. The neural nodes themselves made the supercomputers of the decade before seem crude … their logic was not hardwired. The sonar node was capable of simultaneous handling of massive quantities of data in real time, assembling the data into recognizable interpretations of the outside environment by the higher levels of the layered neural network.

The highest functions of the system were the neural artificial intelligence assembler modules, a part of the collected computer totality that were not specifically mentioned in the training or maintenance manuals except in the broadest terms. The assembler modules’ relation to the lower layers was similar to human brain frontal lobes, the functions that took the data of the sonar modular neural network and interpreted them, comparing them to data received in the past, generating internal questions that required investigation and analysis.

The assemblers were part of the system tasked with interfacing with the human crew, the advanced artificial electronic intelligence assigned the heavy responsibility of making the crew understand what the system understood, making the crew aware of its interpretations of the environment.

The task of interfacing with the crew was the most difficult to design, but in this model, Yokogawa had marched into the frontier of artificial intelligence and extended the science a crucial step.

It would be untrue to say that this level of electronic consciousness was capable of thought, but the unit had a random memory regenerator that reached into the plasma-bubble memory cells for things related to current processing, reaching out across the neural network for associations and previous learning. The unit was not capable of reflection but it did reexamine previous experiences for comparison to present processing and extrapolated such experience into a weighted probability prediction of what future events held in store. It was not capable of confusion, but its processors were designed to suppress action recommendations and slow down neural functions in the event that the comparison of past experience and the prediction of future events were in disagreement with the present reality as perceived by the modular neural networks. When the prediction of reality that benefitted the ship’s mission was correct as sensed by the modular networks, the system was programmed to experience a higher neural flux and the release of electrochemicals at certain neural transmitters giving the system the equivalent of self-satisfaction. When the ship’s encoded mission was not achievable the system’s neural flux was suppressed, turning the processors away from continued attempts to direct analysis at the problem. It had no fear but when the predictor of mission success was perceived as low and system survival was seen as improbable the system’s tolerance for risk increased. It also had no way to experience hope but when mission success as seen by the event probability analysis was perceived as high, the assemblers experienced higher levels of neural flux and a release of electrochemicals associated with higher neural connectivity. It had no equivalent to aggression but when the ship’s mission was threatened by rapidly developing threats it was capable of understanding, the risk-gradient position moved into the high positive values and the nodes that considered action options became highly stimulated.