Nick Gkikas, BAE Systems
For most of human history, engineering artefacts have acted as a means of direct translation of human input to a determined effect. From the simple hand-axe to the huge rudder of an ocean-faring ship, the effect of each device was directly linked to the amount and direction of mechanical force applied by the human controller. It was only with the introduction of servo-mechanisms at the turn of the 20th century that the direct link between human control input and determined effect broke.
Servos didn’t simply multiply the mechanical force of the control input of the human operator; they quantitatively curved the effect line by introducing an intermediate control module with its own mapping and filtering characteristics. A given increase in human control input no longer led to a proportional increase in effect.
A further step-change came with the introduction of electronic control, in the latter part of the 20th century; the introduction of digital electronic control especially, allowed not only the quantitative differentiation of the effect to a given control input, but also to qualitatively alter the effect.
Although the technological step-changes, the effect for a given human control input has been pre-determined – either quantitatively, or qualitatively, or both – within a finite number values – quantitative margins or qualitative values/options. The dramatic increase in computing power since the turn of this century, allowed for the realisation of learning/adaptive technologies and ultimately artificial intelligence, where a given control input not only leads to quantitatively and qualitatively varied effect, but to non-determined one also.
The present article provides examples of each category of the control artefacts above, explains the paradigm shift in the application of non-deterministic technologies, and discusses the early opportunities and challenges those technologies yield.