9The Engineering World View
I want now to move to a more theoretical part of my paper, in which I intend to characterise and define what I call the Engineering world view, and elaborate on where I believe we have cause for concern, in the attempt to amalgamate Engineering and art. In case parts of the ensuing discussion might be found to be affronting to persons trained in engineering professions, I hasten to clarify that my critique is levied not at persons but at the accumulated and often implicit ideology of engineering, an ideology which we are all inoculated with. It would be quite hypocritical of me to criticize engineering per se, since I take part in it every day. What I aim to do is question some of its applications.
Science and Engineering is not an homogeneous entity. Although we might construct an opposition between pure scientific research and the application of such scientific research for the efficient production of goods, it is really a continuum, the boundaries are blurred. Nonetheless, there are core ideas which unite the scientific method, the logic of industrial production and capitalism.
The first of these ideas, reductivism allows that phenomena can be usefully studied in isolation from their contexts. This in turn allows that a holistic system can be rationalized into chosen vectors, vectors which maximise productive output, and hence profit, with respect to input: materials, energy, money and labor. This way of thinking is an 'article of faith' for western culture for very pragmatic reasons: the instrumentalization of this method has led to industrialization, hence to wealth and power in the modern period. I would argue that (contrary to the usual direction of argument) the privileging of scientific discourses in our culture is entirely due to this wealth generating power.
Noah Kennedy has emphasised the structural connection between the computer and the logic of industrial production: "In a sense, the mechanical intelligence provided by computers is the quintessential phenomenon of capitalism. To replace human judgment with mechanical judgment- to record and codify the logic by which the rational, profit maximizing decisions are made- manifests the process that distinguishes capitalism: the rationalization and mechanization of productive processes in the pursuit of profit...The modern world has reached a point where industrialization is being pointed squarely at the human intellect."
That mind is separable from body; that it is possible to observe a system without that observation affecting its outcome; that it is possible to understand a system by reducing it to its components and studying these components (in essence, that the whole is the sum of its parts); that the behavior of complex systems can be predicted: these ideas are hallmarks of a nineteenth and (early) twentieth century scientized approach to the world. When these ideas are instrumentalised, they become the ideology of efficient production, what I call the 'Engineering World View'. Implicit in this discussion is the idea that, futuristic rhetoric notwithstanding, the computer is the pinnacle of achievement of the discipline of engineering and the values that characterise nineteenth century engineering ideology find their purest expression in the digital computer.
Given this, there is a certain irony in the fact that it is the computational capability of modern computers which has, in recent decades, brought key aspects of the engineering world view into question. In the seventies, Benoit Mandelbrot discovered geometrical monsters which he called Fractals by applying the power of the computer to a nineteenth century mathematical oddity. In a similar way, Crutchfield, Farmer, Packard and Shaw shook the scientific establishment with the revelation that simple deterministic systems can give rise to unpredictable and random behavior. They called this phenomenon Chaos and noted that it in principle placed limits on the power of determinism. More recently the reliability of the technique of reductivism has been called into question due to an increasing understanding of Complexity and Emergent Orders.
If the pinnacle of Engineering is the computer, then the pinnacle of the pinnacle is Artificial Intelligence. In the sixties, the perceived failure of the cybernetic approach of modeling organic systems such as reflexes and neural networks had led to the exploration of automated logical systems. The early triumphs of Artificial Intelligence such as Newell and Simon's 'General Problem Solver' found their success in rigorously confined logical domains, but difficulties arose in attempts to generalize these systems to deal with 'real world' problems which have no such bounded domains. Computers were able to excel at logically complex but bounded problems such as playing chess, but were unable to deal with the day to day tasks such as crossing the road. The necessary addition of cascades of contingency conditions led to the phenomenon of 'brittleness'. It became clear that abstract logical reasoning was easy to automate, in comparison to the underlying substrate of learning which we call 'common sense'. I would argue that abstract reasoning is easy to automate because such reasoning is an abstract 'machine'. Like knows like.
Typically, when AI techniques were applied to problems of robot navigation, data was gathered by sensors and a map of the environment of the robot was generated, over which a path was planned. Instructions were then sent to the output devices (usually motors). As the robot proceeded down this path, the environment was re-measured, position plotted on the map, and the map corrected if necessary. This method had come to be known as the Top-Down paradigm. In practice these systems were very slow. It was observed that a cockroach was better at crossing a road than the most powerful computer! This led to the realization that these situations demanded a type of 'intelligence' heretofore unacknowledged by the AI community.
Famously and iconoclastically, Rodney Brooks proposed that AI should stand for Artificial Insects rather than Artificial Intelligence. He argued that a cockroach doesn't 'map', that there was no need for the duplication of the real world in the abstract map, like some kind of floating platonic ideal.This kind of thinking led to a variety of research projects loosely categorised as Bottom-Up robotics. It should be noted that the Top-Down paradigm, in its centralisation of control, inherently perpetuates panoptical models. Furthermore it exactly replicates and reinforces very traditional tropes of master and slave, general and soldiers, boss and workers and more abstractly,nature/culture, body/mind, form/content and hardware/software. Bottom-up theories, on the other hand implicitly oppose authoritarian power structures and endorse horizontal and rhizomatic power structures.