Silicon Domain

1 Logic Circuit

A logic circuit with discrete binary inputs and one binary output presides over the input, the transformation, and the output of data. A closed switch in the circuit will conduct an electric current and an open switch will conduct no electric current. The behaviour of switches (ON or OFF) depends on the voltage level of their gates which is, in turn, controlled by the variables associated with a computer function. If a transmission path exists for a given signal then the output result has the value of 1; else the output is logical 0. Combinational circuits perform a specific function: the OR-gate will have an output value of 1 if either of its inputs are equal to 1; the AND-gate will have an output value of 1 if both of its inputs are equal to 1; the Inverter produces a logical output 1 if its input is 0. In a pull up circuit, the source voltage at the gate is matched by the output voltage; if the circuit is inverted a logical NOT condition is implemented with NOR- and NAND-gates. The speed of a circuit depends, amongst other factors, on the delay of state changes between gate inputs and gate outputs.

Boolean truth values of 1(true) and 0(false) are used in logical expressions to describe the structure of a circuit and the conditions for an electric path. Higher-level primitives and logical macros such as decoders, multiplexers (MUX) and programmable logic arrays (PLA) involve the exploitation of repetitive patterns. Decoding circuitry acquires the output terminal position of a corresponding 1 signal input belonging to an n-bit word that has been 'trimmed' of costly 0s.
A PLA is used to implement sum-of-products expressions for a set of true and complement input variables. It contains programmable switching elements and has two levels of logic (AND/OR) enabling the computation of conjunctions and disjunctions in order to realize a number of output functions.

In a sequential circuit, where values increment at each clock pulse, the output may not be the result of current inputs only.* The flow of data in a dynamic digital system requires memory elements which are controlled by a clocked device such as a microprocessor. ROM and RAM cells provide this storage, utilized by loading the address into a register (MAR) to enable a read/write on the next cycle. The data paths of a processor comprise memories, registers, shared point-to-point connections, etc. to permit the synchronized flow of control signals and information.

* Feedback indicates that input and output are both influencing the current state of the system

2 Semiconductors

a) To achieve clear logic values in switched circuits, a continuously varying quantity is discretized in binary form. In semiconductors, the range of 0-1 volts corresponds to the value of logical 0 and the range of 4-5 volts corresponds to the value of logical 1. Conduction is the result of electron energy gain caused by thermal and electric excitation.
The deposition of impurities during fabrication modifies intrinsic semiconductor properties: in a two-carrier transistor, for example, donor impurities produce an N-type semiconductor where negatively charged electrons carry the current; acceptor impurities produce a P-type semiconductor where flow occurs due to positively charged particles [holes]. Transistor action (charge transfer) takes place at intersections between n-type and p-type layers: the electrons will move towards the p-region and the holes towards the n-region. The n-p-n junction transistor comprised a single crystal of silicon with two n-type layers, the emitter and the collector, separated by a p-type layer called the base.

b) The concept of encoding the structure of arbitrary machines (including the universal Turing machine) inspired the functional elements [input-output device, central processing unit, memory] of the general-purpose computer. A hierarchical IC design decomposed the chip recursively into component instances to the depth of a logic gate, while the integration of reusable modules (or the black box) primarily relied on documented input-output (IO) behaviour. VLSI designs progress from algorithmic descriptions, and technology-independent representation of functionality, to a technology-dependent [n-p-n, FET, CMOS] representation of logic and circuits which can be simulated before their physical placement.

3 Cybernetics

Cybernetics is the science which studies the control systems and information processing found in biology and machines. Phenomena such as entropy-increasing inanimate matter, negative phototropism, and position-guiding by motor neurons,* all belong to a versatile analytic space applicable to chemistry, physics, or information theory.

In Cybernetics, the interaction between an object and its environment is driven by stimuli and responses. A living body [bearing the exchange of atoms and molecules] will differentiate itself by reducing entropy. Cybernetic systems are either equilibrial, oscillatory-equilibrial, or non-equilibrial. Cybernetic modeling enables the scientific discovery of properties in one [unknown] system by means of another [known] system - subject to parallel objects and mapping criteria, e.g. speed-current, mass-inductance, and force-voltage.

* The Nervous System is divided into: i. Receptor Nerves; ii. Effector Nerves; iii. the Autonomic Nervous System and the Central Nervous System. Nerve cells, or neurons, are bipolar and exist in either an excited state or an inhibited state. The Central Nervous System is divided into seven discrete parts listed in order of phylogenetic evolution: 1. Spinal Cord; 2. Medulla Oblongata; 3. Cerebellum; 4. Midbrain (mesencephalon); 5. Diencephalon; 6. Corpus Striatum; 7. Cerebral Cortex. In this hierarchical structure, afferent and efferent links between superior and subordinate parts are vital to coordinated action, but many processes suggest a massively parallel rather than serial or iterative approach.

DCL (conveyor,pallet,down) FIXED POINT; DCL (oneposition,tonextrow) FIXED POINT; DCL (ready,moving,i,j); INTEGER; DCL (palletflag,conveyorflag) INTEGER; DCL (done) BOOLEAN; conveyor=PT(0,0,0) IN LABXYZ; pallet=PT() IN LABXYZ; down=PT() IN LABXYZ; oneposition=PT() IN LABXYZ; tonextrow=PT() IN LABXYZ; ready=1; palletflag=2; moving=1; conveyorflag=3; REFERENCE=HANDXYZ; done:=FALSE; WHILE NOT done DO; BEGIN; WHILE STATUS (palletflag)=ready DO; END; i:=1; WHILE i<=3 DO; BEGIN; WHILE STATUS (conveyorflag)=moving DO; END; j:=1;
WHILE j<=3 DO; BEGIN; MOVE TO conveyor; MOVE BY down; UNLESS HIT DO; OPEN BY -.01; END; MOVE TO conveyor; OUTPUT TO TERMINAL 'restart conveyor';   // robot programming with MAPLE [1985]

4 Autonomous Agents and Robotics

a) As part of a digital environment, software agents can react to changes, process sensed data, and generate actions by means of effectors. Luck and d'Inverno's four-tiered hierarchy carefully differentiates between virtual objects, or entities with capabilities, and more advanced independent agents motivated by goals.

b) A servo-controlled mechanical arm moving along six axes [XYZ, yaw, pitch, roll] can pick up an object and put it down in the right place. Point-to-point Control divides this kind of manipulation task into a series of clearly defined steps, albeit with arbitrary connecting paths. {A,h,r} working volume coordinates [Angular_position, height, radius] are based on Cylindrical-, Spherical-, and Articulated Geometries. In Robotics, each degree of freedom is associated with one actuator, or drive unit, which may be electric, hydraulic, or pneumatic. Continuous Path Control requires that robot motions be a function of time, such that a set of Cartesian coordinates for each anthropomorphic constituent supports an optimum trajectory between the effector origin and its destination (see also, data path bit-slices). A representation of this potential machine haze - given the velocity, positions, heavy-weight accelerations, and a dynamic assembly line - includes algorithmic investigation of safe channels for multiple moving objects.
Robot control combines implicit machine instructions, namely actuated signals that extend the angular position of each drive unit, with a feedback mechanism. Periodic deviations from programmed motion, for instance due to loading, are then incorporated as proportional errors (PE) to regulate the command signal. Position-sensing hardware like Incremental Optical Shaft Encoders comprise a light source, an opaque glass disk with one circular row of slits, and three sensors. To obtain directional detector signals each rotation must activate one of two sensors first, while a reference sensor measures elapsed time via pulse counts.

High-end multitask robots are equipped with machine vision and tactile sensor systems so they can interact with their surroundings cybernetically. An early object recognition method used a 3D-laser-scan to produce 0-level representations of the object in space. This was followed by a triangulation to identify geometric primitives in the object's surface which could be mapped onto the primitives of a possible match. Robot processing of dynamic environment variables has been approached with an organizational hierarchy of computing modules, extending downwards through functional decomposition from goal articulation and strategy to the level of actuated movement. Every module is capable of evaluating feedback received from lower level modules, as well as the environment, with subtasks assigned to the nearest subordinate.

5 Computer Graphics

In Computer Graphics, a mathematical representation of the object is governed by XYZ world coordinates in order to specify a viewbox. Simple two-dimensional transformations that shift, scale, and rotate a geometric outline can be accomplished by adding or subtracting translation amounts to the XY positions of their vertices (with a subsequent redraw of connecting lines).
The [planar geometric] projection of a conceptual object floating in logical 3D space - the volume where points are represented by XYZ triplets - is sliced at the view plane and mapped into the viewport for 2D display. The view plane of a one-point perspective projection bisecting a three-dimensional scene at a distance d from the origin converges parallel lines in the z-direction and also determines the size and distances of its objects with a scale factor derived from the division of XY coordinates by Z.

Computer-generated graphics for 2D display are the result, amongst other processes, of cascading coordinate systems, for example: i. the object coordinate system, whereby objects are conceptually defined; ii. the world coordinate system, where objects are transformed to permit their representation in the computer; iii. the view-reference coordinate system, defining the XYZ orientation of the view-volume; iv. the normalized-projection system, or 3D screen coordinates, which describe the entire view-volume before clipping against the window; v. the 2D-device coordinate system to represent the image via pixels indexed by an ordered pair.

6 Virtual Reality

a) In Quadratura, or Ceiling Painting, the artist creates perspectival architecture to achieve a logical and visually seamless connection between his fictive structure and the physical room. Optimum trompe l'oeil of a flat-surface composition (with one vanishing point at zenith for instance) is determined by the rules of linear perspective and the station-point of the observer. In a volumetric engagement with a stone wall, a 'subtractive' building block [eight vertices and twelve edges drawn at perspectival right angles] can thus be gained for the purpose of an illusory extra room.

b) A virtual reality (VR) system intrinsically combined three types of hardware:  1. the head-mounted display (HMD) to convey stereoscopic vision in a 3-dimensional electronic space; 2. the tracking device to determine the XYZR position of physical objects; 3. the pointing and manipulation device.

An HMD typically contains two SVGA-resolution LCD screens. It can be used in full VR mode to interact exclusively with a computer-generated world or in see-through mode for Augmented Reality applications. A compact version of half-silvered HMD's are head-up displays by Microvision. With a laser, context-sensitive diagrammatic and textual information is superimposed on top of the real world. Virtual Retinal Displays (VRD) project stereoscopic imagery straight onto the retina. At the same time, it records head movements and point-of-regard data so the superimposed electronic information is accurately aligned with physical instrumentation.

In a virtual world, orientation data needs to be exchanged between the user, wearing a head-mounted display with a sensor, and the Simulation Engine, so the graphics can be refreshed on the fly. There are several kinds of position trackers available, such as mechanical armatures, ultrasonic sensors, and magnetic trackers. An inertial system from Intersense offers three degrees of freedom [up, down, rotation]. This tracker is also compatible with design applications and games. The Liberty tracker is a high-end unit. A component called the source emits electromagnetic fields, the intensity and angles of which are gathered via motion-sensors to compute all positional data of the moving subject. Liberty offers six degrees of tracking - XYZ and yaw, pitch, roll. It has an effective range of 180cm. With a latency of less than 4ms the graphic response is almost instant.

Pointing, input, and manipulation often begin with conventional 2D devices, for example the keyboard, mouse, and menu-API. To make volumetric interactions, instrumented gloves are fitted with optical tracking and bend-sensors. For this reason they can be seen and operated in the virtual space. There are many 3D tools which can be taken into a VR environment; with digitizers like the Liberty Stylus, for example, real objects, artefacts, and statuettes can be archived dimensionally and recalled using a desktop computer or PDA.

7.1 Artificial Intelligence

a) Artificial Intelligence (AI) is the study of intelligent behaviour and, like bionics and cybernetic modeling, builds upon isomorphic relations (mental functions, hierarchically wired structures, computer technology) to explain information processes - apart from invoking more highly evolved versions of itself.
AI implied a theory of the domain, that is, a set of primitive elements, connections, and laws to support formal representation based on symbol manipulation.

Tables of behaviour have inspired not only conventional symbol systems (requiring CPU interpretation of machine language, execution of operations [run_ops_standard], and data structures derived from a script) but also the digital representation of biological life-forms. The mind and manifestations of intelligence (recognition, understanding, reasoning) refers to cerebral processes actuated by neural impulses, while the brain is a physiological medium driven by electrochemical processes. Any structural similarities between this organism and a computer system may be expressed on the device level, the circuit level, and the configuration level [IO, memory, brain].
b) Knowledge representations such as labelled data items, property lists, and databases the AI program will access to perform its tasks can be declarative, using statements of facts, or procedural, whereby inference making methods drawn from production rules are selectively activated. In Propositional Logic, the truth or falsehood of a statement is determined by means of argument forms. In Predicate Logic, the additional use of variables and functions allows for data dependencies such as linked facts.
Semantic Networks represent objects, events, and actions with nodes connected by links which are subject to inheritance hierarchies and support domain-specific inference-making. Extensible Markup Languages (XML), used on the World Wide Web, is a meta-language for author-defined tags contained in structured electronic documents that provide automated machine-understandable processing, especially in conjunction with a Resource Description Framework (RDF).

Problem-solving in AI is guided by control strategies: a search space of possible solutions is negotiated by proceeding from the initial state (root node) to the goal state. A 'connecting' search path may traverse successive levels of the data structure [breadth-first], or lead from one individual node to the next [depth-first] until it matches (terminal nodes) or decides to backtrack. Brute force methods developed in adversary game theory attempt to calculate all possible future moves, while selective modes focus on closer sets and goal-directed reasoning. Machine learning involves adding data items to the list of a library, changing the configurations of a simple task, or modifying a plan to address conflicting goals.

"[The box] was sombre, gentle, and somehow childlike. It contained seven objects. The slender fluted bone, surely formed for flight, surely from the wing of some large bird. Three archaic circuitboards, faced with mazes of gold. A smooth white sphere of baked clay. An age-blackened fragment of lace." (Gibson, 1986)

7.2 Expert Systems

Robotic comparison of internal expectations with sensor data to guide the selection of algorithms relates to the way distributed mental models and sensory information are assembled in computer-human interaction (HCI). An Expert System contains and manipulates domain-specific knowledge elicited from interviews with human experts, or activity-based modeling in a sensated environment [ubicom]. Expert systems are structurally divided into: 1. the working memory, to record input data; 2. the rule memory, which is the knowledge base comprising conditional statements; 3. the rule interpreter, also called the inference engine, to direct a search of the knowledge base.
AI inferencing techniques were explored in Caduceus, a medical diagnosis tool designed for navigating a semantic network in order to match manifestations (symptoms, test results) with a known illness. Computational modeling and automation of the microchip design process [high-level partitioning {IO, functions, constraints}, allocation of architectural modules, local placement of registers and data paths] make use of design abstraction, parametric CAD, and prototype editing.

8.1 Artificial Life

a) Artificial Life (AL) investigates dynamic self-organization, behaviour generators, and the emergence of patterns and complexity in the synthesis of life which is itself, according to Christopher Langton, "a set of vital functions implementable in various material bases." Traditional AI involves the human operator, top-down hierarchical principles, and symbol manipulation, while AL begins with a set of simple rules and generates complexity.

b) A biological form (phenotype) resulting from local structural effects due to the instructions (genes) received from a DNA sequence [itself a subset of the genotype space containing all genetic materials] is a biochemical 'machine' instantiated through morphogenesis. The autonomy of living systems, biological and artificial, is influenced by: i. the external environment; ii. internal plan-based mechanisms; iii. thought and selective modification. In genetic programming and evolutionary design the object is represented as genetic code. Operators for reproduction, cross-over, and mutation are applied to a set of character strings, or digital organisms, so as to achieve some genetic variation, e.g. nucleic acid- and 0,1/1,0 replacements, followed by fitness evaluation and natural selection.

8.2 Evolutionary Architecture

Computer Graphics have been used in Strange Attractor visualizations of mathematically amorphous systems and for the fractal geometry of the Mandelbrot set. Parametric software, programming, and growth algorithms enabled pioneering digital artists 'to draw virtual [techno-genetic] circuits through various electronic infrastructures' in search of new cyber-territories. This liquid architecture and rendition of strange asymmetrical structures is characterized by self-similarity, recursive subdivision, and randomness.
The generation of digital architecture involving computational models based on nature emulates evolutionary processes to shape artefacts and design. In John Frazer's evolutionary architecture, the composition of the object [as phenotype] is converted into genetic algorithms, subjected to design criteria, and interactively modelled in a virtual environment. This facilitates the development and iterative adaptation of prototypes for evaluation, as well as crossovers and mutation. The design practice again becomes a problem-solving assignment with objectives, mapping procedures, and constraints. An optimum solution requires the calculation of chromosome/population fitness and aesthetically pleasing forms. The generative script, containing sets of instructions to steer a system of genotypes, can be used in a computer simulation in order to investigate emerging forms. -(see also, Manuel DeLanda, 'Deleuze and the Use of the Genetic Algorithm in Architecture')

Concept, Text, Coding (c) Marcel Ritschel, Sydney 12.06.2005 (edited: 2009)