Background
The history of attempts at automating meaning is a long one. In the 1780s, Count Schempler von Schempstead invented a machine that spat out words and phrases by combining typeset by means of separate "grammar shoots." The device was operated with a foot-treadle (Stipe, 1983; Stripe, 1944; & Yan, 1922). Mainly meant as a parlor trick, von Schempstead's "Word Engyne" showed not merely the feasibility of such a device, but the possibility of practical application
Instead of the long and often messy writing process, fraught with revision, broken quills, and spilled ink, von Schempstead showed that simple messages could be left entirely up to treadle and shoot. Problematically, interpretation was still up to the reader. So a statement like "zebra folly louse manure," an easy possibility given the early machine's structure, could only be successfully interpreted through human intervention. Allowing for the inevitable human error, large and unacceptable amounts of inefficiency are introduced.
 FIG. 1: The "Text Elevator" |
This same problem has plagued attempts at automated meaning since then, from Austin Accostin's steam-powered "Text Elevator" (fig. 1) of 1884 (Cork & Short 1997), to various vacuum tube and transistor models from the 1930s to the advent of the microprocessor, most notably those of Jones (1936) and James (1965). With digital computing has come the possibility of so-called "Artificial Intelligence" (AI), a complete accounting of attempts at this being impossible here. But all modern attempts at AI suffer from the same problem: even if we can develop a program that resembles intelligence, that intelligence still requires human interpretation, and statements within the system must be checked against some index to determine their meaning quotient. True automated meaning implies the actual production of meaning itself. Computing has the further problem of being the result of operator input: even if results seem remote, they are still able to be traced back to specific programming. Results, while sometimes unpredictable, are still due of the parameters designed by human operators. Novel situations may be modeled, but no new meaning is actually created.
Bridging this gap is exactly what this project set out to do. Rather than relying on mere computing, the Automated Meaning Machine (AMM) produces meaning all on its own, with little input by the operator.
Design and Construction
It was acknowledged early on that such a device would only be practicable on a wide scale if the AMM were to use off-the-shelf parts. With this in mind, a supply of surplus electronics, stepper-motors and sheet metal were acquired. This also served to help reduce costs. EastWestern University's existing machine shop was also used.
An electromagnetic diffraction grid is critical to the machine's operation. In this is placed the object or abstraction upon which the meaning operation is to be performed. Fields of 0 to 4.5 gauss are then run through the machine, with energy varying with the angle of the field to the object. Angle-swing is achieved by a series of motors attached to cams actuating arms at each of the field-producing frame's four corners. An eccentric field pattern is thereby produced. Cam variance is created through placement of the cams at the ends of rods moving along an X-Y axis (in relationship to the field-frame's Z) at an ever-changing rate of speed and direction. This creates an essentially random movement of the field, assuring the results are not skewed by overly systematized secondary inputs. The motors themselves are shielded to prevent further skewage.
Results are then recorded via several different media, each representing a different sense in order to make the final output intelligible to the (inevitably human) observers. The sense of sight is represented through two cameras positioned on-axis to the field-producing frame. Microphones on either side of the frame record audio outputs. Tactile outputs are simulated with an electromagnet-sensitive membrane consisting of 1.3 mm polyethylene vinyl (much like that of a comb) impregnated with iron filings and run through with a copper wire grid of 1 cm X 1 cm. The tactile membrane is positioned beneath the field-producing frame in order to facilitate placement of objects or abstractions within the AMM. Olfaction and taste are simulated through the use of a modified gas chromatograph, the sensors for which are positioned in a hood above the field-producing frame, slightly off-center of the access hatch. Results from each meaning-producing scan are recorded into a regular computer hard-drive and interpolations into plain language using the Mars/Fears Plain Language Interpolation Program® are simultaneously printed for archiving purposes.
FIG. 2: The "sleek, clean look" of the AMM. |
Sheet metal was used around the AMM to give it a sleek, clean look (fig. 2).
Findings
The success of our process was mixed. When the AMM was used on a chicken, for instance, the interpolated plain language output read "Insouciance is not a condition, perforce criminologists am." The problem of interpretation being what it is (see above), success is ambiguous in this case. Results from a Hershey's™ brand chocolate bar were more stable: "Form-fitted manhole cover." Responses for the rest of our preliminary runs are outlined on the accompanying table (fig. 3). Of particular interest, however, are such anomalous outputs as "cringing metaphor for dyspeptic seizures" when the AMM was confronted with an alarm clock (Wesclox® model A1M1), and "persimmon mal cakes" when the AMM processed a 4kg solid aluminum bar.
FIG. 3: Results of preliminary runs |
Discussion
It is possible that some fine-tuning of the AMM will be required to
fit specialized circumstances. However, this team stands by the validity
of the process and design. Responses, while enigmatic in certain cases,
are no further from the mark in terms of identifiable truth value than,
for instance, current White House communiques. Outputs from the AMM
actually compare well on the accepted Fog Index of abstruse writing
with the average essay appearing in Lacanian Ink. Furthermore,
the greater efficiency the AMM allows could be of inestimable value
in areas where complex decisions must be made on a routine basis, such
as governance, space-exploration, and corporate management with little
if any increase in randomness or decrease in accuracy. The speed with
which an automated process can create mistakes tends to even out the
severity of any given error, and the separation from system with which
the AMM is able to operate may in fact balance out the relative dearth
of consistency in the decision-making processes current in the above
fields.
References
Cork, M., and Short, P.V. (1997). The problem of meaning in a systematized framework: A fieldbook with illustrations. New Haven: Pistule Press.
James, J.J. (1965). Transistorized resistance to realistic interpretation: A case study. Journal of Implied Heuristics, 33, 234-66.
Jones, B. (1936). Vacuum tube resistance to realistic interpretation: A case study. Journal of Implied Heuristics, 4, 92-122.
Stipe, M. (1983). Giggles the clown: a tale of Victorian intrigue among the Algonquins. New Brunswick: U. Conn. UP.
Stripe, P. (1944). Miraculous inventions of a bygone age. New Canaan: Conn. U. UP.
Yan, Y.V. (1922). Meaning mastered: Another triumph of modern science. Research In Techno-Triumphalism, 10, 34-47.