The Golden Age of DARPA

Kevin Knight (with apologies to Aleksandr Solzhenitsyn)

[Translator's note: This article covers a three-day period in 2003, describing everyday events within the vast national security apparatus. The main actors are the scientists and engineers ("zeks" in their slang) who were set to work on government projects. The zeks worked in special institutes which dotted the nation. These scientific institutes, outside the formal system of forced-labor work camps, were known as "sharashkas."]

Greetings from the zeks at the M_______ sharashka!

This week, all zeks were transported to a work camp in rural New York province. Zeks from the B-collective of the national sharashka-system were transported there first. As a whole, we were tasked with building voice decoders for the leadership. As you may know, a voice decoder taps foreign-language telephone speech, converts these taps into our own national language, and stores the results in vast databases for retrieval and processing.

Other zeks arrived from the S- and I-collectives, each collective being charged with the exact same task. Zeks within each collective believed their voice decoder to be superior to that of the other collectives. In part, this was a psychological necessity, as the leadership made plain that the lowest-performing collective would be liquidated after Phase 2 of the current five-year plan. Each zek was issued a name tag with one of three designators (B, S, or I), indicating his or her collective.

The zeks were relieved to hear that all of the voice decoder program goals for Phase 1 were met. However, as this was totally unexpected by the leadership, goals for Phase 2 were made significantly more difficult, compared to the original five-year plan, effectively raising the accuracy requirements from 80% to 90%. The zeks enthusiastically signed up for the new Phase 2 goals, considering the alternative to be worse.

The leadership demanded a solution to the voice decoding problem within five years. This would have the necessary side-effect of eliminating any further need for the zeks' highly specialized technical expertise, which would free up the zeks for unskilled labor in work camps. Failure to meet or exceed the program goals would of course lead to the same result.

Instead of facing these serious consequences, zeks instead argued nonsensical points with each other. Did the title of a book called "Simplified Chinese Grammar" mean that the grammar was simplified, or that the Chinese was simplified? These were typical zek arguments.

That year, the leadership suddenly changed the way it would evaluate the voice decoders. But the zeks stubbornly clung to their old evaluation method ("Azeul" in their jargon) for their technical work. Regardless of this, in open joint meetings, the zeks joined the officials in denouncing the old ways, including Azeul.

Zeks transported from the all-male acoustics sharashkas were unaccustomed to the novel presence of female zeks. According to the national research structure, a typical male zek worked for years within the same group of five to six other male zeks. His ability to interact with anyone outside his group was limited.

A few zeks came from work units in the national universities. The leadership had recently reduced university participation in the voice decoder project. The reasoning was clear: zeks already serving life terms in university work units ("tenured zeks") did not respond appropriately to the Go/No-Go program goals, having nothing to gain or lose.

A zek's work within a collective consisted largely of participating in sharashka-to-sharashka telephonic conferences. According to a rotation schedule, a given zek would chair the telephonic conference while the other zeks would sleep soundly at the other end of the telephone line. In this way, telephone records would show the zeks to be working in a highly consistent and coordinated manner.

Despite forming an integral part of the national security apparatus, the zeks considered themselves opposed to militarism and foreign intervention. They constantly talked of how quickly their scientific field could advance but for the cost of one aircraft carrier. Of course, the zeks did not appreciate how funding a handful of obscure fields thus would render the national navy devoid of military capability. Nor did they realize that naval zeks were simultaneously calculating how to expand the fleet through the wholesale elimination of linguistics, acoustics, foreign languages, and other fields not directly involved in personnel liquidation.

At noon, the zeks were issued ceramic plates and organized into two lines. Food rations were arranged on a long table, and zeks filed past on either side. If two zeks on opposite sides of the table paused to talk about "multi-pass decoding" or "posteditor variability," other zeks would angrily denounce them for holding up the line.

In order to prevent individual zeks from monopolizing any particular piece of knowledge (thereby becoming indispensable), all were required to submit their research for "peer review." In the distorted zek worldview, such submissions raised a zek's status with the others. The most prolific zeks were given life terms (in ironic zek slang, "tenure"), simultaneously removing their incentives and preventing physical escape. Prolific zeks in no-tenure "special institutes" were treated differently; they were assigned the task of organizing zek conferences and workshops, analyzing technical solutions rather than creating them.

The zeks communicated with each other through transparent slides projected onto a wall. Originally, room lighting was dimmed. Once officials discovered that zeks were falling asleep, they mandated that the lighting be increased, rather than decreased, which rendered the slides unreadable.

Some prisoners were charged with providing linguistic data to others. They used phrases like "exceed our Phase 1 targets for dialectical tree banks by 30%" and "scurry to implement our data plan and respond to ongoing requests." The technical zeks were divided as to the desirability of new data, some insisting that true voice decoding would be impossible without more data, others despairing of having enough calculating equipment to process more data. It was another pointless zek argument. The same data would be sent out, regardless.

The zeks had little contact with the outside world. They attempted to encode secret messages concerning their incarceration by inventing ironic technical terms such as "forced alignment," "bound variables," and "tied parameters." However, these terms made no impact on the free citizens outside, and so the zeks ultimately found themselves saddled with hundreds of ironic technical phrases.

At the end of the meeting, all zeks were put onto aircraft and sent back to their respective sharashkas. The pressure continued to mount on them all, fear of success competing equally with fear of failure.

c 2006