4.03.2005In 2007 Google Inc. launches GLD, the Google Language Database. It is a highly interconnected web of meanings and references skimmed from the surface of the internet. It contains over 400,000 entries of different uses and senses of words, idioms, and other meanings, and can be scoured with amazing speed using Google's search technology. It is, however, simply a semantic network, though it is indexed to real usage of language instead of some artificial network.
Simultaneously, Google Inc releases GAP, the GLD Access Protocol, which is a program that has millions of stored grammatical and syntactic structures, drawn from the best linguistics and cognitive science available at the time, and weighted to reflect the average grammatical structures found on the internet. GAP is the only way to access GLD, although there are many proprietary variations on the protocol (see below). GAP basically provides a bank of sentence 'shells' that can be filled in in various ways (including, of course, recursively) to construct meaningful sentences from the existing GLD network. GAP is a multi-function protocol with two major features:
- GAP can produce a sentence, paragraph, or essay of arbitrary length on any subject in the GLD database. Google Inc provides this service for a substantial fee, though previously produced GAP essays become publicly available after a certain 'hold' period has expired. GAP can also build the essay to be of arbitrary depth of information, as well as adjust its vocabulary and sentence structure to match a variety of target audiences of arbitrary reading level. GAP is similarly structured to allow essays written in any mood desired- it can construct argumentative or persuasive essays, but excels specifically at historical scholarship-type essays (Google Inc publishes its first internet-only encyclopedia series in late 2012). The idea behind GAP's ability to this effect was the breakthrough of specifying one node or group of nodes in the GLD network as the 'subject', and then processing surrounding nodes with respect to their relevance to the subject. Early beta tests of GAP were surprisingly good at constructing compare/contrast type essays on two different topics by analyzing the relations between the two 'subjects'; by the launch date, however, GAP had already made media headlines, not to mention some rather large academic waves, by having several substantial papers published in major academic journals on the subjects of history ("Trade routes and the caste system in pre-industrialized India"), sociology ("The threat of rational organization on contemporary corporate identity"), philosophy ("Aristotle's infinity and its influence on Euclid"), and many other areas. None of it was groundbreaking work by anyone's standards, but it was good, solid scholarship that easily passed blind review peer scrutiny. Initial complains about Google’s inability to properly ‘interpret’ scholarly texts quickly broke down as a rash of papers were published in popular poststructuralist and postcolonial journals; “Feminism and ______” became a popular template in requesting Google’s analysis. Because of this, GLD is widely recognized as the first artificial system to legitimately pass the Turing Test. However, it was the second ability of GAP that ensured its place as the first 'thinking' artificial system.
- Where the GLD network breaks down into ambiguities, GAP produces questions that must be answered to fill in the network. In the process of constructing essays, GAP will often come to highly disputed areas where no clear understanding exists in the network, or where there are specific gaps of knowledge. GAP avoids addressing most of these issues in its essay creation process, but will often produce simple 1 or 2 sentence questions as it processes the network that it finds lacking. In other words, GAP produces as a byproduct a series of important research lines that need addressing in order to complete its understanding. Academics, who initially were quite skeptical of GLD in general, now saw it as a near infinite source of research goals and areas of interest, and became quick to praise it. Scholarship proper fell by the wayside as ‘redundant’, and academia thrived on answering the requests GAP provided for it.
Initially, the questions produced by GAP were in the format that GAP requests took, and answering GAP was accomplished by giving it both the first-hand resources and pieces of analysis (often several pieces or entire anthologies of literature) from which it might construct answers. The first question thus produced was as follows: “What was the status of children and children’s education in Manifest Destiny west?” This delighted researchers, as it was a specific subject area that was already known to be lacking in information, and more importantly many of the early newspapers from that time had not yet been digitally archived, so it was clear and known lack in GLD’s knowledge bank. Several analysis pieces were quickly published by scholars on this area and fed into the GLD network. More questions were spit back out, and academics salivated and tripped over themselves to answer GAP’s request. Sometimes the questions produced were more or less nonsensical (in a more or less uninteresting way), and such errors were traced back to inconsistencies in the organization of knowledge; often with the help of more academic pieces attempting to straighten out these ‘inconsistencies’. Typically these questions had to do with common idioms or turns of phrase that had never been explicitly dealt with in the database; for instance questions regarding food were quite normal- “How hot is room temperature chicken soup on Mars?” was a paradigmatic sentence, and journal articles soon appeared regarding “The maintenance and storage of warm water-based foods on extraterrestrial surfaces”, “The hermeneutics of rooms”, and so on, in a somewhat desperate attempt to address what problems GAP was having.
However, on March 3rd, 2011, GLD produced the question “How old am I?” Researchers were somewhat startled, since all information regarding the construction and organization of Google systems since their inception were part of GLD’s data structure. The answer, it seemed, was clear, and reflected no gaps in GLD- it was a few months shy of turning 5, counting back to its launch date, and a few months older than 7 counting back to the initial conference meetings that began the project. GLD was perfectly aware of this. One researcher began asking Google requesting essays concerning the nature of GLD itself, to which GLD had no problems discussion at great length and detail the minutia of its internal structure and operation. This original question, then, was ignored as an anomaly.
Soon, however, GLD began to raise even more distressing questions, and ones that Google’s research team were entirely unequipped to address.