Who? Machines. (Part 2)

By Charles Henry posted

Part 2 of a 3-part series
My January 9 blog highlighted the contemporary paradox of new computers that will reach the market later this year that are designed to learn, and the digital environment we have inherited that is structured by silos and buckets of tightly sequestered information—in this instance cataloging schemes developed during the last 150 years. The paradoxical aspect relates to a quote from the 1940s by Vannevar Bush, who lamented that library subject categories were artificial demarcations that inhibited the way the human mind wants to explore: across boundaries and through ranging methods of association. In 2014, we may well have machines that can re-code their programs through experience, without recourse to human engineers. Yet still we rely on an organization of knowledge that is frequently segregated by topic and discipline, an organizational method that represents “top level” information about the material in question but not access to the content itself.

It is interesting to ponder the implications of computers modeled on bio-neurological circuitry. Such machines should be able to “read” very large quantities of information (thousands of novels, hundreds of thousands of legal transcripts, millions of images, hundreds of millions of data points) rapidly, and analytically extrapolate patterns from this information. The results of that analysis will change the “weight” of the connections between circuits, much like our reading experience changes the strength of connections between synapses in our brains. For example, I cannot interpret passages of Beowulf without reference to Anglo-Saxon riddles: I read those riddles while studying Beowulf, and the cross-genre linguistic similarities seemed to enrich both the poem and the riddles creating, at least in my mind, a conversation. The sophisticated, polysemous aspects of language pertain to the dark, brooding story of an aging warrior as well as to the dancing metaphors of word puzzles.

A machine that operates on wiring that allows for experienced-based re-wiring would similarly interpret data in light of previous “reading.” Conceptually this is intriguing, but the implications are deeper still. If I own a neurological-designed computational device, then I can direct the reading of, say, all nineteenth-century novels and set parameters on what I would like to investigate: what questions would I like answered? I might well fasten my own interpretation of those novels to the machine's findings, but since these “readings” are not simply key word searches or disconnected instances of discernible patterns, this machine is expressing a modulated sequence of interconnected information. It is a kind of understanding learned over time, a meshed set of cognitive analyses adjusted and recalibrated much like I would do as a human reader.

Such a scenario redefines “personal computer” powerfully. The machine is not proffering disaggregated facts and figures but something closer to theory. Depending on the degree of authority I invest in the machine and its analytical acumen, where does the boundary between us lie? After all, I submitted the questions that structured these inquiries, and the machine performed its rules of reading at my instigation. It can analyze volumes of information in seconds, volumes that would take me several lifetimes, but with quantity may well come a more complex and refined quality of understanding. Who is to say? We are learning together, I presume as a matter of my choosing.

To state the obvious, the new neuro-designed machine can perform optimally only with access to very large arrays of digital content. Surrogate records—bibliographic control—for that content as a locating tool is of no value. Many of our catalogs today, which include dozens of different schemes and organizational rules, can be traced to Anthony Panizzi's "Ninety One Cataloging Rules," published in 1841. It was a set of descriptions that were meant to aid the finding of physical materials in a physical place. Then and now, catalogs, whether on cards or digitally dressed and flashing, have their roots in a system that was fashioned to address the challenges of information proliferation in the Industrial Age. They are artifacts of a three-dimensional reality that is of vanishing consequence for new methods of discovery and interpretation, or so I, with my new machine, think.