A Clear Vision of the Information

"The best window is one so clear you can't see it."

I wish to take a slightly different view point, and propose an "ultimate" library automation system. This provides a sufficiently distant viewpoint from which to examine what we can do in the near term.

The quotation at the start of this paper expresses the philosophy: the automation should be so smooth that it is completely invisible to the user. The information retrieval system that most of us use with least effort is our own mind. We remember things that we have learned in various ways and can see the relationships between all these different items with great facility. It is important to note that each of us has a very personal set of both the information and of the relationships.

If I posit two moderately extreme technological advances, it is possible to construct a truly transparent system. One advance would be a technology that would allow a direct broadband bidirectional connection between the user's nervous system and a computer. The other advance posits a computer with sufficient power to run a real time model of the user's memory strategy. A less radical advance for this model of the user would call for a data base engine with online access to any information sources that the user wishes to access.

Given this technology, a search becomes merely the activity of thinking about a problem. The interface will capture the intention, and the user model will decode it. The decoded clues will be passed to the database engine which will retrieve the requested data. The data will be passed back through the model of the user, which will place it into the user's mind where the user would have placed such data had it been gathered without any technological assistance.

The subjective experience would be like looking into your own memory for information and finding it where you would have put it if you already knew it. The dataspace can be explored in as much richness and detail as desired merely by exerting will. This is the first level of a proper interface to a library.

The second level deals with the problem of how to transparently aid the user in dealing with information that is unfamiliar or has no preexisting context. This is the usual problem we have when we first explore a new area. This breaks down into two large pieces: following the reference chains to all related information and understanding the information in its proper context.

I propose that placing both the article and the pointers to related articles and other forms of information at their "natural" locations in the user's memory lies within the proper definition of the level two interface. Actually understanding the information in its natural context does seem to largely eliminate the need for the user. Systems that truly understand what they are finding are true artificial intelligences as opposed to mere information retrieval aids.

At this point I expect many of the people reading this to have concluded that I read too much science fiction and that this vision cannot be attained. I will first state that even if we never realize this vision, it still provides a standard against which to measure our efforts. Second, what follows if a brief justification for my belief that we can, in time, attain this vision.

We are now at the very beginning of a technological revolution whose results may well dwarf those of both the industrial and computer revolutions. This revolution is our newfound ability to understand and manipulate biological systems at the deepest levels. This will provide us with both the engineering ability to grow direct neural interfaces and the understanding of the working of the brain that will be needed to construct proper "persona analogs" that can run in external hardware, or wetware.

One of the things we are beginning to learn is how the nervous system is assembled. A particularly intriguing aspect of this process is that we customize our nervous systems to our personal environments during the first two years of life. This customization is done by deleting neurons that are not needed to respond to absent stimuli. A well known example of this is the relatively low visual sensitivity of Eskimos to vertical lines.

The space that had been occupied by deleted neurons is filled in with glial cells. If we recreate the chemical gradients that guided the initial assembly of the brain we can, perhaps, induce the growth of new neurons through the available volume of glial cells to provide both input and output taps at interesting locations in the brain. This is a possible scenario for growing a direct neural interface without actually putting hardware in our heads. While I am quite sure that the proposed method will likely be wrong in detail, I still feel confident that with enough work in developmental neurophysiology we will be able to design a workable procedure.

The planetary net is well on its way to being constructed. Currently available technologies will be able to support terabit/second data rates in single optical fibers. Computational speed and storage densities are still rising at exponential rates. These existing trends will deal with the back end database problem.

The really interesting challenge is in the construction of the hardware substrate, and the software, of the persona analogs that will translate between our minds and the back-end processors.

Even after we have managed to place an adequate amount of the right kind of processor in a handy black box, we still need to be able to extract and keep current the model of how each of us thinks, sorts, and remembers. The interaction between ourselves and the analog has many truly intriguing potential pitfalls. They range from avoiding destructive positive feedback loops to establishing the legal status of a black box that thinks like a human being.

Setting the technological details aside and returning to the original thesis, I conclude that developers of library automation systems should aspire to perfect transparency for the user.