By Mick Ashby
The W. Ross Ashby Digital Archive is available at www.rossashby.info. It contains an extensive biography, bibliography, letters, photographs, movies, fully-indexed images of Ross's journal, and other resources. The main resource in the archive is his journal. This paper provides more information about his journal, and explains how it was put on-line.
In May 1928, Ross was a 24 year-old medical student at Barts Hospital in London. He greatly admired Darwin's theory of evolution, which elegantly explained the origin of different species over geological timescales through the competitive selection of those individuals that are best suited to their environments. Ross wanted to understand the brain, and in particular, how individual organisms adapt to their environments, and how they learn and develop behaviors, which are not coded genetically, and are acquired in timescales ranging from perhaps fractions of a second to, at most, one lifetime. He also wanted to find mechanisms that could explain the conditioned reflexes that were described by Pavlov, which had been published in English a year earlier.
Ross started writing a journal. In it, he recorded his thoughts, theorems, and goals that would eventually bring him recognition as a pioneer in the fields of cybernetics and systems theory. It was his retreat away from the realities of life where he would "... weave delightful patterns of pure thought untroubled by social, financial and other distractions". After 44 years, his journal consisted of 7,189 pages, in 25 volumes. However, after Ross passed away at home in 1972, and for the subsequent 30 years, only members of his family had access to his journal.
By the year 2000, computers and scanners had become affordable and powerful enough for the family to digitize his notebooks. This was primarily done as a precaution against the originals being lost or destroyed, and then in January 2003, Ross's family gave the originals to The British Library, in London. In March 2004, at the W. Ross Ashby Centenary Conference, his family, who were very moved that there was still so much interest in Ross and his ideas, announced that they would allow Ross's journal to be made available on the Internet. How this would be achieved wasn't clear. At the time, most digital archives available on the Internet provided very literal versions of the originals, enabling Internet users to browse a collection's physical structure faithfully, as if they could open each box, folder, and envelope. However, that approach wasn't suitable for Ross's journal, which, even divided into 25 volumes, is a sequential monolithic work.
The biggest task for most archivists is probably cataloging; systematically creating descriptions of each item in the collection. Fortunately for us, Ross had already done that work by keeping a card index that contains about 5,500 references to pages in his journal for 678 different keywords. The index was the key to it all – it was our Rosetta Stone. So at the age of 73, and using Microsoft Excel for the first time in her life, Ross's daughter Jill typed in the entire index. Ross's grandson, John, who had done all the scanning, had also created a spreadsheet, which contained information about each scan, including the dates and places that Ross had recorded in some of his journal entries. So this multi-generational effort had produced about 100 GB of digital information consisting of two large spreadsheets and over 5,000 high resolution TIFF images, but it had no useful structure, and was very difficult to use. It quickly became clear that it wasn't feasible for us to create so many HTML pages manually, and we would have to write some programs to generate the pages for the web site. This turned out to be easier and more powerful than I originally imagined.
The first step was to export the two spreadsheets as XHTML, and to turn them into XML files that were suitable for automatic processing. Then, programming in XSL, it was possible to transform the information in the XML files to generate one HTML page for each journal scan, and an HTML index page that allowed users to browse the entire index and click on any page reference to view the corresponding scan's HTML page. This was already much more usable.
From this basic functionality we incrementally implemented improvements. For example, because this XML/XSL approach made it possible to write an XPath expression that returns the set of all keywords for which a given page was indexed, it was quite easy to add above the image of each indexed page, the index text, which Ross had entered in his card index, to provide an effective summary for the hand-written journal pages. We only wanted to make the web site publicly available when we felt that it was good enough that Ross would have given it his approval. And so, using feedback from people who were keen to have access to Ross's journal, we spent about two years of our spare time refining the password-protected beta-testing web site.
The main design priority was to make it as easy and enjoyable as possible to browse purposefully through Ross's vast intellectual legacy. For example, minimizing the click-path distances from any page in the journal to all other related pages – as defined by Ross's index. While we refined and extended the transformations that generate the web site, we also added some new information projections in the form of different views. Fortunately, the effort involved was not proportional to the number of pages in the journal, but to the number and complexity of the classes of information that we wanted to include on the web pages. Each small effort that was made extending a transformation was magnified across the whole set of pages, keywords, or references that it was generating. This made it easy to add tens of thousands of links and contextual pop-ups for related information.
For example, when viewing a journal page that Ross had indexed, clicking on one of the keywords listed above the image of the page takes you to the corresponding entry in the index, where you can see all the page numbers that Ross indexed for that primary keyword, then hovering the mouse pointer over any of those page numbers will display the year that Ross wrote them, and clicking on any of the page numbers displays the image of the appropriate journal page. This XML/XSL approach has certainly paid off, and now whenever we make any corrections or improvements, it takes less than four minutes to consistently regenerate all 10,000 pages, which make up 99.7% of the digital archive web site.
There are ten different logical views of the information in the journal and index:
All these views are densely cross-linked, which makes it possible for anyone who has access to the Internet to browse purposefully through Ross's journal much faster and more effectively than was ever possible for Ross himself, who was limited to using the original physical notebooks and card indexes.
As a result of propagating available information into these different views, it is now possible to quickly find the answers to certain classes of questions that could not generally be answered in a reasonable time using just the original notebooks and index cards. Examples of questions that are now trivial to answer include:
For this reason, it is probably fair to claim that the transformation programs have amplified the original information. On this same subject, on page 3610, written in December 1951, and indexed under “Information, amplifier for”, Ross wrote:
“I have realized that it is possible to build an information amplifier, with all the potentialities it holds.”
On page 3611, he goes on to describe how an information amplifier would work:
“In exactly the same way [as a power amplifier] we can get an amplifier for information. The designer gives up some of his control over the job, divesting that control or information or design into the 'machine'. This has available a great source of undirected information, and is designed to convert it (not all of it [but] by selecting from it) into a useful form.”
He continued...Figure: Journal page 3611.
In the context of generating the digital archive, Ross's terminology and figure illustrates the following:
On page 3613, Ross provides some examples of how questions would be answered within one hundredth of a second. Amazingly, this 57 year-old prediction is very close to the actual time that it takes to refresh my browser's window when I effectively ask a question by clicking on a link in my locally hosted version of the digital archive; Internet users might have to wait for a few seconds, depending on their bandwidth and latency to the public web server. As Ross wrote on page 3614: “Clearly the input is a pattern-demand, the output is a pattern that meets the demand.”
In October 1952, on journal page 4186, Ross wrote that the essence of intelligence is the ability to make an appropriate selection from the set of all possible answers. It was obvious to him that the power of selection could be amplified, and on page 4276, he reflected on the fact that although a man, whose muscles can only produce 0.1 horse power, can easily control a 1,000 horse power engine, which is 10,000 times more powerful than he is, what chance would men have to resist being controlled by a "million I.Q. engine"? He concluded: "They would be as children to it".
On the following page, Ross continued delightfully contemplating the "very strange", almost paradoxical, effects on the observer's perceptions when the relative intelligence of the observer and the observed changes over the range high/low to low/high. Avoiding any implicit insults to the observers, he considered the renormalized high/higher case, which implied that an intelligent observer can only ever perceive a "super-clever" machine as having an "I.Q of zero or infinity!" And as we've all come to expect from Ross, he explained it all using everyday examples that make it much easier for non-academics to understand than my previous two sentences.
When the user of a book, library, or the Internet seeks answers to questions, they are actually using that resource to augment their own intelligence. Clearly, Ross systematically recorded all his ideas on cybernetics and systems theory in his journal so that he could augment or amplify his own intelligence, when trying to solve a problem, by using his card index to efficiently select the appropriate journal entries from the set of all his previous thoughts and ideas.
Now, 80 years after Ross started writing his journal, it is with great pride in Ross, and with appreciation to his students and new generations of scientists and academics who value his ideas, that we make the "amplified intelligence" of W. Ross Ashby available on the Internet so that you too have the opportunity to augment your intelligence with his – before the million I.Q. engine makes children of us all!
Ashby, W. Ross, Journal (1928-1972), The W. Ross Ashby Digital Archive, 2008:
Pavlov, I.P. "Conditioned reflexes." Oxford. 1927.
Copyright 2008, 2010 © Mick Ashby. All rights reserved.
This version was updated in 2010 to reflect the addition of the new views "Other Index" and "Summaries".
An earlier version of this paper was published in the International Journal of General Systems Volume 38, Issue 2, February 2009, Special Issue: "The Intellectual Legacy of W. Ross Ashby", edited by Peter M. Asaro and George J. Klir.