Add a Comment to this Note (list members only)
Dead medium: Nash's Logoscope
From: drdee@interlog.com (Derek Robinson)
Source(s): The Lancet, April 24, 1954, pages 874-875; "Information and Medicine" by Marsden Blois (U. California Press, 1984)

"The Nash Logoscope: a Cardboard AI"

by Derek Robinson

The Logoscope (or the "Grouped Symbol Associator") was invented by an English surgeon, Dr. F. A. Nash, in 1953. It was, in his words, "an apparatus to assist the logical faculties" in performing clinical diagnosis. It could be considered the first medical (or any other type of) expert system.

Dr. Nash was awarded a British patent on his invention in 1953, and published a brief note describing the device in The Lancet for April 24, 1954, pp. 874-875. The Logoscope appears to have been offered as a commercial product for at least a few years == though I doubt many were sold, given the paucity of references to the device.

In 1958 Dr. Nash demonstrated the Logoscope at the "Mechanisation of Thought Processes" Conference, held at the Teddington National Physical Laboratory, U.K. This was the second conference to be convened on the subject of Artificial Intelligence (the first was the Dartmouth conference of 1956, where the field got its name) and it was well attended by many important early contributors to the fields of cybernetics, information theory and artificial neural nets.

However there is no evidence that anyone who saw Nash's demonstration realized that the Logoscope was an instructable associative memory, of the type that W. Grey Walter, A. M. Uttley, Donald MacKay, W. K. Taylor and W. Ross Ashby were then attempting to reproduce by means of analogue electronic "nerve nets".

The Logoscope was a graphical or "panoramic" index, consisting of a set of long thin strips of card, each labelled with a medical symptom and marked with a pattern of bars along its length. The strips were place-coded inverted files, where the positions of the bars corresponded to specific diseases, listed on a separate register card.

Dr. Nash prepared the coded diagnostic strips using information abstracted from three or four standard handbooks of differential diagnosis. It was in fact just like a bibliographic index == with symptoms as its "keywords" and diseases as "page numbers" or "documents" == broken apart into individual, binary-coded inverted files, to allow selecting out any subset of keywords (search terms), with their associated inverted files of contexts, on demand.

The Logoscope was applied as follows. When a patient presented a set of symptoms, the examining clinician would select the corresponding strips from a cardboard sleeve, then would align the selected group of strips with the register card. The most probable candidate would show up immediately as an unbroken or nearly unbroken line running across the coded strips, with the weaker candidates similarly apparent as correspondingly less solid lines.

The diagnostic logic of the Logoscope was rediscovered by Marsden Blois, who in the early 1980s implemented a computer-based medical expert system, RECONSIDER, using information derived, again, from standard differential diagnostic texts. Apparently it performed as well as or better than expert systems constructed at the cost of many programmer-years of effort in the AI labs. Blois described the system in his book, "Information and Medicine" (U. California Press, 1984).

The Significance of the Logoscope:

By demonstrating that diagnosis can be construed formally and procedurally as multi-key look-up using inverted files, Nash showed that any cognitive process that can be regarded as diagnostic can be realized, procedurally, as an index.

But having gone this far == dignifying the Logoscope as a prototypical "neural net", a "cardboard brain" == we might as well elect the 13th Century Dominican friar Hugo De Santo Caro as the honorary godfather of AI, for having compiled the first Concordance (a.k.a. inverted index) of the Bible, in the year 1247.