The other half of Europe

What was missing

The atlas had five records about Kyiv: Ivakhnenko’s method, Vintsyuk in 1968, the Institute of Cybernetics, Glushkov’s book and the TAIR robot. It had none about Poland, Czechoslovakia, Hungary or the GDR before 1991. The Institute of Cybernetics record named the Computing Centre of 1957, yet the machine that centre inherited was nowhere in the atlas. The direction the 6-in-1 set is the same: the prehistory of a line matters more than its tail.

Nine records

Six records are from Kyiv. MESM was first started in Lebedev’s laboratory on 6 November 1950, and on 25 December 1951 a government commission accepted it into regular operation. That same day it solved its first real problem in two and a half hours, nearly 250,000 operations. The Computing Centre grew out of that laboratory, and the Institute of Cybernetics out of the centre. Amosov published a book on modelling thought in 1965. The MIR machines (1965-1969) transformed formulas in a language the machine itself interpreted. In 1968 Schlesinger reduced learning without a teacher to alternating recognition and learning, and proved that the likelihood rises; nine years later statistics would call the same scheme the EM algorithm. In 1970 Glushkov set out the “evidence algorithm” programme for proofs, and in 1971 Vintsyuk carried dynamic programming from single words to connected speech.

Three records are from Prague, Warsaw and Białystok. The GUHA method of 1966 had the machine propose every hypothesis of a given form that holds in the data; it was tried on 230 epileptic patients. Pawlak’s rough sets appeared in October 1982 as an alternative to fuzzy sets. The Mizar library, which admits an article only after a machine has checked every line, began on 1 January 1989.

Where the document said less

As before, I wrote the hypotheses down before opening any source. Three agents searched for the documents, and I checked every claim against a verbatim excerpt and the full text. MESM’s acceptance act was not found anywhere, so its dates come from the list of stages Lebedev kept in his notes and from Malinovsky’s article. The sources give the machine’s speed as three different figures, and the record gives all three. Only the first two pages of Glushkov’s 1970 paper are open, and they are about automata theory. The evidence algorithm record therefore rests on his students’ retrospectives and has medium confidence. For Mizar I expected 1973, but no document of that year exists, so the record is dated by the start of the library. Two claims from my hypotheses are not in the documents. One is that MIR was the first production machine with a high-level language. The other is that Amosov’s 1965 book already describes the system TAIR stands on. Hungary and the GDR got no records: Rényi’s paper was read, but no atlas record builds on it.

The language of the sources

The Kyiv papers of the 1960s are in Russian, and the publisher puts their translations behind a bot check. So they were read from the first pages of the translations, from scans that the Kyiv departments keep themselves, and from the pages of the Kyiv computer history museum. No Russian organisation became an actor in any record, and no Russian website was opened. The records describe the works’ titles in Ukrainian. The bounds of the records, sources and organisations pages and of the first payload were raised as the pass’s first action. After the package the tightest are the axis data (/uk/timeline/data.json, 2,315 bytes left) and the people page.