Germany's Computing Pioneers

Abstract

Germany’s contribution to computing spans four centuries: from Leibniz’s binary arithmetic through the Z3 to the invention of the stack, Petri nets, and LSTM. The state of Hesse passed the world’s first data protection law in 1970, and Karl Steinbuch coined the German word Informatik in 1957, five years before the French informatique. This article profiles German figures from academia, industry, and law who shaped how computers work, how they are programmed, and how their social consequences are governed, many of whom worked in isolation, in wartime, or in a divided country.

The reconstruction of Konrad Zuse’s Z3 at the Deutsches Museum, Munich. Image: Venusianer at German Wikipedia, CC BY-SA 3.0, via Wikimedia Commons.

The Calculating Ancestors

German computing prehistory begins with two figures whose work predates electronics by centuries.

Wilhelm Schickard (1592–1635), a professor at the University of Tübingen, built the earliest known mechanical calculator in 1623, a “Rechenuhr” (calculating clock) designed to assist his friend Johannes Kepler with astronomical computations. Schickard’s device could add six-digit numbers automatically and ring a bell on overflow. He described it in letters to Kepler and had a copy built for the astronomer by a clockmaker, but it was destroyed in a fire before delivery. Schickard died of plague in 1635, during the Thirty Years’ War. No working copy survived; the machine was reconstructed from his letters in 1960. Schickard’s priority over Blaise Pascal’s better-known Pascaline (1642) was established only when those letters were rediscovered in the twentieth century.

Gottfried Wilhelm Leibniz (1646–1716) went further. Between 1671 and 1694 he built the Staffelwalze (Stepped Reckoner), a mechanical calculator capable of multiplication and division, not merely addition. More consequential than the machine was the mathematics behind it. Leibniz independently developed binary arithmetic (the representation of all numbers using only 0 and 1) which he published in 1703 in Explication de l’Arithmétique Binaire. He saw in binary a philosophical elegance: all of creation expressed through two symbols. He did not build a binary calculating machine; binary arithmetic entered computing only in the 1930s, with Zuse’s Z1 and Claude Shannon’s switching-circuit analysis.

Two forgotten designs connect them to Zuse. On 30 September 1926 Emil Schilling, a lawyer in the Reich finance ministry, filed a patent for a “control device for calculating machines,” granted in 1933: a program control using compressed air, with a pneumatic-electric variant, coupled to an existing adding machine. It lapsed in 1934 for unpaid fees but was cited in later German patents, among them one of Gustav Tauschek’s. And in 1933 the Zeitschrift für Instrumentenkunde described a relay and rotary-switch device that Adolf Weygandt, born in 1908, had built as a student in Göttingen to evaluate a mathematical formula automatically.

The Enigma Builder

A German Enigma cipher machine. Image: Rama, CC BY-SA 2.0 fr, via Wikimedia Commons.

Arthur Scherbius (1878–1929) was an electrical engineer in Berlin who, in February 1918, patented a rotor-based cipher machine he called Enigma. The device used a series of rotating electrical rotors to substitute each typed letter through an ever-changing polyalphabetic cipher; reading the traffic required knowing both the machine’s wiring and the daily key settings. Scherbius failed to sell Enigma to the German military immediately after the First World War, and with his partner Richard Ritter developed commercial models instead; the Chiffriermaschinen Aktiengesellschaft that made it was founded in Berlin on 9 July 1923, at the height of the inflation, with a nominal capital of 500 million marks; the Navy adopted it in 1926, the Army in 1928.

Scherbius died after a horse-carriage accident in 1929, years before the Wehrmacht’s mass use of his machine. The British decryption effort at Bletchley Park, building on work by Polish cryptanalysts, broke Enigma with the electromechanical Bombe designed by Alan Turing and Gordon Welchman. The separate effort against the German Lorenz teleprinter cipher produced Colossus, the first programmable electronic digital computer. The full story of Enigma’s breaking is told in Alan Turing and the Enigma.

Germany’s own check on the machine fell partly to the young mathematician Gisbert Hasenjaeger (born in Hildesheim on 1 June 1919), who tested Enigma’s security in the Wehrmacht high command’s cipher department and missed its decisive weakness (see Alan Turing and the Enigma). After the war he became a professor of mathematical logic at Bonn from 1962 and built working Turing machines.

Naming the Discipline: Karl Steinbuch

Karl Steinbuch speaking in Munich, 1975. Image: Storz, Bundesarchiv B 145 Bild-F046476-0002, CC BY-SA 3.0 de, via Wikimedia Commons.

Karl Steinbuch (1917–2005), then head of development at Standard Elektrik Lorenz (SEL) and from 1958 a professor at the Technische Hochschule Karlsruhe, published an article in 1957 titled “Informatik: Automatische Informationsverarbeitung”, the first use of Informatik for automatic information processing. The word is a portmanteau of Information and Automatik. In France, Philippe Dreyfus coined informatique in 1962 for his company Société d’informatique appliquée; the Académie française adopted it in 1967, and the word, German or French, spread to Russian (informatika), Spanish (informática), Italian (informatica) and many other languages. English mostly kept “computer science”.

Steinbuch, born in Stuttgart on 15 June 1917, had led the building of the Informatik-System Quelle at SEL, which handled orders for the mail-order firm Quelle from late 1957 and which the Heinz Nixdorf MuseumsForum counts as Germany’s first transistor computer, before taking his chair at Karlsruhe in 1958. His contributions went beyond terminology. In 1961 (the HNF dates the invention to 1960) he published his design for the Lernmatrix, the main subject of a conference on learning automata at Karlsruhe on 13 and 14 April 1961, a hardware neural network model in which a matrix of connections could be adjusted by experience to recognize patterns. The Lernmatrix was an early hardware associative memory, comparable to the American perceptron work of the same years. In the late 1960s Steinbuch turned to futurology and politics, warning in his 1968 bestseller Falsch programmiert that West Germany was unprepared for the coming information society.

German AI research acquired a meeting of its own in February 1975, when researchers gathered at the University of Bonn.

The First Electronic Machines

West Germany’s first electronic computer came from a physicist’s side project. Heinz Billing (1914–2017), born in Salzwedel, had invented a magnetic drum memory in Göttingen; with Marshall Plan money he built the G1, which went into service at the Max Planck Institute for Physics on 7 June 1952 with 476 tubes, 101 relays and a drum, followed by the G2 and the G3 after the institute moved to Munich in 1958. The PERM (Programmgesteuerte Elektronische Rechenanlage München) at the Technical University of Munich, begun in 1951 under Hans and Robert Piloty, with Friedrich L. Bauer and Klaus Samelson in Robert Sauer’s programming group, went into operation on 8 May 1956 as the first computer at a West German university: 2,400 tubes, 3,000 diodes and a drum for 8,192 words of 51 bits. In Darmstadt Alwin Walther (born 1898), head of the Institute for Practical Mathematics from the 1930s to 1966, had designed the “System Darmstadt” slide rule in 1934 and ran a wartime calculating office of more than thirty women computers that supported the A4 rocket; after the war he became the public face of electronic computing, organised the 1955 Darmstadt conference and built the DERA. Wolfgang Giloi (1930–2009) designed Telefunken’s RA 770, often called the best analog computer ever built, before pioneering computer graphics at the Technical University of Berlin from 1965 and directing the GMD’s FIRST research centre in Berlin from 1983, where the SUPRENUM supercomputer was built. In the East, Nikolaus Joachim Lehmann (1921–1998), a professor at Dresden from 1953 to 1986, built the GDR’s first electronic computer and in 1959 a relay teaching computer with von Neumann architecture (see Computing in East Germany); in the West, the Göttingen firm Phywe sold a didactic relay computer from 1967 that grew out of a Jugend forscht project.

Inventing the Stack: Friedrich Bauer and Klaus Samelson

Friedrich L. Bauer. Image: Hcii, CC BY-SA 3.0, via Wikimedia Commons.

The stack, the data structure that underlies function calls and expression evaluation in nearly every programming language, had several inventors, and two of them worked at the Technical University of Munich.

Klaus Samelson (1918–1980) and Friedrich Ludwig Bauer (1924–2015) developed the concept between 1955 and 1957, calling it the Kellerspeicher (cellar store, or push-down store). Their insight was that nested computational expressions (arithmetic with parentheses, function calls within function calls) could be evaluated correctly and efficiently by maintaining a last-in, first-out stack of intermediate values and return addresses. Alan Turing had described a stack for subroutine returns (“bury” and “unbury”) in his 1946 ACE report, and the Australian Charles Hamblin developed a stack-based expression evaluator independently in the 1950s, but Bauer and Samelson made it the basis of systematic formula translation by compilers.

Bauer and Samelson filed a patent on the concept in 1957 and published it in 1959, and applied it in their work on ALGOL; stack-based translation became standard in ALGOL 60 compilers and their successors. Bauer went on to co-found the Gesellschaft für Informatik (GI) in 1969 (Germany’s computer science professional society) and to establish the Munich school of programming theory, which shaped a generation of German computer scientists.

Bauer’s influence reached well beyond the stack. He served on the committees that designed both ALGOL 58 and ALGOL 60, and he initiated the NATO Science Committee conference “Software Engineering” in Garmisch in October 1968. The term had appeared in print by 1965, but the conference made software engineering the name of a discipline (see The Software Crisis and Friedrich L. Bauer). He later defined it as the “establishment and use of sound engineering principles to economically obtain software that is reliable and works on real machines efficiently.” Bauer was also one of Germany’s leading academic cryptologists: his textbook Entzifferte Geheimnisse (Decrypted Secrets), which ran through four editions, connected the mathematical tradition of cryptology to computer science teaching. The IEEE honored him with the Computer Pioneer Award in 1988.

Info

Every time a program calls a function and returns from it correctly (maintaining local variables, passing arguments, resuming execution at the right point) it relies on a stack. The concept appears in hardware (the x86 stack pointer), in virtual machines (the JVM operand stack), and in every recursive algorithm. The IEEE’s 1988 Computer Pioneer Award to Bauer cited the stack principle.

Concurrent Systems: Carl Adam Petri

Carl Adam Petri. Image: Michael Krapp, derivative work AwOc, CC BY-SA 3.0, via Wikimedia Commons.

Carl Adam Petri (1926–2010) solved a problem that von Neumann architecture sidesteps: how to model systems where multiple processes happen simultaneously and interact with shared resources.

Petri’s 1962 doctoral dissertation at TH Darmstadt, “Kommunikation mit Automaten” (Communicating with Automata), introduced what became known as Petri nets, a mathematical formalism using places (states), transitions (events), and tokens (resources) to model concurrency. A Petri net can represent deadlock, livelock, and race conditions formally, making them analyzable rather than merely observable. The formalism proved applicable to operating systems, communication protocols, manufacturing processes, biological systems, and workflow modeling.

Petri spent most of his career at the Gesellschaft für Mathematik und Datenverarbeitung (GMD) in Bonn, developing the theoretical foundations of concurrent computation in a period when the field was dominated by sequential models. His nets became a standard tool for formal verification of distributed systems, used to prove that a concurrent system cannot deadlock instead of merely testing that it has not deadlocked yet.

The Forgotten Mouse: Rainer Mallebrein

The Telefunken Rollkugel RKS 100-86. Image: Marcin Wichary, CC BY 2.0, via Wikimedia Commons.

Rainer Mallebrein (born 1934) led a Telefunken team in Konstanz that built a trackball console for the Bundesanstalt für Flugsicherung (German air-traffic control). In 1966, prompted by a university customer, he “reversed” the trackball into a movable device, a mouse, developed with no knowledge of Douglas Engelbart’s parallel work at SRI. The Rollkugel RKS 100-86 was finished in early 1968 and publicly offered on October 2, 1968 (two months before Engelbart’s Mother of All Demos) as an optional input device (price: 1,500 DM) for the SIG 100-86 vector terminal of Telefunken’s TR 86 and TR 440 computers.

Technically it was ahead of its American cousin: where Engelbart’s mouse ran on two perpendicular wheels, the Rollkugel already used the rolling ball that Bill English would introduce at Xerox PARC only in 1972, the standard design until the optical mouse. But Telefunken’s patent application was reportedly rejected for “mangelnde Erfindungshöhe” (insufficient inventive step), only 46 systems were sold (mostly to German universities) and the invention was forgotten for decades. In 2019, Mallebrein donated one of the few surviving Rollkugeln to the Heinz Nixdorf MuseumsForum in Paderborn.

The Cipher Foundation: Horst Feistel

Horst Feistel (1915–1990) was born in Berlin, emigrated to the United States in 1934, worked on cryptography for the US Air Force and MIT Lincoln Laboratory, and joined IBM Research in the late 1960s. In the early 1970s he designed the Feistel cipher, a symmetric encryption structure in which plaintext is divided into two halves and processed through multiple rounds of mixing, with each round using a subkey derived from the master key. The structure is invertible: encryption and decryption use the same algorithm with the keys applied in reverse order.

The Feistel structure, first used in IBM’s Lucifer cipher, became the basis for DES (Data Encryption Standard), adopted by the US National Bureau of Standards in 1977 as the first standardized encryption algorithm for civilian use. Triple DES, Blowfish and Twofish are also Feistel ciphers; AES, its successor as the US standard, uses a different structure, a substitution-permutation network. The Feistel construction makes a cipher reversible even when its round function is not. The full history of cryptographic development is covered in Cryptography: The Secret Science.

Industry Builders

Heinz Nixdorf at RWE in Essen, 1952. Image: Heinz Nixdorf MuseumsForum, CC BY-SA 3.0 de, via Wikimedia Commons.

Heinz Nixdorf (1925–1986) founded the Labor für Impulstechnik in Essen in 1952 (bankrolled by a DM 30,000 advance from the utility RWE) and grew it into Germany’s most successful independent computer company; in 1968 he acquired Wanderer-Werke and renamed the firm Nixdorf Computer AG, headquartered in his native Paderborn. Nixdorf specialized in minicomputers and point-of-sale systems, reaching annual revenues of nearly DM 4 billion by the mid-1980s. Nixdorf died of a heart attack at the CeBIT trade fair in 1986; after heavy losses under his successor, the company merged into Siemens Nixdorf Informationssysteme in 1990 (the fuller story is in Germany’s IT Industry).

The SAP founders (Dietmar Hopp, Hasso Plattner, Hans-Werner Hector, Klaus Tschira, and Claus Wellenreuther) left IBM Germany in 1972 to build integrated business software. Hasso Plattner (born 1944) became the most publicly visible of the five, shaping SAP’s technical vision through the transition from mainframe to client-server (SAP R/3, 1992) and later to in-memory computing (SAP HANA, 2010). Plattner was the internal advocate for the HANA architecture, a column-store, in-memory database that eliminated the performance bottleneck separating transactional and analytical systems. SAP’s story as a company is covered in Germany’s IT Industry; the founders had been building such a system for an IBM customer and left to complete it on their own.

Connecting Germany to the Internet: Werner Zorn

Werner Zorn. Image: Reinhard Ferdinand, CC0, via Wikimedia Commons.

On August 3, 1984, the first Internet e-mail reached Germany, at the University of Karlsruhe, over a link to CSNET, the American Computer Science Network. The connection had been built by the group of Werner Zorn, a professor who pursued international connectivity against institutional resistance and the telecommunications monopoly of the Deutsche Bundespost.

The Bundespost treated data communication as a postal monopoly and was suspicious of external network connections that bypassed its infrastructure. Zorn negotiated, lobbied, and demonstrated that academic collaboration required international connectivity, arguments that succeeded incrementally. The 1984 CSNET connection was followed by a full TCP/IP connection in 1989 and by Germany’s participation in the public internet’s rapid growth in the early 1990s. Britain had been connected to the ARPANET since 1973, so Germany came late, but it came through the universities rather than the Bundespost.

Privacy as Technology Policy: Simitis and Steinmüller

Germany’s most distinctive contribution to the social governance of computing came from legal scholars, who argued that the new technology required new rights.

Wilhelm Steinmüller (1934–2013), a professor at the University of Regensburg, introduced the term informationelle Selbstbestimmung (informational self-determination) in a 1971 expert report for the federal interior ministry, written with Bernd Lutterbeck and Christoph Mallmann; it holds that individuals have a fundamental right to control how information about themselves is collected, stored, and used. This framing, grounded in the German constitutional tradition of human dignity (Menschenwürde), was more philosophically radical than American privacy law, which treated privacy primarily as freedom from government intrusion. Steinmüller argued that data about a person is part of that person’s identity; to collect and use it without consent is to appropriate something that belongs to them.

Spiros Simitis (1934–2023) put the idea into practice. As Hessian Data Protection Commissioner from 1975 to 1991, the second holder of the office, he administered the Hessisches Datenschutzgesetz, the Hessian Data Protection Act of 1970, the world’s first data protection statute. Under Simitis, Hesse became a testing ground for data protection enforcement: he challenged corporate data collection practices, advised European legislators, and argued consistently that the distinction between public and private sector data collection was artificial, since both threatened individual autonomy.

The ideas that Steinmüller and Simitis developed became the philosophical foundation of Germany’s 1977 federal data protection law (Bundesdatenschutzgesetz), of the 1983 German Federal Constitutional Court ruling on the census (Volkszählungsurteil) that enshrined informational self-determination as a constitutional right, and, through European law, of the EU’s General Data Protection Regulation (GDPR), applied from 2018 to every company serving users in the EU, including Google, Facebook, Apple, and Amazon.

Warning

The Nazi regime used population registers and census data to identify and persecute Jews, Roma, and political opponents. The East German Stasi kept files on about a third of the GDR’s population, built from informants, intercepted mail, and physical surveillance. German privacy advocates had specific historical evidence for their fears.

Modern AI Pioneers

Jürgen Schmidhuber (born 1963 in Munich) developed, with his student Sepp Hochreiter, the Long Short-Term Memory (LSTM) architecture, published in 1997. LSTMs solved the vanishing gradient problem that had prevented recurrent neural networks from learning long-range dependencies in sequences: gradients used to train the network shrank exponentially as they propagated backward through time, making early inputs irrelevant to the training signal. The LSTM’s gating mechanism (learned controls over what to remember and what to forget) maintained information across long sequences. LSTMs became the dominant architecture for speech recognition, machine translation, and time-series prediction through the 2010s, and were directly used in Google’s speech recognition, Apple’s Siri, Amazon’s Alexa, and early neural machine translation systems. Schmidhuber also published early work on meta-learning, artificial curiosity, and program search.

Bernhard Schölkopf (born 1968), director of the Max Planck Institute for Intelligent Systems in Tübingen, is one of the principal developers of kernel methods (techniques for learning patterns in high-dimensional data): he extended the kernel trick (representing data implicitly in very high-dimensional spaces without computing those representations explicitly) from support vector machines, introduced by Vladimir Vapnik and colleagues, to methods such as kernel PCA, and co-wrote the standard textbook Learning with Kernels (2002). Schölkopf has more recently focused on causality in machine learning: the difference between correlation, which statistical models learn, and causal relationships, which enable reliable generalization to new conditions. It addresses why models that perform well on training data fail in deployment when the world changes.

Kurt Mehlhorn (born 1949), founding director of the Max Planck Institute for Informatics in Saarbrücken in 1990, works in computational geometry, graph algorithms, and algorithm engineering, the science of implementing theoretical algorithms so that they perform well in practice, not just in asymptotic analysis. With Stefan Näher he created LEDA (Library of Efficient Data Types and Algorithms), a software library that demonstrated how theoretically sound algorithms could be packaged for practical use. He led the institute’s algorithms department for decades and trained many of Europe’s algorithms researchers.

Émigrés and Educators

Some of computing’s German contributors made their careers abroad because they had to leave. Among the half-million people who emigrated from Germany and Austria as Jews between 1933 and 1941 were Werner Michael Blumenthal (born in Oranienburg in 1926), who fled via Shanghai, served in President Carter’s cabinet from 1977 to 1979, and from 1981 ran Burroughs, which he merged with Sperry into Unisys, before directing the Jewish Museum Berlin from 1997 to 2014; Ralph Baer, born in Rodalben in 1922, father of the video-game console; Werner Buchholz from Detmold, who coined the word “byte” at IBM; and Emanuel Goldberg, whose microfilm “statistical machine” of the 1920s anticipated search engines. Others built the discipline at home: Christiane Floyd, born in Vienna in 1943, became the first woman professor of computer science in Germany at the Technical University of Berlin in 1978 and taught at Hamburg from 1991 to 2008, and Helmar Frank, professor of cybernetic pedagogy at Paderborn from 1972 to 2006, worked with Heinz Nixdorf on teaching machines.

The GI and the Academic Infrastructure

The Gesellschaft für Informatik (GI), founded in 1969 by Bauer and colleagues, institutionalized German computer science in the way that the ACM and IEEE did for American computing. The GI established standards for the discipline, organized conferences, published research, and lobbied for computer science’s recognition as an independent academic field, at a time when German universities classified computing as a branch of mathematics or engineering.

Germany also has two large research organizations active in computing: the Fraunhofer Society (applied research, with a computing division that spans embedded systems, security, and AI) and the Max Planck Society (basic research, with institutes for informatics in Saarbrücken, software systems in Kaiserslautern, and intelligent systems in Tübingen). Both institutions operate on a model unusual outside Germany: state-funded, with academic independence, pursuing research with long time horizons. Schölkopf’s causality work and Mehlhorn’s algorithm engineering came out of the Max Planck system.