.
1941: Konrad Zuse builds first working general computer—patent application 1936. Juergen Schmidhuber.

Jürgen Schmidhuber (2021)
Pronounce: You_again Shmidhoobuh
German version: Weltwoche, 19 Aug 2021 (also online)
AI Blog
@SchmidhuberAI


1941: Konrad Zuse completes the first working general-purpose computer, based on his 1936 patent application


In 2021, we are celebrating the 80th anniversary of Konrad Zuse's crowning achievement: Z3, the world's first functional program-controlled general computer, based on his patent application from 1936. Today, computers are ubiquitous. Billions of people depend on them. Only 20 years to go until the Z3 centennial in 2041!


Between 1935 and 1941, Konrad Zuse created the world's first working programmable computer 1935-41 Konrad Zuse (1910-1995; pronounce: "Conrud Tsoosay") created the world's first working programmable general-purpose computer: the Z3. The corresponding patent application of the "father of the computer" dates back to 1936.[ZU36-38][Z36][RO98] In 1946, he also founded the world's first computer startup company: the Zuse-Ingenieurbüro Hopferau (IBM provided some of the venture capital for an option on Zuse's patents).

As if that was not enough to cement Zuse's legacy in computing, in the early 1940s, Zuse also designed Plankalkül, the first high-level programming language[BAU][KNU] (compare the first formal language by Gottlob Frege[FRE]). He applied it to chess in 1945[KNU] and to theorem proving in 1948.[ZU48]

In 1967, he also suggested what's known as Zuse's thesis, namely, that physics is computable, and that the universe is computed by some sort of cellular automaton.[ZU67][ZU69][ALL2]

Where does Zuse's Z3 fit in the history of computing? The first known gear-based computational device was the Antikythera mechanism (a kind of astronomical clock) in Ancient Greece over 2000 years ago. 1.5 millennia later, Peter Henlein still made conceptually similar machines—albeit smaller—namely, the first miniaturized pocket watches (1505). But these devices always calculated the same thing, e.g., divide minutes by 60 to get hours. The 1600s brought more flexible machines that computed answers in response to input data. The first data-processing gear-based special purpose calculator for simple arithmetics was built in 1623 by Wilhelm Schickard, one of the candidates for the title of "father of automatic computing," followed by the superior Pascaline of Blaise Pascal (1642).

Leibniz, father of computer science around 1670 In 1673, Gottfried Wilhelm Leibniz designed the first machine (the step reckoner) that could perform all four arithmetic operations, and the first with a memory.[BL16] He also described the principles of binary computers governed by punch cards (1679),[L79][L03][LA14][HO66] and defined the formal Algebra of Thought (1686)[L86][WI48][LEI21,a,b] which is deductively equivalent[LE18] to the later Boolean Algebra[BOO] (1847). Leibniz, one of the candidates for the title of "father of computer science," has been called "the world's first computer scientist"[LA14] and even "the smartest man who ever lived."[SMO13] He was not only the first to publish infinitesimal calculus,[L84] but also pursued an ambitious project to answer all possible questions through computation. His ideas (inspired by Ramon LLull[LL7]) on a universal language and a general calculus for reasoning (Characteristica Universalis & Calculus Ratiocinator[WI48][RU58]) were highly influential.[GOD21,a,b][WI48]

In the early 1930s, however, Kurt Gödel dealt a blow to Leibniz' project. He created a universal language for encoding arbitrary formalizable processes,[GOD][GOD34] and used his so-called Gödel Numbering to show that there are fundamental limitations to what can be decided or computed,[GOD] thus laying the foundations of the modern version of what's now known as theoretical computer science.[GOD21,a,b]

The pragmatic Konrad Zuse was apparently not particularly interested in such theoretical work. In 1936, five years after Gödel's famous publication,[GOD] he filed his patent application on a very practical real computer.[ZU36-38][Z36][RO98] It described the digital circuits required by programmable physical hardware, extending Leibniz' principles of binary computers governed by punch cards (1679),[L79][LA14][HO66][L86][WI48][LEI21,a,b] and predating Claude Shannon's thesis on digital circuit design (1937).[SHA37]

Zuse's Z3 of 1941 lacked an explicit conditional jump instruction "IF ... THEN GOTO ADDRESS ..." (added with little effort to a later variant for ETHZ called Z4). Of course, this did not prevent Z3 from being a universal computer.[RO98] For example, simple arithmetic tricks (e.g., multiplication by 0) can be used to temporarily make a no-op out of every instruction that should not be executed because some condition is not fulfilled.[RO98] Ignoring the inevitable storage limitations of any physical computer, the physical hardware of Z3 was indeed universal in the modern sense of the purely theoretical but impractical constructs of Gödel[GOD][GOD34,21,21a] (1931-34), Church[CHU] (1935), Turing[TUR] (1936), and Post[POS] (1936), which also did not allow for "modern" conditional jumps (they did not even have numbered memory addresses to which an instruction pointer could have jumped).

Zuse's Z3 used electromagnetic relays with visibly moving switches. The first electronic special purpose calculator (whose moving parts were electrons too small to see) was the binary ABC (US, 1942) by John Atanasoff (the "father of tube-based computing"[NASC6a]). Unlike the gear-based machines of the 1600s, ABC used vaccum tubes—today's machines use the transistor principle patented by Julius E. Lilienfeld in 1925.[LIL1-2] But unlike Zuse's Z3, ABC was not freely programmable. Neither was the electronic Colossus machine by Tommy Flowers (UK, 1943-45) used to break the Nazi code.[NASC6]

On the other hand, the concept of programs was well-known by then. Perhaps the world's first practical programmable machine was an automatic theatre made in the 1st century[SHA7a][RAU1] by Heron of Alexandria (who apparently also had the first known working steam engine—the Aeolipile). The energy source of his programmable automaton was a falling weight pulling a string wrapped around pins of a revolving cylinder. Complex instruction sequences controlling doors and puppets for several minutes were encoded by complex wrappings.

The 9th century music automaton by the Banu Musa brothers in Baghdad was perhaps the first machine with a stored program.[BAN][KOE1] It used pins on a revolving cylinder to store programs controlling a steam-driven flute—compare Al-Jazari's programmable drum machine of 1206.[SHA7b]

The first commercial program-controlled machines (punch card-based looms) were built in France around 1800 by Joseph-Marie Jacquard and others—perhaps the first "modern" programmers who wrote the world's first industrial software. They inspired Ada Lovelace and her mentor Charles Babbage (UK, circa 1840). He planned but was unable to build a programmable, general purpose computer (only his non-universal special purpose calculator led to a working 20th century replica). Unlike Babbage, Zuse (1936-41) used Leibniz' principles of binary computation (1679)[L79][LA14][HO66][L03] instead of traditional decimal computation. This greatly simplified the hardware.[LEI21,a,b] Today, most computers are binary like Z3.

In this context it seems worth pointing out the difference between programs and the more limited user-given input data of the 1600s mentioned above. Programs are instruction sequences stored on some medium, e.g., on punch cards, and can be run again and again, without human intervention. Over time the physical objects required to store programs have become lighter. Ancient machines stored them on rotating cylinders; Jacquard stored them on cardboard; Zuse stored them on 35mm film, today we often store them using electrons and magnetizable material.

The first general working programmable machine built by someone other than Zuse (1941)[RO98] was Howard Aiken's decimal MARK I (US, 1944). The much faster decimal ENIAC by Eckert and Mauchly (1945/46) was programmed by rewiring it. Both data and programs were stored in electronic memory by the "Manchester baby" (Williams, Kilburn & Tootill, UK, 1948) and the 1948 upgrade of ENIAC, which was reprogrammed by entering numerical instruction codes into read-only memory.[HAI14b] Already in 1936-38, however, Zuse may have been the first to suggest to put both program instructions and data into memory.[ZU36-38]

Leonardo Torres y Quevedo, the  20th century's first pioneer of practical AI While Zuse's work on automatic chess players (1945)[KNU] and theorem provers (1948)[ZU48] (predating Newell & Simon's work[NS56]) was groundbreaking, it was not the first work on Artificial Intelligence (AI). Already in 1914, the Spaniard Leonardo Torres y Quevedo became the 20th century's first AI pioneer when he built the first working chess end game player (back then chess was considered as an activity restricted to the realms of intelligent creatures). The machine was still considered impressive decades later when the AI pioneer Norbert Wiener played against it at the 1951 Paris conference,[AI51][BRO21][BRU1-4] which is now often viewed as the first conference on AI—though the expression "AI" was coined only later in 1956 at another conference in Dartmouth by John McCarthy. In fact, in 1951, much of what's now called AI was still called Cybernetics, with a focus very much in line with modern AI based on deep neural networks.[DL1-2][DEC]

In 1941, Zuse's Z3 could perform roughly one elementary operation (e.g., an addition) per second. Since then, every 5 years, compute got 10 times cheaper (note that his law is much older than Moore's Law which states that the number of transistors[LIL1-2] per chip doubles every 18 months). As of 2021, 80 years after Z3, modern computers can execute about 10 million billion instructions per second for the same (inflation-adjusted) price. The naive extrapolation of this exponential trend predicts that the 21st century will see cheap computers with a thousand times the raw computational power of all human brains combined.[RAW] Where are the physical limits? According to Bremermann (1982),[BRE] a computer of 1 kg of mass and 1 liter of volume can execute at most 1051 operations per second on at most 1032 bits. The trend above will hit the Bremermann limit roughly 25 decades after Z3, around 2200. However, since there are only 2 x 1030 kg of mass in the solar system, the trend is bound to break within a few centuries, since the speed of light will greatly limit the acquisition of additional mass, e.g., in form of other solar systems, through a function ploynomial in time, as previously noted back in 2004.[OOPS2]

In 1970, long before computers had become ubiquitous, Peter's renowned Atlas of World History already listed Zuse among the 20th century's 30 most important figures, along with Einstein, Gandhi, Hitler, Lenin, Roosevelt, Mao, Picasso, etc. Zuse's historical importance has only grown with the exponential growth of computing since then. By the turn of the millennium, more than 80 streets and squares carried the name of Zuse. A collection of his writings and pictures of his machines can be found in this online archive.

Kurt Goedel, founder of theoretical computer science around 1931 In 2021, we are not only celebrating the 80th anniversary of Zuse's 1941 computer, but also the 90th anniversary of Kurt Gödel's groundbreaking 1931 paper[GOD][GOD21,a,b] which laid the foundations of theoretical computer science and the theory of AI. Gödel identified the fundamental limits of theorem proving, computing, AI, logics, and mathematics itself.[GOD][GOD34,21][BIB3] This had enormous impact on science and philosophy of the 20th century. It seems incredible that within less than a century something that once lived only in the minds of titans has become something so inalienable from modern society. The world owes these scientists a great debt. Ten years to go until the Gödel centennial in 2031, and twenty years until the Zuse centennial in 2041! Enough time to plan appropriate celebrations.


Acknowledgments

Creative Commons License Thanks to Moshe Vardi, Herbert Bruderer, Jack Copeland, Wolfgang Bibel, Teun Koetsier, Scott Aaronson, Dylan Ashley, Sebastian Oberhoff, Kai Hormann, and several other expert reviewers for useful comments on the contents of the four companion articles.[LEI21,a,b][GOD21,a,b][ZUS21,a,b][TUR21] Since science is about self-correction, let me know under juergen@idsia.ch if you can spot any remaining error. The contents of this article may be used for educational and non-commercial purposes, including articles for Wikipedia and similar sites. This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.


References

[GOD] K. Gödel. Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I. Monatshefte für Mathematik und Physik, 38:173-198, 1931.

[GOD34] K. Gödel (1934). On undecidable propositions of formal mathematical systems. Notes by S. C. Kleene and J. B. Rosser on lectures at the Institute for Advanced Study, Princeton, New Jersey, 1934, 30 pp. (Reprinted in M. Davis, (ed.), The Undecidable. Basic Papers on Undecidable Propositions, Unsolvable Problems, and Computable Functions, Raven Press, Hewlett, New York, 1965.)

[GOD21] J. Schmidhuber (AI Blog, 2021). 90th anniversary celebrations: 1931: Kurt Gödel, founder of theoretical computer science, shows limits of math, logic, computing, and artificial intelligence. This was number 1 on Hacker News.

[GOD21a] J. Schmidhuber (2021). Als Kurt Gödel die Grenzen des Berechenbaren entdeckte. (When Kurt Gödel discovered the limits of computability.) Frankfurter Allgemeine Zeitung, 16/6/2021.

[GOD21b] J. Schmidhuber (AI Blog, 2021). 80. Jahrestag: 1931: Kurt Gödel, Vater der theoretischen Informatik, entdeckt die Grenzen des Berechenbaren und der künstlichen Intelligenz.

[BIB3] W. Bibel (2003). Mosaiksteine einer Wissenschaft vom Geiste. Invited talk at the conference on AI and Gödel, Arnoldsheim, 4-6 April 2003. Manuscript, 2003.

[FRE] G. Frege (1879). Begriffsschrift: eine der arithmetischen nachgebildete Formelsprache des reinen Denkens. Halle an der Saale: Verlag Louis Nebert. [The first formal language / formal proofs—basis of modern logic and programming languages.]

[L79] G. Leibniz. De Progressione dyadica Pars I. 15 March 1679. [Principles of binary computers governed by punch cards.]

[L03] G. Leibniz (1703). In: Explication de l'Arithmetique Binaire / Die Mathematischen Schriften, ed. C. Gerhardt, Berlin 1879, vol.7, p.223. English link. [Leibniz documented the binary arithmetics which allow for greatly simplifiying computing hardware and are employed by virtually all modern computers. Binary number encodings per se, however, seem to date back over 4000 years.]

[L84] G. Leibniz (1684). Nova Methodus pro Maximis et Minimis. [First publication on infinitesimal calculus.]

[L86] G. Leibniz (1686). Generales Inquisitiones de analysi notionum et veritatum. Also in Leibniz: Die philosophischen Schriften VII, 1890, pp. 236-247; translated as "A Study in the Calculus of Real Addition" (1690) by G. H. R. Parkinson, Leibniz: Logical Papers - A Selection, Oxford 1966, pp. 131-144.

[LEI21] J. Schmidhuber (AI Blog, 2021). 375th birthday of Leibniz, founder of computer science.

[LEI21a] J. Schmidhuber (2021). Der erste Informatiker. Wie Gottfried Wilhelm Leibniz den Computer erdachte. (The first computer scientist. How Gottfried Wilhelm Leibniz conceived the computer.) Frankfurter Allgemeine Zeitung (FAZ), 17/5/2021. FAZ online: 19/5/2021.

[LEI21b] J. Schmidhuber (AI Blog, 2021). 375. Geburtstag des Herrn Leibniz, dem Vater der Informatik.

[BOO] George Boole (1847). The Mathematical Analysis of Logic, Being an Essay towards a Calculus of Deductive Reasoning. London, England: Macmillan, Barclay, & Macmillan, 1847.

[LL7] A. Bonner (2007). The art and logic of Ramon Llull. Brill Academic Pub, p. 290, 2007.

[RU58] B. Russell (1958). The Philosophy of Leibniz. London: George Allen and Unwin, 1958.

[LE18] W. Lenzen. Leibniz and the Calculus Ratiocinator. Technology and Mathematics, pp 47-78, Springer, 2018.

[LA14] D. R. Lande (2014). Development of the Binary Number System and the Foundations of Computer Science. The Mathematics Enthusiast, vol. 11(3):6 12, 2014. Link.

[BL16] L. Bloch (2016). Informatics in the light of some Leibniz's works. Communication to XB2 Berlin Xenobiology Conference.

[HO66] E. Hochstetter et al. (1966): Herrn von Leibniz' Rechnung mit Null und Eins. Berlin: Siemens AG.

[WI48] N. Wiener (1948). Time, communication, and the nervous system. Teleological mechanisms. Annals of the N.Y. Acad. Sci. 50 (4): 197-219. [Quote: "The history of the modern computing machine goes back to Leibniz and Pascal. Indeed, the general idea of a computing machine is nothing but a mechanization of Leibniz's calculus ratiocinator."]

[SMO13] L. Smolin (2013). My hero: Gottfried Wilhelm von Leibniz. The Guardian, 2013. Link. [Quote: "And this is just the one part of Leibniz's enormous legacy: the philosopher Stanley Rosen called him "the smartest person who ever lived"."]

[CHU] A. Church (1935). An unsolvable problem of elementary number theory. Bulletin of the American Mathematical Society, 41: 332-333. Abstract of a talk given on 19 April 1935, to the American Mathematical Society. Also in American Journal of Mathematics, 58(2), 345-363 (1 Apr 1936). [First explicit proof that the Entscheidungsproblem (decision problem) does not have a general solution.]

[TUR] A. M. Turing. On computable numbers, with an application to the Entscheidungsproblem. Proceedings of the London Mathematical Society, Series 2, 41:230-267. Received 28 May 1936. Errata appeared in Series 2, 43, pp 544-546 (1937). [2nd explicit proof that the Entscheidungsproblem (decision problem) does not have a general solution.]

[TUR21] J. Schmidhuber (AI Blog, Sep 2021). Turing Oversold. It's not Turing's fault, though.

[POS] E. L. Post (1936). Finite Combinatory Processes - Formulation 1. Journal of Symbolic Logic. 1: 103-105. Link.

[NASC6] J. Schmidhuber. Colossus was the first electronic digital computer. Correspondence, Nature, 441 p 25, May 2006.

[NASC6a] J. Schmidhuber. Comment on "Biography: The ABC of computing" by J. Gilbey, Nature 468 p 760-761 (2010). Link.

[ALL2] J. Schmidhuber (2000). Algorithmic theories of everything. ArXiv: quant-ph/ 0011122. More. See also: Hierarchies of generalized Kolmogorov complexities and nonenumerable universal measures computable in the limit. International Journal of Foundations of Computer Science 13(4):587-612, 2002. PDF. More. See also: The Speed Prior: A New Simplicity Measure Yielding Near-Optimal Computable Predictions. Proc. COLT 2002. PDF. More.

[DL1] J. Schmidhuber, 2015. Deep Learning in neural networks: An overview. Neural Networks, 61, 85-117. More.

[DL2] J. Schmidhuber, 2015. Deep Learning. Scholarpedia, 10(11):32832.

[MIR] J. Schmidhuber (AI Blog, 2019). Deep Learning: Our Miraculous Year 1990-1991. Preprint arXiv:2005.05744, 2020.

[DEC] J. Schmidhuber (AI Blog, 2020). The 2010s: Our Decade of Deep Learning / Outlook on the 2020s.

[HAI14] T. Haigh (2014). Historical reflections. Actually, Turing did not invent the computer. Communications of the ACM, Vol. 57(1): 36-41, Jan 2014. PDF.

[HAI14b] T. Haigh, M. Priestley, C. Rope (2014). Reconsidering the Stored-Program Concept. IEEE Annals of the History of Computing. IEEE, 2014. PDF.

[VAR13] M. Y. Vardi (2013). Who begat computing? Communications of the ACM, Vol. 56(1):5, Jan 2013. Link.

[ZU36] K. Zuse (1936). Verfahren zur selbsttätigen Durchführung von Rechnungen mit Hilfe von Rechenmaschinen. Patent application Z 23 139 / GMD Nr. 005/021, 1936. [First patent application describing a general, practical, program-controlled computer.]

[ZU37] K. Zuse (1937). Einführung in die allgemeine Dyadik. [Mentions the storage of program instructions in the computer's memory.]

[ZU38] K. Zuse (1938). Diary entry of 4 June 1938. [Description of computer architectures that put both program instructions and data into storage—compare the later "von Neumann" architecture [NEU45].]

[ZU48] K. Zuse (1948). Über den Plankalkül als Mittel zur Formulierung schematisch kombinativer Aufgaben. Archiv der Mathematik 1(6), 441-449 (1948). PDF. [Apparently the first practical design of an automatic theorem prover (based on Zuse's high-level programming language Plankalkül).]

[ZU67] K. Zuse (1967). Rechnender Raum, Elektronische Datenverarbeitung, vol. 8, pages 336-344, 1967. PDF scan.

[ZU69] K. Zuse (1969). Rechnender Raum, Friedrich Vieweg & Sohn, Braunschweig, 1969. English translation: Calculating Space, MIT Technical Translation AZT-70-164-GEMIT, MIT (Proj. MAC), Cambridge, Mass. 02139, Feb. 1970. PDF scan.

[BRE] H. J. Bremermann (1982). Minimum energy requirements of information transfer and computing, International Journal of Theoretical Physics, 21, 203-217, 1982

[RAW] J. Schmidhuber (AI Blog, 2001). Raw Computing Power.

[OOPS2] J. Schmidhuber. Optimal Ordered Problem Solver. Machine Learning, 54, 211-254, 2004. PDF. HTML. HTML overview. Download OOPS source code in crystalline format.

[NS56] A. Newell and H. Simon. The logic theory machine—A complex information processing system. IRE Transactions on Information Theory 2.3 (1956):61-79.

[RO98] R. Rojas (1998). How to make Zuse's Z3 a universal computer. IEEE Annals of Computing, vol. 19:3, 1998.

[BAU] F. L. Bauer, H. Woessner (1972). The "Plankalkül" of Konrad Zuse: A Forerunner of Today's Programming Languages.

[KNU] D. E. Knuth, L. T. Pardo (1976). The Early Development of Programming Languages. Stanford University, Computer Science Department. PDF.

[Z36] S. Faber (2000). Konrad Zuses Bemühungen um die Patentanmeldung der Z3.

[SHA37] C. E. Shannon (1938). A Symbolic Analysis of Relay and Switching Circuits. Trans. AIEE. 57 (12): 713-723. Based on his thesis, MIT, 1937.

[NEU45] J. von Neumann (1945). First Draft of a Report on the EDVAC.

[AI51] Les Machines a Calculer et la Pensee Humaine: Paris, 8.-13. Januar 1951, Colloques internationaux du Centre National de la Recherche Scientifique; no. 37, Paris 1953. [H. Bruderer rightly calls that the first conference on AI.]

[BRU1] H. Bruderer. Computing history beyond the UK and US: selected landmarks from continental Europe. Communications of the ACM 60.2 (2017): 76-84.

[BRU2] H. Bruderer. Meilensteine der Rechentechnik. 2 volumes, 3rd edition. Walter de Gruyter GmbH & Co KG, 2020.

[BRU3] H. Bruderer. Milestones in Analog and Digital Computing. 2 volumes, 3rd edition. Springer Nature Switzerland AG, 2020.

[BRU4] H. Bruderer. The Birthplace of Artificial Intelligence? Communications of the ACM, BLOG@CACM, Nov 2017. Link.

[BRO21] D. C. Brock (2021). Cybernetics, Computer Design, and a Meeting of the Minds. An influential 1951 conference in Paris considered the computer as a model of—and for—the human mind. IEEE Spectrum, 2021. Link.

[BAN] Banu Musa brothers (9th century). The book of ingenious devices (Kitab al-hiyal). Translated by D. R. Hill (1979), Springer, p. 44, ISBN 90-277-0833-9. [Perhaps the Banu Musa music automaton was the world's first machine with a stored program.]

[KOE1] T. Koetsier (2001). On the prehistory of programmable machines: musical automata, looms, calculators. Mechanism and Machine Theory, Elsevier, 36 (5): 589-603, 2001.

[RAU1] M. Rausch. Heron von Alexandria. Die Automatentheater und die Erfindung der ersten antiken Programmierung. Diplomica Verlag GmbH, Hamburg 2012. [Perhaps the world's first programmable machine was an automatic theatre made in the 1st century by Heron of Alexandria, who apparently also had the first known working steam engine.]

[SHA7a] N. Sharkey (2007). A programmable robot from AD 60. New Scientist, Sept 2017.

[SHA7b] N. Sharkey (2007). A 13th Century Programmable Robot. Univ. of Sheffield, 2007. [On a programmable drum machine of 1206 by Al-Jazari.]

[ZUS21] J. Schmidhuber (AI Blog, 2021). 80th anniversary celebrations: 1941: Konrad Zuse completes the first working general computer, based on his 1936 patent application.

[ZUS21a] J. Schmidhuber (AI Blog, 2021). 80. Jahrestag: 1941: Konrad Zuse baut ersten funktionalen Allzweckrechner, basierend auf der Patentanmeldung von 1936.

[ZUS21b] J. Schmidhuber (2021). Der Mann, der den Computer erfunden hat. (The man who invented the computer.) Weltwoche, Nr. 33.21, 19 August 2021. PDF.

[LIL1] US Patent 1745175 by Austrian physicist Julius Edgar Lilienfeld for work done in Leipzig: "Method and apparatus for controlling electric current." First filed in Canada on 22.10.1925. [The patent describes a field-effect transistor. Today, almost all transistors are field-effect transistors.]

[LIL2] US Patent 1900018 by Austrian physicist Julius Edgar Lilienfeld: "Device for controlling electric current." Filed on 28.03.1928. [The patent describes a thin film field-effect transistor.]
.

Highlights of over 2000 years of computing history. Juergen Schmidhuber.