RRisieseooff tthhee SeSlfe-lRf-Repeplilcicaattoorrs EaEralyrlyVVisiisoinosnsofofMMacahcihninese,sA, AI IaannddRRoobboottss ThTahtatCCaannRReepprroodduucceeaannddEEvvoollvvee TimTiTmayTaloylroarnadndAAlalannDDooriinn Rise of the Self-Replicators Tim Taylor is a scientist, author, and coder, based in Edinburgh, Scotland. He pursues research and development in artificial life, agent-based mod- elling and artificial intelligence, and also has a deep interest in the history of these fields. He has held research positions in various leading univer- sities in the UK and Australia, including Edinburgh, York and Goldsmiths. He currently works part-time (and remotely) as a Senior Research Fellow in the Department of Data Science and AI at Monash University, while also progressing his own research and writing interests as an independent researcher. He is an associate examiner for the University of London, and an elected board member of the International Society for Artificial Life. Alan Dorin is an Associate Professor at Monash University in Australia where he leads the Computational and Collective Intelligence group within the Department of Data Science and AI. His research interests include ar- tificial life, ecological modelling, insect-plant interaction and simulation, biologically-inspired electronic media art, generative art, and the history of technology. He promotes an understanding of the relationships between technology, society and the natural environment by encouraging interdis- ciplinary curiosity. Rise of the Self-Replicators Early visions of machines, AI and robots that can reproduce and evolve Tim Taylor and Alan Dorin RISE OF THE SELF-REPLICATORS EARLY VISIONS OF MACHINES, AI AND ROBOTS THAT CAN REPRODUCE AND EVOLVE Tim Taylor and Alan Dorin c Springer Nature Switzerland AG 2020 This is an author-formatted version of the text (v1.0.1) produced for personal self-archival purposes. There is no ISBN associated with this version of the book, and it is not intended for wide distribution or resale. Cover illustration: MC Escher – Building Hands by Simon Liu, used under the Creative Commons CC-BY-NC-SA 2.0 licence. https://www.flickr.com/photos/si-mocs/49833525251/ The Version of Record of this book can be purchased from the Springer website: https://www.springer.com/gb/book/9783030482336 Contents Preface v 1 Self-Reproducing Machines: The Evolution of an Idea 1 1.1 Two Central Questions . 1 1.2 The Promise of Self-Reproducing Machines . 2 1.3 Diverging Visions in the Early History of the Idea . 4 1.4 Relevance Today . 7 1.5 A Note on Scope and Terminology . 9 1.6 Outline of the Rest of the Book . 11 2 Animals and Machines: Changing Relationships in the 17th & 18th Centuries 15 2.1 Animals as Machines, Machines as Animals . 15 2.2 Mechanism and Design . 18 3 Babbage Meets Darwin: Mechanisation and Evolution in the 19th Century 25 3.1 Samuel Butler: Darwin Among the Machines (1863) . 28 3.2 Alfred Marshall: Ye Machine (c. 1867) . 33 3.3 George Eliot: Impressions of Theophrastus Such (1879) 37 3.4 The Late 19th Century . 39 4 Robot Evolution and the Fate of Humanity: Pop Culture and Futurology in the Early 20th Century 41 4.1 Literary Work . 41 4.1.1 E. M. Forster: The Machine Stops (1909) . 42 4.1.2 Karel Capek:ˇ R.U.R.: Rossum’s Universal Robots (1920) . 42 4.1.3 Early American Science Fiction (1920s–1950s) 43 i 4.2 Scientific Speculation in the Early 1900s . 48 4.2.1 J. D. Bernal: The World, The Flesh and the Devil (1929) . 49 4.2.2 The Widening Impact of Ideas . 54 5 From Idea to Reality: Designing and Building Self- Reproducing Machines in the Mid-20th Century 57 5.1 Theory of Logical Self-Reproduction . 58 5.1.1 John von Neumann (1948) . 58 5.2 Realisations of Logical Self-Reproduction . 64 5.2.1 Nils Aall Barricelli (1953) . 64 5.3 Realisations of Physical Self-Reproduction . 72 5.3.1 Lionel Penrose (1957) . 73 5.3.2 Homer Jacobson (1958) . 80 5.4 Scientific Speculation in the 1950s . 82 5.4.1 Edward F. Moore (1956) . 82 5.4.2 Konrad Zuse (1957) . 84 5.4.3 George R. Price (1957) . 86 5.5 Self-Reproduction in the Cybernetics Literature . 88 6 More Recent Developments: Signposts to Work from the 1960s to the Present 95 6.1 Theoretical and Philosophical Work . 96 6.2 Software Implementations . 100 6.3 Physical Implementations . 103 6.4 Addressing the Risks Associated with Self-Replicators 110 7 The Next Evolution: Reflection and Outlook 113 7.1 Narratives of Self-Replicators . 113 7.1.1 Takeover by Intelligent Machines . 114 7.1.2 Implications for Human Evolution . 115 7.1.3 Implications for Human Society . 116 7.1.4 The Narratives in Context . 117 7.2 Purpose and Goals of Research on Self-Replicators . 121 7.3 The Process of Self-Reproduction . 123 7.3.1 Maker-Replicators: The Top-Down Approach . 124 7.3.2 Evo-Replicators: The Bottom-Up Approach . 128 7.3.3 Top-Down and Bottom-Up Approaches Compared . 130 7.3.4 Drive for Ongoing Evolution . 132 7.4 Looking Forward . 134 Acknowledgements 139 Image Credits 141 Bibliography 143 Preface The origins of this book date back to 2014, during a period when one of the authors (TT) was working in the other’s (AD’s) lab at Monash University in Australia. We had first met at a conference on Artificial Life in 2002; our shared interests in the subject meant that we had kept in touch regularly since then, despite living on opposite sides of the world (TT in Edinburgh, Scotland, and AD in Melbourne, Australia). Artificial Life (or “ALife” for short) is the application of biolog- ical design principles for building complex, intelligent systems. It might be studied in software, hardware or “wetware” (molecular systems), and might be used for a variety of purposes. The two most common reasons people pursue ALife are as an approach to understanding biological systems and as an approach to building intelligent robots and artificial intelligence (AI) systems. It is also used by philosophers, social scientists, artists, and many others besides. The ALife conferences are an exciting interdisciplinary melting- pot of ideas. At the conference in 2002, we soon discovered that we shared very similar interests in designing artificial worlds in software. We both used computational analogues to the processes of biological self-reproduction, evolution and natural selection to populate our worlds with interesting creatures. We had both com- pleted PhDs in this area in the 1990s. Beyond our experimen- tal work, we also shared an interest in early mechanical models of living systems, and in the history of thought about technology inspired by biology. During our period of working together in 2014, our original inspiration for writing this review came from reading a recently published account [40] of Alfred Marshall’s ideas of machine self- v reproduction and evolution from the 1860s (we discuss his work in Sect. 3.2). Although we were aware of Samuel Butler’s writing on the subject at around the same time (Sect. 3.1), we had not come across Marshall in this context before. We were therefore curious whether there might be other work from around this time, or even earlier, that discussed such ideas. Another motivation was to highlight the pioneering work in the early 1950s of Nils Aall Barricelli on self-reproduction and evolu- tion in software (Sect. 5.2.1). While John von Neumann’s theo- retical work on self-reproducing machines from around that time is widely discussed in the literature (Sect. 5.1.1), personal experi- ence suggested that Barricelli was still a relatively unknown figure within the ALife community, despite having a strong claim to be regarded as one of the field’s founding fathers. Our purpose in writing this book was therefore to review the early history of the idea of self-reproducing and evolving machines, tracing it back as far as we could. This being the case, much of the book (Chaps. 2–6) is written as a guide to the literature on the subject, presented in chronological order from the earliest inklings of the idea up to the present day. While we provide some commen- tary and suggest classifications of the work in terms of the goals of the authors we survey, our primary aim is to present a compre- hensive archive of thought about self-reproducing machines. These chapters represent the most extensive early history of the subject published to date and include coverage of many works that have not been widely discussed elsewhere. We do provide a synthesis and summary of the concepts discussed in Chap. 7, and it is there that we offer more of our own views on the field and where we see it heading. The audience we have in mind includes anyone wishing to learn about the origins of the idea of self-reproducing and evolving ma- chines, especially those interested in drawing lessons from this early work regarding likely future developments in the field. Most obviously, the audience will be Artificial Life and Artificial Intelli- gence practitioners. We also believe the subject will be of interest to many philosophers, biologists, engineers, historians of science, and those working in the emerging field of AI safety and ethics. We hope the content will be of value in informing a wider general readership too. For that reason, in Chap. 1 we discuss the profound future implications of the technology and explain why it is a subject of broad relevance. We have tried to make the text accessible and to avoid technical jargon, although this has not always been possible. In particular, in Chap. 5 we discuss at greater length the details of the first realisations of self-reproducing machines in the 1950s, and in Chap.