Genesis Machines E-Book

Martyn Amos

• Herausgeber: Atlantic Books
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 573

Veröffentlichungsjahr: 2014

Beschreibung

Silicon chips are out. Today's scientists are using real, wet, squishy, living biology to build the next generation of computers. Cells, gels and DNA strands are the 'wetware' of the twenty-first century. Much smaller and more intelligent, these organic computers open up revolutionary possibilities. Tracing the history of computing and revealing a brave new world to come, GenesisMachines describes how this new technology will change the way we think not just about computers - but about life itself

Leseprobe

Genesis Machines

DR MARTYN AMOS was awarded the world’s first Ph.D. in DNA computing; he is currently a Senior Lecturer in Computing and Mathematics at Manchester Metropolitan University, UK.

His website is at http://www.martynamos.com.

‘To me this is perhaps the most fascinating and potentially important area of science there is at the moment and this book is an excellent introduction’ Richard Jones FRS, Professor of Physics, University of Sheffield, author of Soft Machines: Nanotechnology and Life

‘Fascinating… Amos describes such experiments beautifully, combining laboratory drama with technical explanations. His lucid and punchy prose conveys a genuine excitement of the frontier.’ Steven Poole, Guardian

‘What do encryption, the double helix and sudoku have in common? They are all bound together… in rather surprising ways, as Martyn Amos masterfully shows in this compendious volume. Amos is a born communicator, that rare breed among scientists who write fluently in an understandable and approachable way about difficult concepts.’ Tony Valsamidis, Times Higher Education Supplement

‘It is hard not to share Amos’s excitement as the computational possibilities of the DNA revolution become clear… Amos makes the science accessible, with well-plotted and nicely structured explanations. It’s clear that this field will continue to throw up dramatic advances, even if we don’t quite know what, yet… Genesis Machines provides a fine introduction to those wishing to follow its progress.’ Roly Allen, New Statesman

‘An elegant primer on a mind-blowing technology that could change our lives out of all recognition… On the way to describing… this new science, Amos provides lucid histories of mathematics, computing, the invention of the integrated circuit and discovery of DNA, all of which are improbably knockabout and entertaining… As compelling as anything by Isaac Asimov or Philip Dick… For an early and intriguing glimpse of one possible future, Genesis Machines is highly recommended.’ Andrew Smith, Mail on Sunday

‘Amos has witnessed the early years of this nascent science and writes about them with affection and enthusiasm. To make the case that computers might one day be made out of DNA, he recounts potted histories of computing, mathematics, molecular biology and various pieces of physics, which he does with a sure touch.’ Matt Ridley, Sunday Times

‘Sounding like something out of a futuristic science fiction thriller, this book tackles the topic of what kind of computers the future holds in store for us – based on fact, and not fiction… A fascinating insight into tomorrow’s world.’ Good Book Guide

First published in Great Britain in 2006 by Atlantic Books, an imprint of Grove Atlantic Ltd.

This papback edition published by Atlantic Books in 2007.

Copyright © Martyn Amos 2006

The moral right of Martyn Amos to be identified as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act of 1988.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of both the copyright owner and the above publisher of this book.

Every effort has been made to contact copyright holders. The publishers will be pleased to make good any omissions or rectify any mistakes brought to their attention at the earliest opportunity.

ISBN 978 1 84354 225 4 E-Book ISBN 978 1 78239 491 4

A CIP catalogue record for this book is available from the British Library.

1 2 3 4 5 6 7 8 9

Design by Lindsay Nash Typeset by Avon DataSet Ltd, Bidford on Avon, Warwickshire Printed in

Atlantic Books An imprint of Grove Atlantic Inc. Ormond House 26–27 Boswell Street London WC1N 3JZ

For Alice, with love

/ Contents

Acknowledgements

Prologue

Introduction

1The Logic of Life

2Birth of the Machines

3There’s Plenty of Room at the Bottom

4The TT–100

5The Gold Rush

6Flying Fish and Feynman

7Scrap-heap Challenge

Epilogue

Notes

Index

/ Acknowledgements

First thanks must go to my publisher, Toby Mundy, who saw the first seeds of a book in my entry to the Wellcome Trust prize. Without Toby’s enthusiastic support, constant encouragement and gentle marshalling of the project, this book would never have been written. Thank you Toby, for the opportunity. I have benefited greatly from the input of my two editors; Alice Hunt helped me considerably in the early stages of the project, before she left Atlantic for academia, and Sarah Castleton provided marvellous moral support, editorial advice and a friendly ear at all times. I was privileged to have Annabel Huxley handle the early publicity for the book, which she did admirably. On that note, I also thank Jonathan Black at the Royal Institution, Stephen Emmott of Microsoft Research, Oliver Morton at Nature, Johnjoe McFadden, Caspar Hewett and John Burn for helping with book-related events.

My own work has benefited enormously from the support and collaboration of many gifted people. Foremost among these is Alan Gibbons, my Ph.D. supervisor at Warwick, and now a good friend and colleague. His unstinting support, friendship, guidance and willingness to go off-piste in search of new scientific terrain were the primary factors in my being in a position to write this book from the perspective of an insider. David Hodgson has also proved to be a wonderful collaborator, gently demolishing our more outlandish ideas while being open to new challenges, offering incisive biological analysis and supervising the laboratory work with impeccable precision. On that note, I must also thank Gerald Owenson and Steve Wilson, our indefatigable post-docs, who stuck with it long after most would have thrown in the towel.

I thank the following for conversations, collaborations, support and advice: Len Adleman, Charles Cantor, Dave Corne, Chris Cox, Paul Dunne, Brian Goodwin, Lila Kari, Laura Landweber, the late Ray Paton, Mike Poppleton, Somasundaram Ravindran, Grzegorz Rozenberg, Paul Sant, Dennis Shasha, Mike Simpson Ron Weiss and Eric Winfree.

I am immensely grateful to Alan Gibbons, Jim Shapiro and Paul Rothemund for reading and commenting on drafts of the manuscript; any errors that remain are, of course, my sole responsibility.

I thank my parents, for their unfailing love and support, but most of all I thank my wife, Justine Ashby, for everything.

/ Prologue

Stanford, California – June 2015

The shiny black slab stood on a low mound of grass, marble glinting in the hazy early morning sunlight. Inscribed in gold on one side of the sign were the initials ‘ABC’, and, beneath these, the full corporate name, ‘Advanced BioComputing’. The ABC labs and administrative offices were housed in a low, U-shaped white structure surrounding a paved courtyard, where early starters congregated to drink coffee and discuss science.1

As bioengineer Neal Mendal pulled his car around the long, gentle sweep towards the main car park, his mind began to focus on the day’s work that awaited him. He worked in a second-floor Level 2 containment laboratory at the heart of the complex. Each lab was graded according to the relative risk of the organisms manipulated within it; laboratories with the highest 4-rating were used by specialist teams in oxygen suits working on microbes such as the deadly Ebola virus. Neal’s corporation, on the other hand, dealt with relatively benign creatures, and no such elaborate containment facilities were required. Even so, he still had to swipe his card through a reader at the main door and then pass a biometric retinal scan to gain entry to his laboratory.

As he sat in his office, waiting for his morning coffee to brew, Neal began to muse on the nature of his work. Back in the twentieth century, software engineers had implemented programs by meticulously designing components and implementing them in an unambiguous programming language. How different the job was now, Neal thought. The processing units that he wrote his programs for were not built from silicon, but from colonies of living cells. Neal’s job was to develop methods of phrasing instructions in the language of the cells, so that they could then go about their work. Instead of learning a traditional programming language, Neal had been trained in the language of biological differentiation and pattern formation. By manipulating the genetic material of cells and then altering their environment, Neal could coax colonies of cells into performing human-defined computations that ‘traditional’ computers struggled with. As the smell of fresh coffee filled his office, Neal found himself pondering, as he often did, the ‘magical’ process occurring within the organic modules. He could still barely imagine data being transformed and manipulated by living cells, however hard he tried. Somehow it was easier to imagine the much simpler operation of symbol processing in traditional computer systems.

Neal’s first task of the day was to replace the nutrients in the main processing unit. He flipped open the covers on a couple of nutrient cases, tossed the old cartridge in the trash and, rather more gently, dropped the replacement into place. He waited to see the clear liquid seep down the inclined surface, just in case the cartridge seal hadn’t punctured properly. The organic modules were far too valuable to risk letting them run dry.

As Neal waited for the nutrient broth to fill the processor case, he wandered around his lab, noting the usual mix of smells. He had always been told that the organic computing modules were sealed units, but nevertheless they always seemed to exude some low-level odours that gave a unique sensory profile any modern-day system administrator would recognise. In any case, the chemicals were harmless and at low concentrations, posing no threat to the human staff. Neal was more concerned that contaminants might inadvertently enter the organic computing modules and affect their proper functioning. With relief, he noted that everything appeared to be normal, as each module was exhibiting the typical patterns of fluorescent green scintillations with which he had become so familiar. He could now judge by eye when a module had been contaminated or had developed some aberrant behaviour. In any case, the modules were inherently self-healing in nature, and would adapt to any minor problems by reconfiguring themselves. Neal chuckled to himself as he recalled that, decades ago, people would complain that their computers ‘had a life of their own’. His computer was different. It was alive.

/ Introduction

In 1985, Greg Bear published Blood Music,2 a novel that established its author’s reputation and led him to being heralded as ‘the next Arthur C. Clarke’. The science-fiction magazine Locus lauded it as ‘A Childhood’s End for the 1980s, replacing aliens and mysterious evolution with the effects of genetic engineering run wild.’ In the book, a brilliant microbiologist works to develop biochips, using DNA as the next ‘quantum leap’ in computer technology. As his work progresses, Vergil I. Ulam3 develops intelligent cellular colonies that appear to exhibit intelligence way beyond that of ‘higher’ creatures. In one memorable section, Ulam observes groups of trained cells running through a complex miniature glass maze to earn nutritional rewards, just like laboratory rats scurrying for food.

I was sent a copy of Blood Music in 1999 by a thoughtful delegate who had recently attended a talk I’d delivered to a computer conference in Madrid. This was not simply a random act of generosity by a stranger who had just happened to enjoy a presentation of mine. The particular choice of book was motivated precisely by the content of my talk, in which I had described ongoing work that, only a decade or so previously, had been mere fantasy, imagined only in the pages of a science-fiction novel.

My talk was part of a ‘Frontiers of Computing’ event organized by the Unisys Users’ Association, during which several speakers were invited to present their vision of the future of computers in the years and decades to come. Nicholas Negroponte, the founder of MIT’s Media Lab, spoke about Being Digital,4 while Wim van Dam from Oxford University gave a presentation on quantum computing, the notion that quantum states of atoms could somehow be harnessed to build machines of almost unimaginable power.5 I was invited to speak about a growing research area that had existed in practice for just five years.

This book tells the story of a whole new connection between two ancient sciences: mathematics and biology. Just as physics dominated the second half of the twentieth century with the atomic bomb, the moon landing and the microchip, it’s becoming increasingly clear that the twenty-first century will be characterised and defined by advances in biology and its associated disciplines. Cloning, stem cells and genetic engineering have become the new hot topics of debate in newspapers and on the Web. New genome sequences are being produced at an astonishing rate, capturing the genetic essence of organisms as diverse as the orang-utan and the onion.6 This flood of sequence data is transforming biology from a lab-based discipline into a whole new information science. The pioneers at the forefront of this genomic unravelling speak of ‘networks’, ‘data mining’ and ‘modelling’, the language of computer science and mathematics. The sequencing of the human genome, one of the triumphs of the modern scientific age, was only made possible through the development of sophisticated mathematical algorithms to piece together countless DNA sequence fragments into a coherent whole. The growth of the Web has led to unprecedented levels of scientific collaboration, with researchers across the globe depositing gene sequences into communal databases in a distributed effort to understand the fundamental processes of life. These advances have been facilitated by mathematicians and computer scientists training their analytical armoury on the big biological questions facing us today.

However, simple biological organisms existed millions of years before we humans appeared on the scene, with our calculus and computers. Genetic sequences have been exchanged, copied and corrupted for at least three billion years, according to the fossil record. Biological systems have always processed information in one form or another; only now do we have the tools and techniques available to begin to analyse it. By continually swapping, chopping, splicing and mutating blocks of genetic information, nature has evolved an incredible array of organisms from an initially unpromising primordial sludge. Individual creatures have refined intricate strategies for survival and procreation, from the shifting colours of the chameleon to the peacock’s feather. Over the past few decades, humans have adopted nature’s strategies in a conscious effort to emulate the rich problem-solving capabilities of living systems. Models of natural selection are used to organically ‘grow’ car designs; simulations of the brain recognize patterns in the stock market; artificial ant colonies are used to route mobile phone traffic through congested networks of base stations and exchanges.7 All these solutions are examples of natural processes being successfully abstracted and distilled to yield novel problem-solving methods. This activity is now a central theme in computer science, with major international conferences and learned journals dedicated to the study of ‘biocomputing’, ‘natural computing’ or ‘nature-inspired architectures’. And yet, the flow of information in nature-inspired computing has been, until very recently, one-way traffic. Researchers have dissected natural systems, both literally and metaphorically, in order to identify the key components or processes that can then be harnessed for the purposes of computation. Recently, though, a growing number of scientists have posed the question: ‘Is it possible to directly use existing natural systems as computational devices?’ That is, can we take wet, ‘squishy’, perhaps even livingbiology and use it to build computers and other useful devices? The fields of molecular and cellular computing (and, even more recently, synthetic biology) have emerged in the last decade to investigate further this very question.

Although anticipated as early as the 1950s, the idea that we could somehow build working computers from organic components was merely a theoretical notion, of interest to a tiny and disparate collection of scientists. That all changed in November 1994, when a scientist, better known for helping to build the encryption scheme that safeguards financial transactions on the Internet, announced that he had built the world’s first molecular computer. Emerging from a laboratory in Los Angeles, California, his collection of test tubes, gels and DNA lay at the heart of a totally new and unexplored region of the scientific landscape.

Scientists across the globe have set out to map this terrain in a truly international effort. Millions of dollars are being invested worldwide in molecular computing and synthetic biology research, both by governments and by private corporations. Every month, new results are reported, molecular algorithms developed, exotic organic complexes constructed. DNA, the code of life, is right now being used at the heart of experimental computers. Living cells are being integrated with silicon nanotubes to create hybrid machines, as well as being routinely manipulated to add entirely new capabilities. Brain cells are being used to build real, ‘wet’ neural networks. Preparations are being made to build entirely new organisms, never seen before in nature. The fields of computer science, biology and engineering are constantly morphing and merging to accommodate this radical new enterprise. Traditional boundaries between disciplines are breaking down, as computer scientists move between laptop and laboratory and biologists routinely talk in terms of logic and genetic circuits.

Nobody knows for sure where the journey will take us, and undoubtedly there will be pitfalls along the way – scientific, technological and ethical. What is certain, however, is that a whole new vista is opening up before us, where revolutionary discoveries await. The scenario played out at the start of this book is much closer to fact than fiction. The research I shall describe has the potential to change our lives in profound ways. The genesis machines will change the way we think about not only computers, but about life itself.

1 / The Logic of Life

At the end of 2005, the computer giant IBM and the Lawrence Livermore National Laboratory in the USA announced that they had built the world’s fastest supercomputer. Made up of over 130,000 computer chips wired up into 64 air-cooled cabinets, the machine known as Blue Gene/L cost one hundred million dollars and was capable (at its peak) of performing more than 280 trillion calculations per second.1 Computer scientists salivated at the thought of such vast computational power, forecasters anticipated the creation of global weather models capable of predicting hurricanes weeks in advance, and astrophysicists dreamed of simulating the very first fiery instant after the birth of the universe. Biologists, on the other hand, had other ideas. The problem they had earmarked for the machine was rather more interesting than any of these other projects. They wanted to work out how to unscramble an egg.

What could possibly justify spending hundreds of millions of dollars of American taxpayers’ money on reverse engineering an omelette? The answer lies in just how proteins form their particular complex shapes, and the implications are huge, not just for chefs, but for the whole of mankind. When preparing scrambled eggs, we begin by cracking a couple of eggs into a bowl. What we generally see is the orange-yellow yolk, and its surrounding liquid ‘egg white’. This white (known as albumen) is essentially made up of water and a lot of protein. Individual protein molecules are made up of long amino-acid chains, like beads on a string. The amino-acid ‘beads’ are sticky, so the whole thin string repeatedly folds in, on and around itself when it’s first made, twisting and turning to form a compact spherical ball (proteins can take many wierd and wonderful forms, as we’ll see, but egg-white proteins are generally globular). In their normal state (i.e. inside the egg), these globular proteins float around quite happily in the albumen, bouncing off one another and the various other molecules present. However, when heat is introduced into the equation, things begin to get messy. This new energy begins to shake the egg-white molecules around, and they start to bounce off one another. This constant bashing weakens the sticky bonds holding the protein balls together, and they quickly begin to unfurl back into their original long, stringy shape. With so many molecules bouncing around in solution, the sticky beads begin to stick to their counterparts in other molecules, quickly binding the protein strings together into the dense, rubbery mesh we see on tables the world over.2

Why is this process so interesting to biologists? The reason is that our understanding of protein structure formation is infuriatingly incomplete. We can take a folded protein apart, mapping the precise location of every individual atom, until we have a complete three-dimensional picture of the entire structure. That’s the easy part, and it was done decades ago. Putting it all back together again – well, that’s rather more difficult. As we’ll see, predicting in advance how an arbitrary chain of amino-acid beads will fold up (that is, what precise shape it will adopt) is one of the main driving forces of modern biology. As yet, nobody knows how to do this completely, and the problem of protein structure prediction is taxing some of the best minds in science today. A complete understanding of how to go from bead sequence to 3-D molecule will have massive implications for the treatment of diseases such as cancer and AIDS, as well as yielding fundamental insights into the mechanics of life.

As we can begin to appreciate, nature is often remarkably coy; huge proteins fold up in fractions of a second, and yet the biggest human-built computer on the planet could take over a year’s worth of constant processing just to predict how a single, simple protein might adopt its particular shape. We should not be surprised that simulating even simple natural processes should come at such a high cost, and advances in computing technology and its application to biology will reap huge dividends in terms of a deeper understanding of natural systems. Such knowledge, though, is also beginning to suggest an entirely new way of thinking about how we build computers and other devices. ‘Traditional’ computers are shedding new light on how living systems process information, and that understanding is now itself being used to build entirely new types of information-processing machine. This new form of engineering lies at the heart of what follows.

Nature has computation, compression and contraptions down to a fine art. A honeybee, with a brain one twenty-thousandth the size of our own, can perform complex face recognition that requires state-of-the-art computer systems to automate.3 A human genome sequence may be stored on a single DVD, and yet pretty much every cell in our body contains a copy. Science-fiction authors tell stories of ‘microbots’ – incredibly tiny devices that can roam around under their own power, sensing their environment, talking to one another and destroying intruders. Such devices already exist, but we know them better as bacteria. Of course, the notion of biomimicry – using nature as inspiration for human designs – is nothing new. Velcro, for example, was patented in 1955, but was originally inspired by plant burrs. Spider silk is now used as the basis for bulletproof vests. Away from the realm of materials science, nature-inspired design permeates our modern way of life. Telephone traffic is now routed through the global communications grid using models of how ants communicate using chemical signals. Computer systems based on the operation of the human brain detect fraudulent trading patterns on the stock market. As author Janine Benyus explains, ‘The core idea is that nature, imaginative by necessity, has already solved many of the problems we are grappling with. Animals, plants and microbes are the consummate engineers. They have found what works, what is appropriate, and most important, what lasts here on Earth. After 3.8 billion years of research and development, failures are fossils, and what surrounds us is the secret to survival.’4

Biocomputing – the main focus of this book, building computers not from silicon but from DNA molecules and living cells – has emerged in the last decade as a serious scientific research area. In his book The Selfish Gene5, Richard Dawkins coined the phrase gene machine to describe early life forms in terms of their being nothing more than ‘replication devices’ to propagate their genetic programs. In the title of this book I use the similar phrase ‘genesis machine’, but with exactly the same intention as Dawkins: to emphasize the fact that there are direct parallels between the operation of computers and the gurglings of living ‘stuff’ – molecules, cells and human beings.6 As Dawkins puts it, ‘Genes are master programmers, and they are programming for their lives.’5 Of course, the operation of organic, biological logic is a lot more noisy, messy and complex than the relatively simple and clear-cut execution of computer instructions. Genes are rarely ‘on’ or ‘off’; in reality, they occupy a continuous spectrum of activity. Neither are they arranged like light switches, directly affecting a single, specific component. In fact, as we’ll see, genes are wired together like an electrician’s worst nightmare – turn up a dimmer switch in London, and you could kill the power to several city blocks in Manhattan. So how can we possibly begin to think about building computers from (maybe quite literally!) a can of worms? State of the art electronic computers are unpredictable enough, without introducing the added messiness, ambiguity and randomness that biology brings. As computer scientist Dennis Shasha puts it, ‘It’s hard to imagine how two scientific cultures could be more antagonistic than computer science and biology . . . In their daily work, computer scientists issue commands to meshes of silicon and metal in air-conditioned boxes; biologists feed nutrients to living cells in petri dishes. Computer scientists consider deviations to be errors; biologists consider deviations to be objects of wonder.’7 But, rather than shying away from the complexity of living systems, a new generation of bioengineers are seeking to embrace it – to harness the diversity of behaviour that nature offers, rather than trying to control or eliminate it. By building our own gene(sis) machines (devices that use this astonishing richness of behaviour at their very core) we are ushering in a new era, both in terms of practical devices and applications, and of how we view the very notion of computation – and of life.

If you believe the considerable hype that has surrounded biocomputing in recent years, you could be forgiven for thinking that our desktop PCs are in imminent danger of being usurped by a whole new generation of bio-boxes, thousands of times more powerful than the silicon-based dinosaurs they will replace. This is, of course, absolute nonsense. What concerns us here is not simply the construction of much smaller bioelectronic devices along the lines of what has gone before. We are not just in the business of replacing silicon with organic ‘mush’. Silicon-based machines will, for the forseeable future, be the weapons of choice for scientists probing the fundamental mysteries of nature. Device miniaturisation may well be one of the main side benefits of using molecules such as DNA to compute, but it is certainly not the major driving force behind this work. Instead, researchers in the field of biocomputing are looking to force a fundamental shift in our understanding of computation. In the main, traditional computers will still be important in our everyday lives for the forseeable future. Our electricity bills will still be calculated using silicon-based computers built along existing principles. DNA computers will not do your tax return in double-quick time. Nobody, at least not in the forseeable future, will be able to buy an off-the-shelf organic computer on which to play games or surf the Web.

This may sound like an abruptly negative way to begin a book on biocomputing. Far from it. I believe that alternatives to silicon should be sought if we are to build much smaller computers in the near to mid term. But what really interests me (and what motivated me to write this book) is the long term – by which I mean, not five or ten years down the line, but decades into the future. As Len Adleman, one of the main researchers in the field, told the New York Times in 1997, ‘This is scouting work, but it’s work that is worth pursuing, and some people and resources should be sent out to this frontier to lay a path for what computers could be like in 50 years as opposed to Intel’s explorations for faster chips only a few years down the road.’8

The key phrase here is ‘what computers could be like’. The question being asked is not ‘Can we build much smaller processor chips?’, or ‘How do we run existing computers at a much faster pace’, but what sorts of computers are possible in the future? This isn’t tinkering around the edges, it’s ‘blue-sky’ research – the sort of high-risk work that could change the world, or crash and burn. It’s exhilarating stuff, and it has the potential to change for ever our definition of a ‘computer’. Decades ago, scientists such as John von Neumann and Alan Turing laid the foundations of this field with their contemplation of the links between computation and biology. The fundamental questions that drive our research include the following: Does nature ‘compute’, and, if so, how? What does it mean if we say that a bacterium is ‘doing computation’? How might we exploit or draw inspiration from natural systems in order to suggest entirely new ways of doing computation? Are there potential niches of application where new, organic-based computers could compete with their silicon cousins? How can mankind as a whole benefit from this potentially revolutionary new technology? What are the dangers? Could building computers with living components put us at risk from our own creations? What are the ethical implications of tinkering with nature’s circuits? How do we (and, indeed, should we) reprogramme the logic of life?

I hope that in what follows I can begin to answer at least some of these questions. By tracing the development of traditional computers up to the present day, I shall try to give an idea of how computers have evolved over time. It is important that we are clear on what it means to ‘compute’. Only by understanding what is (and what is not) computable may we fully comprehend the strengths and weaknesses of the devices we have built to do this thing we call ‘computation’. By describing the development of the traditional computer all the way from its roots in ancient times, it will become clear that the notion of computation transcends any physical implementation. Silicon-based or bio-based, it’s all computation. Once we understand this fact – that computation is not just a human-defined construct, but part of the very fabric of our existence – only then can we fully appreciate the computational opportunities offered to us by nature.

Life, the Universe and Everything

Descartes was dying. The once proud mathematician, philosopher, army officer and now tutor to the Queen of Sweden lay in a feverish huddle in his basement room. Racked with pneumonia, his already frail body could no longer bear the intolerable illness, and at four o’clock on the morning of 11 February 1650, he passed away. Barely five months after being summoned to court by Queen Christina, the merciless chill of the Stockholm winter claimed the life of the man who had coined the immortal phrase, ‘I think, therefore I am’. Christina had summoned Descartes for tuition in the methods of philosophy. The Queen was a determined pupil, and Descartes would be regularly woken at 5 a.m. to begin the day’s work. During one such gruelling session, Descartes declared that animals could be considered to be no different from machines. Intrigued by this, the Queen wondered about the converse case; if animals are nothing more than ‘meat machines’, could we equally consider machines to be ‘alive’, with all of the properties and capabilities of living creatures? Could a steam turbine be said to ‘breathe’? Did an adding machine ‘think’? She pointed to a nearby clock and challenged Descartes to explain how it could reproduce. He had no answer.

Thomas Hobbes, the English philosopher most famous for his work Leviathan9, disagreed with the notion of Cartesian duality (body and soul as two separate entities), in that he believed that the universe consisted simply of matter in motion – nothing more, nothing less. Hobbes believed that the idea of an immaterial soul was nonsense, although he did share Descartes’s view that the universe operates with clockwork regularity. In the opening lines of Leviathan, Hobbes gives credence to the view that life and machinery are one and the same:

Nature, the art whereby God has made and governs the world, is by the art of man, as in many other things, so in this also imitated – that it can make an artificial animal. For seeing life is but a motion of limbs, the beginning whereof is in some principal part within, why may we not say that all automata (engines that move themselves by springs and wheels as does a watch) have an artificial life? For what is the heart but a spring, and the nerves but so many strings, and the joints but so many wheels giving motion to the whole body such as was intended by the artificer?

A slim volume entitled What is Life? is often cited by leading biologists as one of the major influences over their choice of career path. Written by the leading physicist Erwin Schrödinger, and published in 1944, What is Life? has inspired countless life scientists. In his cover review of the 1992 edition (combined with two other works),10 physicist Paul Davies observed that

Erwin Schrödinger, iconoclastic physicist, stood at the pivotal point of history when physics was the midwife of the new science of molecular biology. In these little books he set down, clearly and concisely, most of the great conceptual issues that confront the scientist who would attempt to unravel the mysteries of life. This combined volume should be compulsory reading for all students who are seriously concerned with truly deep issues of science.

At the time of the book’s publication, physics was the king of the sciences. Just one year later, the work of theoretical physicists would be harnessed to unleash previously unimaginable devastation on the cities of Hiroshima and Nagasaki. Lord Rutherford, an intellectual giant of the early twentieth century, was once quoted as saying, ‘All science is either physics or stamp collecting’11. Biology was definitely seen as a poor relation of the all-powerful physics, and yet Schrödinger, one of the leading physical scientists of his day, had turned his attention not to atoms or particles but amino acids and proteins. As James Watson, co-discoverer of the structure of DNA, explains: ‘That a great physicist had taken the time to write about biology caught my fancy. In those days, like most people, I believed chemistry and physics to be the “real” sciences, and theoretical physicists were science’s top dogs.’12

Schrödinger’s book had an immediate impact on the young Watson. As he explains, ‘I got hooked on the gene during my third year at the University of Chicago. Until then I had planned to be a naturalist, and looked forward to a career far removed from the urban bustle of Chicago’s south side, where I grew up.’ At the time, Watson couldn’t ever have imagined how different his life’s trajectory would turn out to be. A quiet existence spent studying birds in a rural idyll would be replaced, in time, by worldwide fame, the fanfare of a Nobel Prize ceremony, and perhaps the highest profile of any scientist of his age.

The underlying motivation of Schrödinger’s book, as its title suggests, was to capture the very essence of life itself. Most of us have a common-sense notion of what it means to be alive, and of what constitutes a living thing. We see a tree growing in a field, and can all agree that it stands there, alive. A rock lying in that tree’s shadow, however, is unanimously defined as ‘not alive’. But where do we draw the line between life and non-life? Does there exist a sliding scale of ‘aliveness’, with inert matter such as rocks, dirt and water at the ‘dead’ end, and fir trees, humans and whales at the other? If this scale is balanced, where might we find the fulcrum? What lies just to the right and the left of the tipping point? Are viruses alive? Are computer viruses alive? Fanciful questions on the surface, but they mask a deep and fundamental question. What is life? In order to be able to attach the label ‘living’ to an entity, we must first define the measure by which we come to such a decision. Simply obtaining a definition of life that meets with universal approval has taxed biologists, poets and philosophers since ancient times.

The Greek philosopher Aristotle defined life in terms of the possession of a soul: ‘All natural bodies are organs of the soul,’ he wrote.13 Birds have one, bees have one; even plants have one, according to Aristotle, but only humans have the highest-grade soul. The prevailing belief was that this divine spark was breathed into the egg at the point of conception, thus creating life. This notion remained unchallenged until the rise of thinkers like Descartes, who believed that life could be thought of in terms of clockwork automata (machines). The most famous realisation of this mechanistic interpretation of life was de Vaucanson’s duck, an artificial fowl constructed from copper, which delighted audiences in Paris in the mid 1700s with its ability to quack, splash, eat and drink like a real duck (it even had the ability to defecate).14 Steven Levy describes the disappointment felt by Goethe on encountering the duck, then somewhat aged and forlorn, which speaks tellingly of its ability to give the impression of life: ‘We found Vaucanson’s automata completely paralysed,’ he sighed in his diary. ‘The duck had lost its feathers and, reduced to a skeleton, would still bravely eat its oats but could no longer digest them.’15

Others remained unconvinced. It would take more than a mechanical mallard to persuade them that life could ever exist inside a collection of springs, pulleys and wheels, however complicated or ingenious it might appear. The vitalists took their collective term from the ‘vital force’ or ‘élan vital’ that was, they believed, only possessed by living creatures. Proposed by the French philosopher Henri Bergson, the vital force was an elusive spirit contained only within animate objects.16 The exact nature of the vital force remained the subject of some debate, but many thought it to be electricity, citing the work of the Italian scientist Luigi Galvani. The phenomenon of ‘galvanism’ (and, later, the term ‘galvanized’, meaning to stir suddenly into action) was named after the Bolognese physician, who, while dissecting a frog with a metal scalpel that had picked up a charge, touched a nerve in its leg, causing the dead amphibian’s limb to jerk suddenly. Galvani used the term ‘animal electricity’ (as opposed to magnetism) to describe some sort of electrical ‘fluid’ that carried signals to the muscles (although Galvani himself did not see electricity as a vital force). Nowhere is the claim for electricity to assume the mantle of the ‘élan vital’ made more strongly than in Mary Shelley’s novel Frankenstein, in which the eponymous creator breathes life into the monster by channelling lightning into its assembled body.

In more recent times, scientists have struggled to agree on a definition of life that embraces forms as diverse as the amoeba, the elephant and the redwood pine without resorting to vagueness or spirituality. One framework that has found favour involves ticking off several boxes, each corresponding to a specific criterion. An entity may only be considered as a life form if it meets every one of these conditions at least once during its existence: growth (fairly self-explanatory); metabolism (that is, sustaining oneself by the conversion of energy from one form to another, as we do when we eat and digest food); motion (that is, either moving oneself or having internal motion; trees qualify, even though they are at first glance immobile, as they may twist their leaves and branches to face the sun); reproduction (the ability to create similar yet separate entities, like tadpoles or babies); response to stimuli (the ability to detect and act upon properties of one’s environment, as we do when we snatch our hands away from a hot stove).

These rules, on first inspection, appear to be rigorous yet accommodating. However, the nature of science is to test and probe continuously, and it is not difficult to find counter-examples (that is, things that we would naturally assume to be living and yet fail to meet one or more of the criteria, or a ‘dead’ entity that somehow ticks all of the boxes). For example, no sane person would dream of labelling another as ‘non-living’ due to infertility, and yet the subject of scrutiny would, if we applied the rules strictly, fail the reproductive criterion. Similarly, a bush fire would qualify as a living thing: it certainly grows, it ‘metabolizes’ by converting plant matter into heat and light, it is capable of moving large distances with astonishing speed, it may reproduce by sending cinders into the air that light other fires, and it may respond to environmental conditions (for example, dampness of the ground, or dryness of the undergrowth, by shrinking or expanding).

So, it would appear that we are no further forward in our search for a satisfactory definition of life. However, rather than simply discarding the rules listed above, we may yet find something on which to build, with a little modification. As it turns out, the word ‘modification’ is significant in more than one sense. If we take the reproduction rule and tweak it slightly, therein lies the key to a definition of life. The notion of ‘descent with modification’ is central to Darwin’s theory of natural selection; a parent (or parents) may produce an offspring that is similar to them, and yet also slightly different. These chance variations may endow on the offspring some evolutionary advantage over its competitors (for example, a baby frog may grow up with a slightly stickier tongue than its parents, allowing it to catch more flies than its pondmates), and thus that child and its child will prosper in the future. This study of heritability is known as genetics, and we shall return to it in detail shortly. For now, though, we’ll look at the contributions of two of the founding fathers of modern computing – John von Neumann and Alan Turing. In the 1930s and 40s, these two were instrumental in laying the foundations of biocomputing through their wider work on computation and natural systems.

Artificial Life

Natural organisms are, as a rule, much more complicated and subtle, and therefore much less well understood in detail, than are artificial automata. Nevertheless, some regularities which we observe in the organisation of the former may be quite instructive in our thinking and planning of the latter.17

Three centuries after the death of Descartes, the Hungarian mathematician John von Neumann showed for the first time how a machine could indeed reproduce,18 laying the foundations for a field that was anticipated by Hobbes in 1651 – the field of artificial life. Like Descartes and Hobbes, von Neumann believed that the behaviour of living organisms, although many times more complex, was similar to that of the most intricate machines of the day. He believed that life was based on logic.

The young genius was one of the first people to be appointed to a professorship at the elite Princeton Institute for Advanced Study (IAS), housed in an impressive Georgian-style building in the New Jersey town. Even by the eccentric standards of the massive intellects populating the Institute, von Neumann was considered a little odd. On one famous occasion, he visited the Grand Canyon with his wife, and insisted on taking a mule-train trip right down to the bottom. Everyone in the party was kitted out in the regulation cowboy boots and sombreros – apart from von Neumann. Dressed in shirt and tie, jacket and handkerchief, he cut an impressive dash. Born in Budapest in 1903, von Neumann was a precocious child; at the age of 6 he could divide two eight-digit numbers in his head. On one famous occasion he approached his mother, who was staring blankly into space, and asked her ‘What are you calculating?’ A man described as having a mind that was ‘a perfect instrument whose gears were machined to mesh accurately to within a thousandth of an inch’, von Neumann would make fundamental contributions to science, not only in computing, but in mathematics, quantum physics and economics.

A few years after its foundation, the IAS hosted a two-year visit from a young English mathematician called Alan Turing, who had just made an astonishing debut in the world of mathematics. In 1936, Turing published a paper19 that would lay the foundation of modern computation, as well as prompting an invitation from von Neumann to study at Princeton. Turing spent most of his visit building an electromagnetic cipher machine (a portent of his later work) before returning to England in 1938 after gaining his doctorate. The significance of his contributions – both to computing and to human history – would prove to be profound and long-lasting.

Modern computing, as we understand it today, was born partly out of urgent necessity. The Second World War was raging, and the Allies desperately needed to be able to decode the German miltary’s encrypted communications. By this time, Turing had returned to Bletchley Park in the UK to join the headquarters of the British decryption effort. The cornerstone of the German’s secrecy strategy was the Enigma machine, a portable machine used to encode and decode messages. The key to cracking the Enigma code lay in finding the unique internal configuration of key settings used to encode a particular message. The Germans’ false confidence in the security of their system lay in the astronomical number of possible settings that the machine could have. Even in the worst-case scenario, where the Allies had limited knowledge of the way the machine was set up, there were still more than 1023 possible ways to set up the machine. If a human user attempted to try to crack the code manually using a ‘brute force strategy’ (that is, trying one possible ‘key’ every second), a full-frontal attack on the machine would still take longer than the lifetime of the universe.20

The Turing Machine

The great programming pioneer Edgar Dijkstra once famously remarked that ‘computer science is no more about computers than astronomy is about telescopes’. The point he was trying to make was that computer science is not about the tools or technology; it is all about the scientific acquisition of knowledge about computation. Computation may be defined as ‘The procedure of calculating; determining something by mathematical or logical methods.’ Another definition may be ‘The manipulation of numbers or symbols according to fixed rules.’ Yet another definition may be ‘Finding a solution to a problem from given inputs by means of an algorithm.’ All of these definitions are equally valid – the important thing to note is that none of these definitions even mentions a computer! The reason for this is that computation is an abstract process that may be divorced from its physical implementation. Computation can occur on paper, on an abacus, in a computer, inside a cell, between a bunch of molecules, or even in your head. The foundations of the theory of computation were laid long before electronic computers had been built.

In 1936, at the age of 24, Alan Turing21 published his famous paper which, in the words of his biographer Andrew Hodges, ‘went outside classical mathematics . . . to lay down a new kind of logical engineering.’22 His contribution was to define precisely a device for computing, which we now refer to as, simply, the Turing Machine. This abstract machine provided, for the first time, a framework for the precise definition of an algorithm. Remember, though – computers as we now understand them still hadn’t been built at the time of Turing’s remarkable work, although it placed everything that would follow on a rock-solid theoretical foundation.

Fundamentally, Turing was interested in what it means for something to be computable. A task is said to be computable (other terms that may be used include ‘solvable’ or ‘decidable’) if it is possible to specify a sequence of instructions that, if followed, followed, lead to the completion of the task. This set of instructions constitutes what is known as an effective procedure (or, simply, algorithm) for the task. These notions of computability and effective procedures allow us to make fundamental statements about the extent and limitations of what may be computed.

Turing’s stunning contribution in 1936 was to prove, in an elegant fashion, that there exist certain problems that can never be solved by mechanical means such as a computer. The archetype of such problems is known as the Halting problem, and can be stated informally as ‘Given a computer program and some input data, can we predict in advance whether the program will finish its execution (halt) or run for ever?’ Turing proved that there could not exist a general method to solve the halting problem for all possible combinations of program and input data.

We’ll consider programs in more detail shortly, but for now let’s just say that a program is a list of instructions used to control the operation of a computer. Instructions can do something simple, like multiplying two numbers, or they can control the way the program behaves, such as comparing two numbers and jumping to a different part of the program if the first number is less than the second. Programs are used to control not just desktop computers and game consoles, but cars, traffic lights, washing machines and microwave ovens. A simple program might read in twelve numbers, add them together and then divide the total by twelve to give the average monthly rainfall for a region. Another program might ask the user to input an amount of money and then instruct a cash dispenser to spit out the appropriate banknotes. The first program runs for a fixed number of ‘steps’ (in this case, twelve). The number of steps required by the second program depends on the amount requested and the denominations available to the cash machine. But let’s consider a third type of program: one that runs for ever. The obvious example is a program controlling traffic lights that, at least in England, will cycle through ‘red, red and amber, green, amber’ in an infinite loop until the end of the universe (or until the power is switched off).23 An observer from a foreign country could sit at a junction for a few moments and quickly ascertain a pattern in the cycling behaviour of the lights. However, Turing proved that, in general, we have no way of knowing if the cycle will continue for ever, or if it will eventually come to an end with, perhaps, a previously unseen configuration of lights. If we take a program and some input and it spits out an answer, all well and good. However, if the program takes in an input and then churns away for hours without producing an output, we cannot conclude that the program will never end; maybe we just haven’t waited long enough.

The central core of Turing’s proof was the idea of an idealized, or abstract computer that would become known as the Turing Machine (TM). Turing demonstrated that a universal TM exists that is equal in power to any other possible TM. When we talk about ‘power’, we don’t mean the processor speed or the amount of memory; in computer-science terms, ‘power’ refers to what a machine can and cannot compute. The TM is a way of abstracting away from the details of actually physically building a computer, and instead considering its fundamental processing abilities. More importantly, the TM gives us a precise mathematical framework for the description and analysis of algorithms. An algorithm is a mathematical procedure for performing a computation. To phrase it another way, an algorithm is ‘a computable set of steps to achieve a desired result’.

An automaton (like the TM, or the mechanical duck we met earlier) may be defined as ‘a self-operating machine or robot designed to follow a precise sequence of instructions’. The duck’s complex machinations were dictated by the set-up of its internal configuration of cogs, wires and switches, which defined its behaviour in different situations. The TM, on the other hand, was a machine of the mind – it existed purely as a theoretical construction, although we’ll see, in a later chapter, how such machines may be physically built using biological molecules. A TM may be visualized as a particular type of tape player, operating on a tape of unlimited length (such possibilities being allowed in Turing’s theoretical scheme). The tape is partitioned off into discrete sections or ‘squares’, each of which may contain a bit of information. The TM head may be moved over the tape and, once positioned over any particular square, either read, write or erase the contents of that precise section. The real power of the TM derives from a control mechanism in the device, which tells the tape head what to do when it reads a piece of information. This control mechanism may be represented by what mathematicians call a finite state machine (FSM).

Such a device is used to represent the behaviour of a given system, where that system may be in one of a number of states (the ‘finite’ just means that the number of possible states, although possibly very large, is not infinite). A state represents the current ‘condition’ or ‘status’ of the system (for example, a light switch has two states, ‘on’ and ‘off’, whereas the set of traffic lights we encountered earlier may be in one of four states: ‘red’, ‘amber’, ‘red and amber’ or ‘green’). We assume, for the sake of simplicity, that time moves in discrete ‘ticks’, and during each tick the FSM is in only one possible state. The key thing to note is that, to the FSM, information is divided into two camps: that obtained from its internal state, and that obtained from ‘outside’. The FSM has some sort of abstract ‘sensory’ capability, which allows information to be fed into it from outside (in the example of the TM, input comes from the tape). The structure of the FSM encodes a ‘rule table’, or program, which is used by the machine to decide how to behave, given both its sensory input and its current state. This rule table/program dictates, for any given combination of state and input, both how the machine should behave in terms of its observable actions, and into which state the machine should then move. Author Steven Levy gives an excellent example of an FSM description of the children’s game of ‘musical chairs’, which I quote below.

Here, the world is broken down into obvious time steps, defined by pauses in the music. Players could be in any of four states: sitting, standing, moving and leaving the game. The rules of the musical chairs universe are as follows:

•If one is sitting, and there is no music, remain in that state.

•If one is sitting, and there is music, change to the moving state.

•If one is moving and there is music, remain in that state.

•If one is moving and there is no music, change to the sitting state if there is a chair, and change to the standing state if there is no chair.

•If one is in the standing state, leave the game.

•If one has left the game, remain in that state.

In the game of musical chairs, each participant acts like a finite state machine, noting two things: his or her internal state, and the external condition of the world (i.e. the sensory input) – the music. At each point in the game the players apply this information to determine both how to behave and their subsequent states.

The Turing Machine works on exactly these principles, having a set of internal states and a set of symbols that may be read from or written to the tape. Depending on the current internal state of the machine and the symbol (sensory input) being read from the tape, the machine can do one of six basic operations: it can change its state, write a symbol (or write a blank) to the current square, move the head one square to the left or right, or halt. A rule may be written as ‘When in state 3 and reading symbol 1, write 0 to the current square, move the head one square to the left and go into state 4.’ When the machine comes across a combination of state and symbol for which it has no rule in its table, the TM is said to halt. The TM head starts on the ‘leftmost’ square of the tape, and the tape may already have symbols written in its squares (we call this the ‘input string’). The TM then repeatedly applies the rules in its table, the head zipping backwards and forwards, reading and writing, all under the control of the rule table.

The key thing to note about the TM is that it has unlimited storage capacity on its tape. A Turing Machine with a fixed rule table behaves exactly like a computer with a fixed program, like that controlling a set of traffic lights. The behaviour one can expect from such a system is of limited interest, but Turing’s insight was that the rule table itself could be encoded as a string. In this way, a TM could be set up to read in a rule table (encoding, in essence, a second TM) from a string on its tape, followed by its input string, and then compute the string that the encoded machine would have computed. Such a machine is known as a Universal Turing Machine (UTM), and is capable of simulating the operation of any other computer – that is, anything that can be computed by a ‘real’ computer may be computed by a UTM. Just as you can download emulator programs to allow your PC to ‘pretend’ to be a Commodore 64, an Acorn Electron or a Nintendo Gameboy, the UTM can emulate any other machine.

This startling conclusion means that any program that can be run on a computer, regardless of its make or model, can be translated into a program that will run on a Turing Machine. Any computer program can be converted into a sequence of steps taken from the six basic operations listed above – that is, any piece of computer software, from the simplest program to the most complex package, can be converted into a (long) list of very simple instructions, such as move left or right, write a symbol, or change state. Of course, no sensible person would ever use a Turing Machine for the purposes of practical computing.24 For a start, Turing’s definition states that the machine should have unbounded memory, when this is clearly impossible for a physical machine (you may think that your desktop PC has a lot of memory, but this is not the same as it having unbounded memory). In addition, writing a program for a Turing Machine is a very ‘low-level’ activity that yields an awkward and non-obvious result. It’s rather like describing the route from some landmark to your house by listing every single footstep a person must take along the way. A much more natural way of describing the route (which may be thought of as an algorithm for finding your house) may go something like ‘Start facing the clock tower and turn left. Keep walking until you see a post office on your right. Turn right at the next junction, and my house is the third one on the left.’ This high-level, abstract method of description gives us the ability to describe easily, in a