The atomic human, p.42
The Atomic Human, page 42
Another intervention goes directly to the source of the machine’s power. The emergent digital oligarchy derives its power from aggregation of our personal data. Data trusts are a form of data intermediary designed to return the power to the originators of the data – that is us.17 Sharing our data brings benefits, but also exposes our digital selves. From the use of social media data for targeted advertising to influence us, to the use of genetic data to identify criminals, or to find natural family members. In current data protection law, we are referred to as data subjects. Our data is managed by a data controller who has an obligation to protect our interests. This leads to a power structure reminiscent of the medieval feudal system where the data controller is like the Lord of the Manor, and we are the vassals. Just like in a medieval manor, the overlord has a duty of care for the data subject, but as data subjects we may only discover failings in that duty of care, such as data leakage, when it’s too late.
Personal data trusts18 are inspired by land societies that formed in the nineteenth century to bring democratic representation to the growing middle classes. A land society was a mutual organization where resources were pooled for the common good. A personal data trust is a legal entity where the trustees’ responsibility is entirely to the members of the trust. This ensures that the motivation of the data-controllers is aligned with the data-subjects’ aspirations. How data is handled is subject to the terms under which the trust is convened. The success of any one individual trust would be contingent on it satisfying its members with appropriate balancing of individual privacy with the benefits of data sharing. In 2019 we launched the Data Trusts Initiative,19 funded by the Patrick J. McGovern Foundation, which has sponsored three pilots that explore data trusts in medical contexts and for local communities. We worked with the Ada Lovelace Institute, the Open Data Institute and the Office for AI to characterize the different forms of data intermediaries.20
On the fortieth anniversary of the Moon landings,21 Neil Armstrong said, ‘History is a sequence of random events and unpredictable choices, which is why the future is so difficult to foresee.’ Each of us sees a different piece of these random events, and each of us is faced with different choices. It is only by bringing different voices and perspectives together that we can provide good steerage. In periods of uncertainty, we often grasp for those who confidently give answers, when we should be paying heed to Norbert Wiener’s theory of ignorance and his solution of considering the many different possible paths. That is the spirit of the open society.
The publication date of this book is 6 June 2024, 80 years to the day after Eisenhower launched the invasion of a million men across the English Channel, based on intelligence from that first computer at Bletchley Park. Of those millions, many hundreds of thousands of people died. We owe a great debt to those who lost their lives defending the open society. Not just soldiers taking part in the invasion, but all those who stood against totalitarianism and for the atomic human and our diverse cultures. That debt imposes an obligation on us to hear the voices of all their descendants as we face new challenges and construct a modern information society that we hope would have made them proud.
Acknowledgements
‘Mkono mmoja haulei mwana’ is a Kiswahili proverb meaning ‘one hand cannot bring up a child’. Similarly, one hand cannot write a book.
I am highly indebted to a number of friends and colleagues for their help. Firstly Jonathan Price, as our long conversations in the Fork Deli laid the foundation of my understanding of regulation. Those ideas led to articles in the Guardian, ably subedited by Adam Davidi and Oscar Williams. Their editing transformed my writing from an academic style to something more readable. If I write well it is due to the foundation laid by Adam and Oscar. But the usual provisions apply in terms of inadequecies: errors of style remaining my own, although I also got to tell everyone that’s ‘my voice’.
The Royal Society Machine Learning working group was where I was first able to build an understanding of the wider policy landscape. I am indebted to staff at the Royal Society for convening such a breadth of expertise for that working group. From that period I got to know Claire Craig and Julie Maxton, who generously shared their understanding of how to navigate and communicate with government. Claire and Julie provided my policy foundation in the same way Adam and Oscar provided my writing foundation. Through that group, and later efforts giving advice during the Covid-19 pandemic, I’ve been lucky to work closely with Jessica Montgomery. Her leadership, friendship and support both ground and inspire the projects we lead together.
Many other friends and colleagues have also acted both knowingly and unknowingly as sounding boards for the ideas presented here. To the extent that these ideas have depth and polish it is due to their feedback and encouragement. From an encouragement perspective, I’m particularly grateful to Richard Rex, Sebastian Nowozin, Lucy Cavendish and Rich Turner. They’ve each been generous with expectations about the final result. I hope it comes close to matching them. Inspiration has come from individual conversations with Sarah Haggarty, David Hogg and Hong Ge as well as ongoing conversations with Bernhard Schölkopf. Many ideas came from, or were tested during, conversations on bicycles. Thank you to Alistair Morfey, Rick Cotgreave, Peter Greenhalgh, Steve Marsden, Tony Ryan, Mike Hounslow and Jonathan Tenney for tolerating my ramblings. Francis O’Gorman, John Stringer, Mansur Boase, Andrei Paleyes and Lucia Reisch all kindly reviewed the manuscript for errors.
Kathy Weeks carefully guided me through the first steps of the contractual process, leading me to contact Robert Kirby. His advice and support led me to my wonderful agent Max Edwards. That brings me to the team at Allen Lane. Keith Mansfield has believed in me and the book from the start, his enthusiasm and excitement for the book were coupled with comprehensive editing that lifted the text and the narrative. Thanks also to Sarah Day for copyediting and Rebecca Lee for guiding the book into production and Hanni Sondermann for bringing some sanity to everything.
Joaquin Quiñonero Candela is in the unusual position of being a colleague whose ideas have influenced me and also a key person in the book. I thank him for his generosity in allowing his experiences to be shared, as well as his generosity as a friend and colleague.
Sylvie Delacroix is not only an academic inspiration to me, but she is a generous enough friend to have read very early drafts, provided encouraging feedback, friendly admonishment when I overreached and deep insight on the world that has broadened my perspectives and deeply enriched my thinking.
Matthew Syed has also been extremely generous with his time, thoughts and reading of the book across lunches and discussions around artificial intelligence, sport and the wider world of politics, policy and just life in general.
The final groups to thank are the two different kinds of family: first my academic family, my students and post-docs who are my closest collaborators. You are too numerous to name individually, but Carl Henrik Ek was also kind enough to provide feedback on early manuscripts as well as encouragement, not just for the book, but for life in general. Denis Thérien and Ciira wa Maina, thank you for your wisdom and leadership.
And then my direct family, Marta, Frederick and Paolo. Your patience, not just with the book, but with me, is the foundation on which all of this is laid. Thank you.
Discover Your Next Great Read
Get sneak peeks, book recommendations, and news about your favorite authors.
Tap here to learn more.
Credit: Onur Pinar
Neil D. Lawrence is the DeepMind Professor of Machine Learning at the University of Cambridge where he leads the university-wide initiative on AI and a senior AI fellow at the Alan Turing Institute. Previously, he was director of machine learning at Amazon, deploying solutions for Alexa, Prime Air, and the Amazon supply chain. Cohost of the Talking Machines podcast, he has written a series for the Guardian and appeared regularly on other media.
Praise for
The Atomic Human
“Neil D. Lawrence’s The Atomic Human is a brilliant technological and philosophical tour de force by one of the world’s foremost authorities on AI and machine learning. Anyone interested in the great promise and potential dangers of AI and machine learning would do well to read this book. The Atomic Human is at once a fascinating, entertaining, and deeply serious study on the most consequential emerging technologies humans have ever developed. Lawrence has plenty of computer science laced through the book, but he makes it understandable to the non-specialist by historical examples and analogy. It is also a book of ethics and philosophy that argues we must always ensure machines and AI are viewed and used as tools to assist humans, and we must never concede control of fundamental decisions of great consequence. A great book by an obviously brilliant author.”
—General Mark A. Milley, former chairman, US Joint Chiefs of Staff
“The Atomic Human is a brilliantly panoramic celebration of the vast expanses of human cognition as well as the ingenious, flawed, and often bizarre attempts to replicate it artificially. Refusing easy answers, Lawrence cuts a huge swath across the history of computation with passion, erudition, skepticism, and hope. Cognition, he shows us repeatedly, is not an abstract formula but an impossibly eclectic phenomenon that manifests differently in myriad contexts. From amoebae to the brain to information theory, from Isaac Newton to Alan Turing to ChatGPT, Lawrence shows our approximations of the mind leave out as much as they leave in. He reminds us of the plumbed and unplumbed depths of what is really at stake and the unexpected consequences that will accompany the increased integration of society and technology, the uncontrolled behemoth he calls System Zero. What he demonstrates is more relevant and more urgent than most supposed metrics of AI’s capabilities today.”
—David Auerbach, author of Meganets and Bitewise
“Lawrence is one of the world’s foremost authorities on AI and one of the few who has deployed AI in large-scale industrial systems. He is also a rare technical leader who understands AI as part of a long evolution of human beings interacting with other intelligences in a cognitive landscape. In this thoughtful and engaging book—ranging from James Watt’s steam engines to World War II gunners and the Apollo lunar landings—Lawrence shows what’s novel and what’s human about AI. A must-read for anyone seeking to understand AI’s place in our world and how to harness it for human flourishing.”
—David A. Mindell, Dibner Professor of the History of Engineering and Manufacturing, MIT
“In the wide-ranging intellectual sweep of The Atomic Human, Lawrence invites the public to understand and contrast human and machine intelligence and what AI means for society, effortlessly bridging C. P. Snow’s ‘two cultures’ with lucid, accessible explanations of mathematics and computer science and resonant human and cultural stories from Democritus to Ernest Hemingway.”
—Dr Jean Innes, CEO, Alan Turing Institute
“This is an utterly absorbing account of humans, computers, and how much they differ. It explains why AI cannot substitute for human intelligence even as machine intelligence poses enormous challenges for how information is used and societies are organized.”
—Diane Coyle, Bennett Professor of Public Policy, University of Cambridge
“This is a book for anyone and everyone interested in what makes humans different from machines by one of the world’s experts in AI research. Understanding our differences more may help us live in harmony alongside very intelligent machines, so that we can worry less about existential threats and more about how we work with intelligent machines to make the world a better place.”
—Dame Wendy Hall, Regius Professor of Computer Science, University of Southampton
“The Atomic Human concludes that whatever AI becomes, and whether or not it ultimately poses a threat to our species, it will never replicate or penetrate the essence of what it means to be human.”
—Matthew Syed, Sunday Times
Notes
1. Gods and Robots
1 See e.g. B. Singler (2020), ‘The AI creation meme: a case study of the new visibility of religion’, in Artificial Intelligence Discourse: Religions, 11(253); https://doi.org/10.3390/rel11050253
2 The English title refers to a diving bell, but a more accurate translation of scaphandre would be an old-fashioned diving suit, one of the type that Tintin and Captain Haddock used during their hunt for Red Rackham’s treasure.
3 See United Nations, Department of Economic and Social Affairs, Population Division (2019). World Urbanization Prospects: The 2018 Revision (ST/ESA/SER.A/420). New York: United Nations; https://population.un.org/wup/Publications/Files/WUP2018-Report.pdf
4 See e.g. E. E. Greenwald, L. Baltiansky and O. Feinerman, ‘Individual crop loads provide local control for collective food intake in ant colonies’, eLife, 16 March 2018; https://doi.org/10.7554/eLife.31730
5 See M. Beekman and F. L. W. Ratnieks (2000), ‘Long-range foraging by the honey-bee, Apis mellifera L.’, in Functional Ecology, 14: 490–96; https://doi.org/10.1046/j.1365-2435.2000.00443.x
6 Nick Bostrom, Superintelligence: Paths, Dangers, Strategies (Oxford: Oxford University Press, 2014), p. 22
7 I. J. Good, ‘Speculations concerning the first ultraintelligent machine’, Advances in Computers, vol. 6, 1965. https://doi.org/10.1016/S0065-2458(08)60418-0
8 At the Web Summit conference in Lisbon, Portugal; https://www.cnbc.com/2017/11/06/stephen-hawking-ai-could-be-worst-event-in-civilization.html
9 Arthur R. Miller, The Assault on Privacy (Ann Arbor: The University of Michigan Press, 1971) p. 12
2. Automatons
1 T. H. Flowers, ‘D-Day at Bletchley Park’, Chapter 6 of Colossus: The Secrets of Bletchley Park’s Codebreaking Computers, ed. B, J. Copeland (Oxford: Oxford University Press, 2006), p. 81.
2 Michael Smith, The Debs of Bletchley Park (Aurum Press, 2015), p. 12.
3 Common sense would say that a machine that does nothing is not a machine at all. But machines and mathematics don’t share in our ‘sense of the common’, so defining the nothing machine turns out to be important to make everything work.
3. Intent
1 A. Conan Doyle, A Study in Scarlet (London: Ward Lock & Co., 1888), Chapter 2.
2 Richard E. Susskind, Expert Systems in Law: A Jurisprudential Inquiry (Oxford: Clarendon Press, 1987).
3 G. E. P. Box (Dec. 1976), ‘Science and statistics’, Journal of the American Statistical Association, 71(356), 791–99; https://doi.org/10.2307/22868416
4 This is only partially true. At the time we were talking to Ferrari, racing cars were using their engine’s exhaust to improve their aerodynamic performance in something called a ‘blown diffuser’. But the effect of the blown diffuser could also be re-created in the model without including a full engine.
5 In a Vernam cipher the two streams are compared. If the binary digits are different, then the locked message encodes a 1. If they are the same, the locked message gives a 0. The message then encodes ‘how it was originally different’ from the key. If you have the key, and you know how the original message was different from the key, then you can decode the original message.
6 In the US such machines are known as Rube Goldberg machines after a cartoonist who drew similar machines.
7 See J. Dunn, ‘Introducing FBLearner Flow: Facebook’s AI backbone’, Facebook Engineering Blog, 9 May 2016; https://engineering.fb.com/2016/05/09/core-infra/introducing-fblearner-flow-facebook-s-ai-backbone/
4. Persistence
1 Barley was domesticated around 10,000 years ago in the Levant. See e.g. A. Badr et al. (April 2000), ‘On the origin and domestication history of barley (Hordeum vulgare)’, Molecular Biology and Evolution, 17(4), 499–510; https://doi.org/10.1093/oxfordjournals.molbev.a026330.
2 P. Lee, ‘Learning from Tay’s introduction’, Official Microsoft Blog, 25 March 2016; https://blogs.microsoft.com/blog/2016/03/25/learning-tays-introduction/#sm.00000gjdpwwcfcus11t6oo6dw79gw
3 The quote is from an 1871 essay on strategy by Graf von Moltke. The original German quote can be found in Moltke’s Militärische Werke: II. Die Thätigkeit als Chef des Generalstabes der Armee im Frieden (Berlin: Ernst Siegfried Mittler und Sohn, 1906), p. 291.
4 Each brick is a chemical compound called a nucleobase. The four nucleobases in DNA are called guanine, adenine, cytosine and thymine and their first letters give us their representation as G, A, C and T respectively.
5 This estimate is the current estimate of the mutation rate of the human germline: see e.g. A. Scally, ‘Mutation rates and the evolution of germline structure’, Philosophical Transactions of the Royal Society of London: B, Biological Sciences, 19 July 2016, 371(1699):20150137; doi: 10.1098/rstb.2015.0137. PMID: 27325834; PMCID: PMC4920338.
6 Charles Babbage, On the Economy of Machinery and Manufactures (London: Charles Knight, 1832), Chapter 19, p. 131.
7 Of course, this is just my subjective opinion. To properly rank these two extraordinary achievements, I would have to convert the term ‘extraordinary’ into objective measures. One reason I find it more extraordinary is because it is helpful in communicating the message of this chapter to do so. So the objective measures I would be tempted to create turn out to be subjective in origin.
5. Enlightenment
1 Isaac Newton, Philosophiae Naturalis Principia Mathematica (London: Royal Society, 1687)
2 Isaac Newton Letter to Robert Hooke, on 2 February 1675. Simon Gratz collection, Permanent ID: 9792. Available from https://digitallibrary.hsp.org/index.php/Detail/objects/9792
