How Alan Turing invented the computer, helped win World War II and left us with one of the greatest puzzles of our time: are humans simply computers or are we more than that? Many scientists think we have a tenuous hold on the title, “most intelligent being on the planet”. They think it’s just a matter of time before computers become smarter than us, and then what? This book charts a journey through the science of information, from the origins of language and logic, to the frontiers of modern physics. From Lewis Carroll's logic puzzles, through Alan Turing and his work on Enigma and the imitation game, to John Bell's inequality, and finally the Conway-Kochen ‘Free Will’ Theorem. How do the laws of physics give us our creativity, our rich experience of communication and, especially, our free will? Can a computer win the imitation game and pass the Turing Test? Why do creative people make better mates than rich people? Why are humans bad at mathematics, yet so creative? Could an infinite number of monkeys write Hamlet? Is our brain a quantum computer? Is free will an illusion? James Tagg is an inventor and entrepreneur. A pioneer of touchscreen technology, he has founded several companies, including Truphone, the world’s first global mobile network. He holds numerous patents, filed in over a hundred countries. He studied Physics and Computer Science at Manchester University, Design at Lancaster University and Engineering at Cambridge University. He lives with his family on a farm in Kent, England. www.jamestagg.com HOUSE_OVERSIGHT_015675

“T can't tell you when the last time was that I had this much fun reading and using my brain. From the very beginning, James Tagg had me hooked with the premise; the question of whether or not humans are the most intelligent beings on the planet...” Janet, Netgalley “This is a fantastic book. It seams together cutting edge neuroscience, psychology, thought experiments, artificial intelligence/ ij » machine learning, mathematics and even some history!..’ PF] H., Amazon “Hard work to read, but makes you think about the nature of human intelligence and AI.” Brian Clegg, Popular Science “This is a fat book that covers a huge amount of ground. James’ topic is primarily the brain and how we think, but there is a running theme contrasting the human brain with computers. His thesis is that computers can never think like humans (for example, that they can never be truly creative) and he explores many fields from philosophy and logic to mathematics in pursuit of this proof...” R. Hanbury, Amazon If you have enjoyed reading this book please leave a review and if you would like to hear more, or come to one of my talks, please join the mailing list at: www.jamestagg.com/updates. HOUSE_OVERSIGHT_015676

Are the Androids Dreaming Yet? Amazing Brain. Human Communication, Creativity & Free Will. HOUSE_OVERSIGHT_015677

HOUSE_OVERSIGHT_015678

Are the Androids Dreaming Yet? Amazing Brain. Human Communication, Creativity & Free Will. JAMES TAGG Hurst Farm Books An Imprint of Hurst Farm Enterprises HOUSE_OVERSIGHT_015679

Published by Hurst Farm Books Hurst Farm, Dairy Lane, Crockham Hill, TN8 6RA. +44 1732 807246 12 Williams Road, Chatham, NJ 07928 +1 646 355 1250 www,jamestagg.com [email protected] Copyright @ James Tagg 2015 The Moral Right of the author has been asserted. All rights reserved. Without limitation to copyright, no part of this publication may be reproduced, stored, or transmitted in any form without the prior written permission of the copyright owner and the publisher. A catalogue record for this book is available from the British Library. Publisher’s Cataloging-in-Publication Data Tagg, James, 1964- Are the Androids Dreaming Yet?: Amazing Brain. Human Communication, Creativity & Free Will. / James Tagg. pages cm Includes bibliographical reference and index ISBN: 978-1-910464-03-8 (softcover) ISBN: 978-1-910464-01-4 (ebook) 1. Creative ability. 2. Communication—Social aspects. 3. Technology—Social aspects. 4. Mind and body. 5. Computers and civilization. _ I. Title. T174 .T24 2015 303.48°34—dce23 Library of Congress Control Number: 2014945686 (hardback) p2 220415 poste HOUSE_OVERSIGHT_015680

To my family, who have patiently listened to my interminable ramblings about ‘Elephantine’ Equations. HOUSE_OVERSIGHT_015681

HOUSE_OVERSIGHT_015682

PREFACE ACPMM, Wolfson College, Cambridge ‘A man may have twenty years of experience, or one year of experience twenty times.” Mike Sharman “Rules are for the obedience of fools and the guidance of wise men. Douglas Bader am an inventor. I’ve always been an inventor. Ever since childhood I’ve tinkered with electronics and computers, taking things apart and putting them back together. There is no academic course for inventing, so I had to choose my own path through school and University. I studied design, physics and mathematics at secondary school, and engineering and management at University. Part of that time was spent in the Engineering Department of Cambridge University on a particularly special course. HOUSE_OVERSIGHT_015683

x Are the Androids Dreaming Yet? Mathematical Bridge, Cambridge Every autumn about thirty graduate students arrive at the Engineering Department in Cambridge to join the Advanced Course in Design, Manufacturing and Management. They expect to spend the year walking among the city’s hallowed spires, attending lectures, bumping into Stephen Hawking and punting on the River Cam. Instead, they get quite a shock! In 1989, I joined the course. There were twenty-six engineers, a psychologist and a physicist — me. There was no prescribed syllabus; instead the course used learning-by-experience and lectures from the experts ina given field. To study advertising, you might visit a top London agency, for shipbuilding a shipyard on the Clyde. If you were unlucky enough to find these two lectures scheduled for the same week, you had to travel the length of Britain. The course runs a half dozen minibuses to solve this transport problem. Every four weeks we would undertake a project in a different company. I remember designing pit props for coal mines and imaging software for a weaving company. At the end of each project we presented our findings to each other and, with eight projects and thirty students, this made for a great many presentations. To keep the process manageable, the course put great store in teaching us the art of communication. These days I design large complex systems, and clear communication is extremely important. My ideas are often turned into working products and, if those products have flaws, a post-mortem usually shows the cause HOUSE_OVERSIGHT_015684

Preface xi was a breakdown in communication. Of course, this may be a purely personal failing, but when I talk to people in other companies they report the same problem. It seems we all find communication difficult. have wondered for many years why it is called the ‘art of communication. Surely it’s a science, governed by bits, bytes and bandwidth. That might be true of the symbols in an email — they are clearly encoded symbolically — but is the understanding in our brains simply encoded by symbols? What is the physics that underlies human understanding? Each summer I go on holiday to escape engineering for a couple of weeks. While away I indulge my passion for reading books by the likes of Douglas Hofstadter, David Deutsch and Stephen Hawking. One book that struck me years ago was Roger Penrose’s The Emperors New Mind. In it, he tackles the question of what happens in the human brain when we understand something. He extends an idea put forward by J.R. Lucas of Oxford University that minds must be more powerful than computers because they do something computers cannot: namely to step beyond mere rules and see truth. Colloquially we call this ‘common sense’ or ‘stepping outside the box’ The Lucas argument uses the theories of Gédel and Turing to show computer algorithms have limitations. Some things are simply not computable. Computers can do many useful things, but they cannot discover new mathematical theorems, such as a proof of Fermat’s Last Theorem. In 1996, Andrew Wiles succeeded in finding a solution to this problem. This presents a paradox, solved only if we conclude Andrew Wiles is not a computer. Indeed, since most mathematicians discover at least one theorem during their lives, we must conclude no mathematician is a computer! This is controversial. Most philosophers tend to the view put forward by Daniel Dennett that the Universe is an entirely determined place and any personal sense of free will and creativity is an illusion. In Dennett’s worldview, Andrew Wiles is a special purpose machine that was always destined to solve Fermat’s Last Theorem. I believe this model is flawed. It is my aim in this book to show you why. Indeed I am going to go further and argue all human creativity is non- computational; art, communication, understanding - all are based on non-algorithmic principles. If you consider creative thinking deeply enough you're inevitably drawn into the question of whether we have free will. When I get to work each morning, the first thing I do — after a cup of coffee, obviously — is choose which creative task to tackle first. I feel this choice is freely made, but the determined determinists assure me I am wrong and my HOUSE_OVERSIGHT_015685

xii Are the Androids Dreaming Yet? decision was already made. As Daniel Dennett says, “You have no free will. Get over it!” They say I am effectively an avatar in some giant cosmic computer game, going about my business in an entirely predefined way. I do not agree! If they are right all the coincidences and chance actions of my life were fixed at the time of the Big Bang. I feel this must be wrong, but finding a chink in the determinist armor is hard work; the laws of physics as we know them today are almost exclusively deterministic. This book lays out the options — the chinks — that would allow free will to enter our Universe. To understand human thinking we would really like to look inside a working human brain. We can’t do this yet. All we can do is observe minds at work when they communicate with one another. If our minds think non-computationally — as I believe — we should be able to see them struggle when they have to translate thoughts into symbolic form. The more symbolic, the harder it will be. This is indeed what we observe: face- to-face communication is easy, while formal written modes are much harder. We will explore the difference between human and computer communication as our first step in locating the weakness in the armor of determinism. What do I Believe? As a scientist, I ought not to have beliefs. I should have theories and working assumptions. But, as a human being, I must admit believing certain things are true. Science does not forbid beliefs. It just requires you to be prepared to have one overturned if a better one comes along. Richard Feynman summed this up in a lecture he delivered at Cal Tech: “If you want to discover a theorem,” he said, “first, you guess, then you work out some effect predicted by the theorem. Finally, you see if the effect happens in the real world. If it does, you have a good theory. If the effect happens a little differently, you will need to look for a better theory.’ Here are some of my overturn-able beliefs. HOUSE_OVERSIGHT_015686

Preface X1il Beliefs e We have true free will. We consciously decide our actions and these decisions are in no way predetermined. We shape the future. Allowing for free will is, therefore, a boundary condition for any theory of our Universe. * The world is an amazing place, but understandable. We can understand the Universe through the application of thought and reason. * ‘There is only one Universe and it appears to make sense. * Humans think creatively, computers do not. * ‘The process of understanding and communication is complex, much more complex than the digital theorems of Claude Shannon and Harry Nyquist. « Understanding is hard. « The communication of understanding is even harder. HOUSE_OVERSIGHT_015687

CONTENTS Preface ix Introduction - Experiments, Multimedia and Puzzles 1 Chapter 1 - Mind Over Computer 3 Deep Blue Man v Machine 11 Intelligence 25 The Learning Brain 35 Determinism Al Creative Theories 49 Chapter 2 - Understanding 53 Bad Understanding Can Kill 59 The Imitation Game 65 Chapter 3 - Body Language & Banter 77 Chapter 4 - The Brain 95 Thinking 117 Chapter 5 - Knowledge 127 Chapter 6 - Kittens & Gorillas 147 Chapter 7 - Complexity & Chaos 161 Chaos 171 Chapter 8 - 00 177 Chapter 9 - Known Unknowns 191 The Game of Math 199 HOUSE_OVERSIGHT_015688

Chapter 10 - Turing’s Machine The Machine Chapter 11 - Software Silver Bullets Can't be Fired Consequences Chapter 12 - Hyper-Computing Chapter 13 - Hyper-Communication Chapter 14 - Creativity Chapter 15 - Free Will Schrédinger’s Cat Twins Does God have Free Will? The Free Will Theorem Free Will Universe Chapter 16 - The Quest for Knowledge Awards for Discovery Chapter 17 - The Future Appendix 1 — Acknowledgments Appendix 2 — Bibliography Appendix 3 — Puzzles and Experiments Appendix 4 — Conventions in the Book Appendix 5 — Index of Theorems Index 209 221 229 233 257 273 285 295 313 325 331 339 343 351 355 365 371 374 382 395 397 401 405 HOUSE_OVERSIGHT_015689

“It is no good getting furious if you get stuck. What I do is keep thinking about the problem but work on something else. Sometimes it is years before I see the way forward. In the case of information loss and black holes, it was 29 years.” Stephen Hawking HOUSE_OVERSIGHT_015690

Introduction EXPERIMENTS, MULTIMEDIA AND PUZZLES r Mhroughout this book you will come across experiments to try, multimedia references to track down, and puzzles to solve. You can get additional information at www.jamestagg.com/ understanding. If you undertake an experiment I would appreciate your leaving a note of your results on the website and making useful comments on the blog. Most of the experiments and puzzles are quick and simple. The puzzles I have set often benefit from creative thinking. I have made finding the answers to these problems a little hard, so you are not tempted to cheat. I want you to try to solve the problems and ‘feel’ your brain working. This book argues that intuitive thought solves problems in a different way to analytical thought. The process takes time and often benefits from putting a problem to one side while you use your mind to process foreground tasks. I hope you read this book at a time when the website is not available — or at least don't peek. Give your intuitive thought processes time to work. Graham Wallas described the process of creative thinking in 1926 and I think it is still one of the best models we have: First you must prepare and become fully acquainted with the problem. It might seem impossible but don't despair, just commit to it. Next, you should leave the problem to stew — incubation, he called it. After a while, you will feel a solution is at hand. You don’t quite have it yet but you are HOUSE_OVERSIGHT_015691

2 Are the Androids Dreaming Yet? sure you will. This is intimation. Finally, some inspiration or insight will pop into your head — this is the Eureka moment. Now you have a solution but intuitive thinking is far from infallible. You will need to check the solution and may find your answer wrong the first few times. Persevere; you will get there in the end. As a warm-up exercise, let me give you a simple childhood riddle to solve. A man lives on the twentieth floor of a skyscraper with an old elevator. Each morning he gets into the elevator and goes down to the ground floor, but each evening he gets into the elevator, travels up to the tenth floor, gets out, and walks the rest of the way. Why? ANSWER IN YOUR OWN TIME HOUSE_OVERSIGHT_015692

Chapter 1 MIND OVER COMPUTER Computer versus Human “I visualize a time when we will be to robots what dogs are to humans, and I'm rooting for the machines.” Claude Shannon “The question of whether computers can think is just like the question of whether submarines can swim.” Edgar Dijkstra HOUSE_OVERSIGHT_015693

Kasparov versus Deep Blue “The Three Laws of Robotics: 1. A robot may not injure a human being or, through inaction, allow a human being to come to harm; 2. A robot must obey the orders given it by humanbeings except where such orders would conflict with the First Law; 3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law. The Zeroth Law: A robot may not harm humanity, or, by inaction, allow humanity to come to harm.” Isaac Asimov, I, Robot HOUSE_OVERSIGHT_015694

Deep Blue York, watching a chess match. It’s no ordinary match. Two men sit opposite each other. One, a neatly suited figure, stares intently at the board. You can almost see the heat rising from his head as he processes the possibilities before him. The other, sits implacably calm and, before each turn, looks to a screen at the side of the board, reads the instruction, and makes his move. This is the famous match between Garry Kasparov and IBM’s Deep Blue. Kasparov, a child prodigy, became world chess champion at the age of fifteen and, to this day, holds the record for the highest chess ranking ever achieved. Some consider him one of the most intelligent people on the planet. His opponent, Deep Blue, is a massively parallel chess-playing computer built by IBM’s Watson Research Laboratory. The machine itself sits a few blocks north of the tournament in an air-conditioned room, and relays the moves over a phone line to Joe Hoane, the IBM researcher who moves the pieces. Six months earlier, in Philadelphia, Kasparov won against Deep Blue. This is the rematch and has generated a worldwide media frenzy. Tickets to the event are sold out and most news organizations give a blow-by- blow report each day. On the eighth day of the tournament Kasparov and Deep Blue are level pegging. Kasparov is playing an opening he knows well. It’s one designed to be hard for computers to play and has been tested extensively against Fritz, a chess computer Grand Masters use for practice. But Deep Blue doesn’t seem fazed. Kasparov is visibly tired. On the 16 move he makes a dreadful blunder and sinks into despair. An hour later, after some moments of quiet contemplation, he tips over his ] t is 1997 and we are on the 39" story of the Equitable Center in New HOUSE_OVERSIGHT_015695

6 Are the Androids Dreaming Yet? king, gets up, and leaves the room. Kasparov has resigned, Deep Blue has beaten him 3% to 2% points and is now the most powerful chess player on the planet. Later, when interviewed about his experience, Kasparov thought Deep Blue must have been assisted by humans during the games because the program appeared to play intuitively. The rules of the tournament allowed humans to work on the program between matches, but not during actual play. The argument has never been settled, and Deep Blue was long ago dismantled. These days chess players avoid big public matches against computers, arguing it is really a different sort of game. A computer’s ability to crunch mathematically through all the many possibilities means a chess player must play without error against a machine, but can play a more interesting and fluid match against a fellow human. Chess is computer-friendly because it is a finite problem. You always win, lose or draw. The game can't go on forever because any position that repeats itself more than three times is declared a draw, and if a player makes 50 moves without moving a pawn or taking a piece, the game is also declared a draw. In a typical game, each player makes 40 moves, and on each turn you can choose from 30 possible moves. Although this equates to a huge number of options, it is still a finite number. It is possible, therefore, to create a perfect chess-playing machine. Such a machine would project any position it encountered through every permutation to the endgame. But, although chess is solvable using brute force this might not be practical in our Universe. The storage required to hold all the possible positions being analyzed would be vast — needing most of the atoms in the Universe. You would need to pack this information into a small enough space to allow fast retrieval in order to play the first 40 moves in two hours. This would require storing all the information within a sphere no larger than three light minutes. Putting that much data in such a small space would exceed the Hawking Bekenstein bound - a limit on the information carrying capacity of space-time put forward by Stephen Hawking and Jacob Bekenstein — causing the region of space-time to collapse to a black hole! Despite these minor technical problems, an ingenious algorithm could be made that was unbeatable: chess is essentially computable. The term algorithm will often arise in the book, so it is worth giving a little history. The word comes from the name of an 8" Century Persian mathematician, Al-Khwarizmi, and means a step-by-step procedure. We use one whenever we do long division or look up a phone number on HOUSE_OVERSIGHT_015696

Mind over Computer F The Music of Emily Howell our mobile phone. It is any mechanical procedure you perform without thinking about it. Computers are always executing an algorithm; that’s what they do. Fast forward to 2010 and Centaur Records releases a new classical music CD featuring the piano music of Emily Howell. Critics are enthusiastic about the new talent. She has composed music in a broad range of classical and contemporary styles. You can find some examples on my website. But, it transpires, Emily is a computer, the brainchild of David Cope from the University of Santa Cruz. On hearing this news critics revise their opinion of the compositions - “repetitive and formulaic,” “not real music,” “pastiche” Listen again to the music and see whether you have changed your opinion. Whatever you think, Emily has made a good attempt at composing in the style of several great composers: J.S. Bach and Franz Liszt, as well as modern ones such as Stockhausen and HOUSE_OVERSIGHT_015697

8 Are the Androids Dreaming Yet? Philip Glass. The compositions would get a reasonable technical score in an exam, better than many of my attempts, but are these compositions truly art? There's no question computers are gaining ground on us in certain mathematically oriented tasks — playing chess, musical composition, and various modeling tasks. But attempts to have them work with words and ideas have generally produced dismal results. Until now. In 2008, IBM unveiled Watson: a computer capable of answering general knowledge questions. Watson has an enormous database of human knowledge: the Encyclopedia Britannica, a billion web pages, the entire text of Wikipedia and millions of books. It uses artificial intelligence to trawl through this vast reservoir of knowledge and answer questions using a statistical approach. In 2011, Watson featured as a contestant on Jeopardy, the American quiz show, where it beat the two record-holding contestants — the one with the highest number of wins and the one with most consecutive wins. Let me give you a few sample questions and see how you fare. Question 1. _ It can mean to develop gradually in the mind or to carry during pregnancy. Question 2. William Wilkinson's “An Account of the Principalities of Wallachia and Moldavia” inspired this author's most famous Novel. Question 3. _ Its largest airport is named for a World War II hero; its second largest, for a World War II battle. Watson answered questions one and two correctly but failed on question three. You can probably see the final question is posed in poorly structured English and this threw off Watson’s comprehension algorithm. HOUSE_OVERSIGHT_015698

Mind over Computer 9 IBM’s Watson Plays Jeopardy Ignoring the odd hiccup, Watson is much better at Jeopardy than a human. Should humans be worried? First chess, then music, now general knowledge, will all human endeavors succumb to a computer? What will be our purpose on the planet if this happens? HOUSE_OVERSIGHT_015699

Steve Wozniak “Machines will run the world, humans will become idle pets.” Steve Wozniak HOUSE_OVERSIGHT_015700

Man v Machine re humans advanced computers with a temporary hold on the title, ‘most intelligent being on the planet; or are we fundamentally different? We are extraordinarily creative, but we can't add up as well as a cheap pocket calculator. We have poor memories, but we can use common sense to solve problems we have never seen before. Our communication skills are woefully imprecise, but we can tell jokes that send our fellow humans into paroxysms of laughter. We might conclude humans are not computers, but the scientific consensus is that brains are ‘wet computers. I don’t agree with this and I’m going to set out the argument to show why man is not a computing machine. There is an urban legend we think with only 10% of our brains. This is not true. Science has mapped the vast majority of the human brain using two methods. The first, an amazing set of noninvasive imaging techniques, allows us to ‘see’ the brain as it thinks. The second is more macabre: with seven billion humans on the planet, enough accidents occur through sports injuries, car crashes and surgical mistakes to provide a large enough sample to conduct research. Questioning patients with brain-damage allows us to work out what the injured part did before the accident. One famous patient had an accident where the blade of a toy sword went up his nose and damaged a small part of his amygdala and hippocampus, the area of the brain responsible for storing memory. This rendered the man unable to lay down permanent memories after the accident. Events before the accident remained clear but he could not memorize new information. You could tell a joke and he would find it HOUSE_OVERSIGHT_015701

12 Are the Androids Dreaming Yet? Turning Images to Music funny and laugh uproariously. A few minutes later, you could tell the same joke and he would find it just as funny as the first time. For him, every time was the first time, because he had lost the ability to record long-term memories. The syndrome is wonderfully depicted in the film 50 First Dates starring Adam Sandler and Drew Barrymore. Another patient with specific stroke damage was unable to recall the names of fruits but, oddly, could still name vegetables. Interestingly tomatoes presented a particular problem. He had probably never known how to catalogue them so they were partially remembered in both areas. There are many such medical cases. In Oliver Sachs’ The Man who Mistook his Wife for a Hat, the author relates the tale of a man with visual agnosia who could not reliably name familiar objects, including his own wife! He had a perfectly loving relationship with her but simply could not name her from a picture. Sachs, Professor of Neurology at New York University School of Medicine, provides many such fascinating stories, along with their medical backgrounds. The fruit and vegetable case suggests our brains are organized like a filing cabinet. When we damage a part of the brain, it’s like losing a drawer: All the information stored in that drawer is lost. Quite a few experiments contradict this model and indicate many tasks are distributed around the brain. The curious case of blindsight is one such example. People with a damaged visual cortex can often recognize objects despite reporting they have no sensation of vision. Show them a shape and they will report they can see nothing. Ask them to name the shape and they might even get a little irritated by the question; they are blind after all. But, ask them to guess the shape and they will get it right every time. Seeing is more HOUSE_OVERSIGHT_015702

Mind over Computer 13 Brain Image of Fish Hunting Prey widely distributed in the brain than was first thought. Conscious seeing is based in the visual cortex, but there are older pathways still active in the brain that facilitate this unconscious seeing. The brain is very plastic. Lose your sight through damage to the eye or optic nerve, and the brain can repurpose the visual cortex to other uses such as processing sound or touch. Daniel Kish has developed this to such a high level that he can ride a bicycle despite being blind. He clicks his tongue against the roof of his mouth and uses echolocation to form an auditory model of the world around him. Using a similar approach, Amir Amedi from the Hebrew University of Jerusalem has built an audio imager that turns pictures of the world into musical sound patterns. CAT scans of people using this system show they use the visual cortex to convert these sound images into models of the world in similar parts of the brain to a sighted person. We now know roughly what each part of the brain does, but we have no idea how it does it. The scale of an individual thought is too small to see in a brain scan. All we can do is observe large-scale electrical activity associated with those thoughts. A video, from a group at Tokyo University, shows an example of electrical activity filmed in real time as a fish hunts for its prey. Fish have transparent bodies and thin skulls facilitating this sort of imaging. Humans are much harder subjects to work with! The most popular theory to explain how brains work is as some form of computer. Computers are easy to study because we manufacture them. They tend to crash quite frequently — usually at the most inconvenient HOUSE_OVERSIGHT_015703

14 Are the Androids Dreaming Yet? moments — so we have packed them with diagnostic monitoring systems. These systems allow us to watch a computer think and, since they think symbolically, we can easily read their minds. Unfortunately computers don't display many human-like thoughts. They don’t laugh and cry, they don’t report consciousness and they don't appear to exercise free will or display creative impulses. This is frustrating because these are the thoughts we would most like to study. It might be that computers are not yet powerful enough, and in another few years they will be giving Mozart a run for his money. But there may also be a fundamental difference which renders them incapable of this sort of thinking. This is the crux of the modern scientific debate: do humans think differently? Computer Brains On the face of it, humans and computers behave very differently. Our memories are poor, but we understand things. We are creative, but bad at mathematics. We learn by example, computers are programmed. We are emotional, impulsive and appear to have free will. Computers are ordered, predictable, but lack common sense. Both humans and computers appear to be physical, discrete systems. We both take inputs, generate outputs and are capable of solving similar problems. Indeed, each time we examine a problem solved by humans we usually find we can automate it. This is known as ‘knowledge engineering’ and there are many examples; from aerospace to finance, and architecture to medicine. An example of where computers excel is in medical diagnosis. ISABEL is a clinical diagnosis program designed to help ER doctors quickly diagnose critical patients. It was created by the parents of Isabel Maude, a little girl who presented with multiple symptoms to an ER unit. Doctors were initially confused by the symptoms and misdiagnosed her condition. She was later diagnosed with meningitis. Isabel suffered multiple organ failure but survived. Her parents realized there was something wrong with the ER triage process. They got together with some computer scientists and built the expert system ‘ISABEL. When ER doctors are presented with symptoms, they must mentally scan a vast array of literature to rule in and out possible diagnoses. The problem- solving process is not linear; if you've ever watched the TV series House it gives a great dramatization of the process. Certain symptoms might suggest a diagnosis but are not conclusive, and there are many paths to explore. Programmers have taken the heuristic rules from many doctors and codified them into software. ISABEL allows a doctor to input a set HOUSE_OVERSIGHT_015704

Mind over Computer 15 of symptoms and it will spit out a range of possible alternative diagnoses with probability weightings and suggested further tests. Similar systems are widely deployed in other fields, to build racing cars, design dams and fight crime. Even the game consoles in our living rooms implement artificial intelligence to make the aliens more believable and our hearts pump faster. Origin of Computers Alan Turing effectively invented the modern day computer in a paper he submitted to the London Mathematical Society in the summer of 1936. He was not the first person to come up with the idea — that honor probably goes to Charles Babbage — but he was the first to fully understand its power. When we talk about computers today we mean machines, but it is worth noting computers in Turing’s time were more often humans using pencil and paper. The mechanical computers before Turing were elementary at best. Rudimentary calculating machines were developed in Greece, Persia and China as far back as the Ming Dynasty. An astrolabe recovered froma ship wreck off the Greek Island of Antikythera had cogs and gears and could accurately predict the motions of the sun and planets. Many Babbage’s Difference Engine No. 2, Computer History Museum, CA HOUSE_OVERSIGHT_015705

16 Are the Androids Dreaming Yet? of these skills were lost in the Dark Ages but, once the Renaissance was underway in the 16" and 17" centuries, complex mechanical clocks were devised that were capable of predicting the motions of the planets to a high degree of precision. Mechanical, hand-cranked calculators appeared in the mid-18" century, and in 1886 Charles Babbage conceived the first programmable computing machine, The Analytical Engine. It was designed to read programs from cards, and used cogs and wheels to perform the calculations. His first machine - The Difference Engine — was designed to help the Admiralty calculate tide tables, but Babbage realized he could generalize it to compute almost any function. He ran out of money to complete any of his machines, but in the 20th century a dedicated band of enthusiasts built a working model of Difference Engine No.2. One copy sits in the London Science Museum and another in the Computer History Museum in California. These difference machines are not Turing complete and his Analytical Engine has never been built. 19" Century Calculators HOUSE_OVERSIGHT_015706

Mind over Computer 17 In 1935, Turing was made a Fellow of King’s College, Cambridge, and became interested in whether mathematical proofs could be found automatically. He wanted to know whether solving a mathematical puzzle was simply a matter of working through all the possibilities in a methodical manner, or whether something more subtle was required. Although chess is a fantastically complex game, it is finite, a big enough, fast enough computer can play the perfect game. Is this the case with discovering knowledge? Could a big enough, fast enough computer calculate all the knowledge in the Universe? Is Douglas Adams’ fabled computer Deep Thought a possibility, able to calculate the answer to the ultimate question of ‘life, the Universe and everything, albeit with a more enlightening answer than 42? Turing boiled down the process of pencil and paper computation to a = Model of the Antikythera systematic program — a computer program. Mechanism He proposed a thought experiment where he would run every possible program and see if such a procedure would yield the solution to every imaginable mathematical problem. He was able to show this would lead to a paradox and concluded the universal problem solver could not exist. His discovery is one of the most important of the 20" century — in the same league as relativity and quantum mechanics - and I will use it as my main tool in trying to explain the difference between brains and computers. Although Turing’s original paper was not intended as a blueprint for a practical device, he was one of those rare mathematicians who also liked to tinker with real world machines. The outbreak of the Second World War made the practical application of his work very important, and in Chapter 8 I will relate some of the code breaking stories that were to make him famous and caused Churchill to credit him with shortening the war by two years. Calling Turing’s work an ‘invention’ is probably the wrong term; ‘discovery might be more appropriate. Whatever you call it, people immediately equated human brains with computers. This is not new. HOUSE_OVERSIGHT_015707

18 Are the Androids Dreaming Yet? Each time a new advance in technology is made, people use it to explain the working of the brain. The ancient Greeks thought the brain was a fire consuming oxygen. When Alexander Graham Bell invented the telephone, the nervous system resembled a maze of wires and the brain an exchange. Brains were obviously a sophisticated telephone system. This idea has some potentially frightening consequences, particularly in light of the speed at which computers are improving. The most striking feature of computer technology is the rate of development. Cars travel faster than a person’s legs will carry them, machines manufacture things faster than our hands are capable of working. If brains are computers, surely it is just a matter of time before they will think faster than humans. Turing predicted this would happen when computers reached the level of storing around 10 billion units of information. This happened some time in mid-2000. But today, in the year 2014, I can report that although my computer can beat me at chess, it still cannot fill out my expense report for me. So I am still ahead! Maybe Turing just got the mathematics wrong. The human brain has about 10,000 times more neurons than our most powerful computers have logic gates. By this calculation, it’s not a billion units of storage we need but, a trillion trillion units to put the computer on a par with a human brain. It’s just a matter of time! The worrying thing - especially for fans of the ‘computers taking over the world’ science fiction genre — is that computers are improving exponentially fast in line with Moores Law, and the parity point is coming soon. Gordon Moore founded Intel with Andy Grove, and ran the engineering department there for more than 20 years. According to Moore's Law, the power of a computer doubles approximately every 18 months. The next significant event in the computer versus human competition is the gate count parity point — the moment when the number of logic gates and the number of neurons become equal. By my reckoning this will happen some time in 2053. Don't despair. There may be a few dodges yet. The gate parity point assumes a logic gate and a neuron are equally powerful. However, some single cell organisms with only one neuron are capable of complex behaviors, such as hunting prey and avoiding obstacles. To perform these simple behaviors, a computer would need as many as 10,000 logic gates, about the complexity of my TV remote control. This gives us a bit more breathing space. The extra four orders of magnitude push the gate parity point out to around 2080, too late for me to see, but certainly within the bounds of some readers of this book. HOUSE_OVERSIGHT_015708

Mind over Computer 19 To give you some idea of how Moore's Law works, the graph shows growth in computing power over time; the y-axis is a logarithmic plot using engineering notation. Because the growth is exponential we rapidly end up with very large numbers. Scientists use a special notation to cope with these large and small numbers. In scientific notation a number is written out in a compact form. For example, three hundred can be written as 3.0 x 10’. To expand it back to a regular number you move the decimal point in 3.0 two spots to the right, making the number 300.0. A similar technique is used for small numbers. To expand 3.0 x 10? move the decimal point 2 points to the left, giving 0.03. Why use scientific notation? Well, once the numbers get large they would no longer fit on a page! We can shorten the representation of numbers even further by dropping the ‘3.0 x’ part and just looking at the order of magnitude. The number 10°, one with eighty zeroes after it, is the number of atoms in the Earth, and 10’ the number of particles in the known Universe. 10° meters is the ‘plank number’ believed to be the smallest dimension you can have, and 10!” is called a googol, named by Milton Sirotta, the Moore’s Law is only one example * Exponential Growth of Computing for 110 Years Moore's Law was the fifth, not the first, paradigm to bring exponential growth in computing o fe) ° d eo a @ se ao) < ° 1) @ ep) — @ o. w < 2 ra po > ae oO W Moore’s Law Extended by Ray Kurzweil HOUSE_OVERSIGHT_015709

20 Are the Androids Dreaming Yet? nephew of the famous American mathematician Edward Kasner, and subsequently the inspiration for the name ‘Google; the Internet search engine. Ray Kurzweil, the prolific inventor and futurologist, is fascinated by this exponential growth. Exponential curves grow slowly to start with but they pick up speed rapidly and, in the end, growth tends towards infinity. We are all painfully acquainted with one example of exponential growth: The common cold. Each infected cell in our body releases virus particles into the blood which infect further cells, leading to an exponential increase. This makes us feel rotten. Luckily our immune system can also respond exponentially, albeit somewhat delayed, so we survive. In the case of computer power there is no opposing immune system fighting back, so Kurzweil thinks computers will achieve almost limitless processing power; perhaps even within our lifetime. He thinks this will lead to some interesting consequences, for example, allowing people to live forever! Far-fetched? Follow his argument. The two most important elements in keeping us alive are medical imaging, to see what is wrong; and genetic engineering, to fix those things. Both are improving in line with digital technology, doubling in power every 18 months. As computers get better at seeing into our bodies, and our ability to sequence and synthesize spare parts improves, we will reach a point where we can fix almost any problem. Kurzweil figures technology is improving and his body is decaying at just the right rate to mean by the time he needs heavy duty medical intervention it will be available. Barring a traffic accident or mad-axe-murderer, he should live forever. Even if his calculation is slightly off, the next generation will definitely have this option. You might dismiss this as science fiction, but some amazing things are already happening. Recently a female patient in the USA suffering from bone cancer had her jaw replaced with a 3D printed component. Doctors were able to scan her head and take an image of the good side of her jaw, flip it right to left within the computer and repair any problems they saw. Then they sent the image to a 3D printer. The printer made a new jaw from tungsten powder, which was fused in a kiln. The final stage was to cover the metal part with an inert bone-like substance to give the human body a scaffolding on which to build real bone. They then performed the operation to remove her old jaw and replace it with the new one: result, brand new healthy jaw. There are some practical limits to the power of computers on the horizon. Currently, the wires in a silicon chip are about twenty-two nanometers wide. That’s around a thousandth of the width of a human HOUSE_OVERSIGHT_015710

Mind over Computer 21 hair, or approximately two hundred atoms wide. To match the complexity of a brain we will need to pack an order of ten million more gates into a silicon chip. One way to achieve this is to simply shrink the wires, but when we get down to around ten atoms wide, quantum effects begin to dominate. Signals in today’s chips involve tens of thousands of electrons. We normally think of these electrons as a group, but in these tiny circuits we need to consider the behavior of each individual electron. Problems arise as this behavior is subject to quantum uncertainty. With only ten electrons there is a finite probability that none of them will be where you were expecting them to be. This causes problems for digital logic. You can't put a ‘1’ ina memory location and be sure when you come to read it you will get a ‘1’ back. You have to factor in the possibility of error. Quantum effects can be annoying - requiring us to devise all manner of error checking hardware — but they can also be helpful. Richard Feynman proposed using quantum bits, ‘qubits, to perform computation. Quantum computers can calculate many times faster than a classical computer because a single bit can represent more than one piece of information. Enterprising entrepreneurs are making use of this effect to build the next generation of devices, and you can already buy a 512 qubit computer from a Canadian company called D-Wave. The biggest problem with building more powerful conventional chips is their area is reaching the manufacturing limit for economic viability. Silicon wafers contain random spots of damage and, as a chip gets larger, the chance it will have one of these spots approaches certainty. One solution is to use the third dimension and print the logic HOUSE_OVERSIGHT_015711

22 Are the Androids Dreaming Yet? gates so that they communicate in the vertical direction as well. Intel demonstrated the first three-dimensional chip in 2004, and these chips should begin to appear in our laptops by around 2020. Taking a chip into the third dimension solves the economic problem but adding logic gates to a 3D chip presents a new problem — heat. Heat is generated in proportion to the volume of the chip but it can only be lost through the surface area. Result: the chip overheats. Large animals have the same problem which is why elephants have huge ears, filled with blood vessels, they can flap to cool themselves and really big mammals, such as whales, live in the ocean. The thermal problem is now the biggest problem in most computer designs. One data point suggests we could solve this problem, the human brain. We pack huge processing power into our skulls without overheating by using a variety of techniques, including folding the surface of the brain, running each neuron very slowly and maybe even using quantum mechanics. A very recent discovery is that brains could be using quantum effects to transmit signals. If true — and the research has only been recently published - it means we may use a form of high-temperature superconductivity to avoid overheating. More on this in Chapter 4. Excluding exotic quantum effects, the main difference between computer and human brains is their processing architecture. Brains use slow, asynchronous logic to process information rather than the fast, synchronous type used in modern day computers. Logic gates in today’s computers work all the time, even when there is nothing to do. For example, if I multiply 2 by 3 on my laptop the entire multiply circuit, designed to work on 20 digit numbers will still operate, and, even worse, it will operate on every tick of the master clock even if there is nothing to multiply. The brain, by contrast, works only as it needs; unused gates don't operate. This gives a massive reduction in unnecessary power consumption. We'd like to use this technique in modern computers but itis very difficult to implement. Tiny changes in timing cause completely different operation and this makes them hard to test. We accept this sort of problem in humans, calling it ‘human error’, but we count on computers to behave absolutely reliably, so full-blown asynchronous logic is not likely to appear anytime soon. Some of these ideas, however, have made their way into today’s consumer devices. For example, the chips in the latest iPhone contain two CPUs: a small slow one operating when the phone is dormant in your pocket, and a large fast one that switches on when you need the power for a game or other high performance task. HOUSE_OVERSIGHT_015712

All things considered there doesn’t appear to be a fundamental engineering problem that would prevent computer hardware reaching the complexity of the human brain. But, even with all this hardware, it’s not clear we would have something intelligent. HOUSE_OVERSIGHT_015713

Richard Branson “I was dyslexic, I had no understanding of schoolwork whatsoever. I certainly would have failed IQ tests. And it was one of the reasons I left school when I was 15 years old. And if I - if I'm not interested in something, I don't grasp it.” Richard Branson HOUSE_OVERSIGHT_015714

Intelligence Ithough my computer appears to havea mischievous sense of humor — deleting files and crashing just before important presentations — it does not appear to enjoy listening to Bach, nor does it ever ask me to put it near the window to admire the view. I think this means it’s not intelligent, but what is ‘intelligence’? Intelligence is hard to define; I usually conclude someone is intelligent if I have an interesting conversation with them. In the workplace we typically hire people based on interview, rather than practical tests. Could a computer hold a conversation with me and persuade me it is intelligent? Meet Eliza. Invented by Joseph Weizenbaum of the Massachusetts Institute of Technology, and named after George Bernard Shaw’s character in Pygmalion, Eliza runs an algorithm pretending to be a psychiatrist offering a sympathetic ear to anyone who cares to talk. Try her out at my website or read the sample dialogue below! User: Men are all alike. Eliza: In what way? USER: Men are all alike. ELIZA: In what way ? USER: They’re always bugging us about something or other. ELIZA: Can you think of a specific example ? USER: Well, my boyfriend made me come here. ELIZA: Your boyfriend made you come here ? USER: He says I’m depressed much of the time. ELIZA: | am sorry to hear that you are depressed. USER: It’s true. | am unhappy. HOUSE_OVERSIGHT_015715

26 Are the Androids Dreaming Yet? ELIZA: Do you think coming here will help you not to be unhappy ? USER: | need some help, that much seems certain. ELIZA: What would it mean to you if you got some help ? USER: Perhaps | could learn to get along with my mother. ELIZA: Tell me more about your family. USER: My mother takes care of me. ELIZA: Who else in your family takes care of you ? USER: My father. ELIZA: Your father ? USER: You are like my father in some ways. ELIZA: What resemblence do you see ? USER: You are not very aggressive but | think you don’t want me to notice that. ELIZA: What makes you think | am not very aggressive ? USER: You don’t argue with me. ELIZA: Why do you think | don’t argue with you ? USER: You are afraid of me. ELIZA: Does it please you to believe | am afraid of you ? USER: My father is afraid of everybody. ELIZA: What else comes to your mind when you think of your father ? USER: Bullies. ELIZA: Does that have anything to do with the fact that your boyfriend made you come here ? If you think Eliza a bit odd, you are right. Perhaps you won't hire her just yet. The program looks for keywords in speech, and parrots them back at you interspersed with a few stock phrases. If you type a sentence containing the word “job”, the computer will say, “Tell me more about your job? It’s a simple program and runs to only three pages of text. It shows how good a job you can do with very few resources and how far we are away from a real human being. More complex programs with a larger repertoire have been built and can hold a plausible conversation. People are occasionally taken in by them and are quite shocked when told they have been conversing with a machine. If you’re on your guard, they are easy to spot. So far, no one has built a computer capable of holding an extended human conversation and fooling a human into thinking it’s a person. Incidentally, having a sympathetic listener is so important to human beings these programs are used in psychotherapy and can be as effective as drugs. Turing proposed a test to tell whether a computer had truly achieved human intelligence called the imitation game. His argument is as follows: HOUSE_OVERSIGHT_015716

Mind over Computer DT IQ Test Humans are intelligent. (If you disagree with this premise then you're going to have a problem with this argument!) If you talk to a computer and cannot tell it from a human, it must also be intelligent: QED. The logic is sound but somehow feels wrong. It neatly, but irritatingly, sidesteps the whole problem of defining intelligence. In 1912, William Stern devised a method for measuring intelligence in children. He named it ‘IQ’ from the German Intelligenz-Quotient. You may have taken one of these tests at school. The tests consist of a series of abstract reasoning problems that minimize cultural references. For example, you might be asked to look at a set of blocks with dots on them and identify which is the odd one out. Numerous versions of the test have been developed over the years, but nowadays we mostly use one of three standard tests, Wechsler being the most common. Measuring intelligence is complicated. Culture and language play a big part. If we take a tribe of Amazonian Indians and ask them to list the presidents of the United States, they will fail. That does not mean they’re stupid. Drop me into the Amazon Rainforest and I will probably starve to death; they, on the other hand, can live off the land as hunter-gatherers with only a few hours work per day. Who is more intelligent? One problem with IQ is that individual candidate scores can differ wildly from test to test, sometimes by as much as 20 points. That’s huge. At the high end of the scale it can be the difference between being classified as smart or as a genius; and, at the low end, between being average or mentally subnormal. These variations don’t usually matter and most universities and colleges take IQ with a pinch of salt, preferring more specific tests such as SATs in America, the Baccalaureate in Europe or A levels in the UK. IQ can be very important; and is sometimes a matter of HOUSE_OVERSIGHT_015717

28 Are the Androids Dreaming Yet? life or death. In Atkins v. Virginia, the US Supreme Court found a person with mental disability, defined as having an IQ of less than 80, cannot be executed. IQ is not really a measurement, in the normal sense. Most measurements in life are absolute, for example, distance, weight, and time. I can prove my house is bigger than yours using a tape measure. We each ensure our measures are the same by calibrating them against a common reference. In the 1900s we could have walked down to the local town hall and checked our measurements against a ‘yardstick. As measurements became standardized, these sticks were compared with a common central reference. For example, the metre was a platinum-iridium bar kept at the Pavillon de Breteuil near Paris. In the 1960s, a laser superseded the metal reference, and today a metre is defined as 1,650,763.73 wavelengths of the orange-red emission line in the electromagnetic spectrum of krypton-86 in a vacuum. Measurement has become very precise! Intelligence is different. It has no yardstick. If I were to ask, “How much intelligence does it take to design a building?” there’s no simple answer. IQ is not an absolute measurement - it’s a relative score. Test 100 people and list their scores in order. The ones in the middle get a score of 100; the top 5 a score of at least 130 and the top person a score of 140. Similarly at the lower end. A person with a high IQ is probably smarter than one with a low IQ, but it doesn’t tell you if the building they designed will stand up. It’s rather like quoting the odds of a horse winning the Derby. Quoting the odds does not give the speed of the horse, nor often the winner of the race! Despite attempts by test creators to remove cultural bias, it can never be completely eliminated. Certain Amazonian tribes have no concept of counting above five. For them, numbers are an alien idea and serve no useful purpose in their habitat. In the jungle there are always enough trees to make spears, and as a hunter-gatherer you simply need to know where to find your prey. There is no need to count animals into an enclosure at night. Another interesting environment is the Australian Outback. Aboriginal Australians appear to have a remarkable aptitude for visio-spatial memory and can remember maps or collections of objects much better than you or I. Tests for this skill involve playing a variant of Pelmanism. A collection of objects is placed on a tray and covered with a cloth. The cloth is lifted for 60 seconds to reveal the location and type of objects and then replaced. Subjects are then given a bucket full of objects and asked to recreate the tray. You and I do a modest job. Native Australians do this almost perfectly. Why? HOUSE_OVERSIGHT_015718

Mind over Computer 29 In the vast, inhospitable Outback it is vitally important you remember that water can be found at the two rocks near the old gnarled tree. Forget this and you will die of thirst. It was once thought the skill evolved through natural selection, but this might not be the correct explanation. Recent studies show many of us can use mnemonic tricks to significantly improve our memory. Aboriginal skills might actually be learned and passed on from generation to generation. IQ gives us a way to sum up intelligence using a single number but is this too simplistic? We all have friends who would be our first call if we met that special someone or lost our jobs. They are often not the smartest people we know, but they are highly empathetic. These people have ‘social intelligence. Other friends may fail academic tests yet demonstrate wonderful musical or artistic ability. They have creative intelligence. As we dig deeper, more talents emerge: sporting prowess, organizational brilliance, the ability to inspire loyalty. All these traits appear independently of academic brilliance. During the last century, scientists worked hard to understand these different intelligence traits. The most influential theory came out of studies done at the United States Army Educational testing service, by Raymond Cattell and John Horn, and later added to by John Carroll. Their initials give the theory its name. CHC theory breaks down the general idea of intelligence into many different subgroups: ‘G’ factors. If you are good at recalling all the kings and queens of England in chronological order, or can name every member of the 1966 English World Cup team or, perhaps, all the members of the baseball Hall of Fame, you would have high ‘crystalized intelligence’ — ‘Gc. It measures the sum total of all the things you have learned and retained in your long-term memory, your store of useful, and useless, facts. On the other hand there is innate intelligence, the sort that allows you to solve problems where tapping memory banks is not useful. My family often buy me puzzles for Christmas, the sort where you manipulate bits of bent metal that appear linked, but can be separated with a little ingenuity. These puzzles test our ability to work with problems we have never seen before and is called ‘fluid intelligence’ — ‘Gf? tn * We can go further. A good \ tennis player will have high ‘Gt > and ‘Gv’ scores: ‘t’ for time and ‘Vv for vision, a good pub quiz Metal Puzzle HOUSE_OVERSIGHT_015719

30 Are the Androids Dreaming Yet? contestant a high ‘Glr’ score — ‘Ir denoting for long-term retrieval. Carol Vorderman, a UK game show presenter famous for mental arithmetic, would have a good ‘Gq’ score, ‘q for quantitative numerical skills. With all these types of intelligence to choose from it begs the question, “Is there a single master intelligence from which the rest follow?” Political correctness plays a part here. It feels rather elitist to say smart people are good at everything. It is far nicer to think we each have our individual talents and some just have a few more than others. But that’s not what the science tells us. ‘Group Intelligence’ — the overall G score — does appear to be the underlying cause of the other types of intelligence, and smart people do tend to be good all-rounders. However, there is one major flaw in the analysis; the studies only measure the subjects’ ability to pass academic tests, they don't look at our success in real-life, nor our creativity. Lewis Terman began the longest running study of intelligence and its relationship to life success back in the 1920s. It continues to this day. A group of 1500 children with high IQs were selected and tracked throughout their lives. Terman assumed their high IQs would result in them being very successful. They certainly did well, but studies show they did no better than if they had been chosen randomly from the same area (all the children came from around Stanford University). Famously two children, William Shockley and Luis Alvarez, tested too low to be chosen for the study but went on to win Nobel Prizes for Physics in 1956 and 1968, respectively. There are many similar anecdotes: Apparently stupid people go on to great things. Einstein’s teacher famously stated he would never amount to anything and Sir John Gurdon’s school report said he was ‘too stupid’ for science. He went on to discover monoclonal antibodies for which he was awarded a Nobel Prize! Scientists have now devised the alternative theory of an intelligence tidemark. Once above this level — an IQ of about 130 - you can pretty much do anything you want to. This might be because one very important type of intelligence — creative intelligence — is not highly correlated with the rest. Creative people tend to be sufficiently intelligent for their field but once above that threshold the relationship breaks down. Success in creative endeavors seems to reflect strength of character and creative aptitude rather than raw brainpower. Physical Basis of Intelligence The high correlation between different sorts of academic intelligence suggests we might find a physical process within the brain leading to HOUSE_OVERSIGHT_015720

Mind over Computer 31 high IQ. Functional MRI scans show intelligent people use more neurons when tackling a given mental task, perhaps bringing to bear greater raw horsepower, but this is not really an explanation. It is akin to saying Usain Bolt runs faster because he gets more power to his legs. This is obvious. What we want to know is how. The problem with looking at brains for a common cause is the variation from brain to brain. We all have different genes and life experiences. On top of this, we really only see brains post mortem and this tends to confound comparisons of brain structure. One way to minimize the variation is to use separated identical twins. Twins have identical genes so their fundamental hardware is the same. We should be able to see features of the brain that are common to smart sets of twins but absent in less smart pairs. If a feature is not shared it can be discounted as something accidental, caused by disease, environment, or the like. When we examine smart twins, they appear to have greater myelination of their neurons. Myelin is a flat protein that acts as an insulating sheath, wrapping the nerves and the neurons in our brain. Myelination appears to be part of the mechanism involved in laying down long-term memory — more myelin, more memories. It may also help sustain signals and allow them to move faster over a longer distance: the increased insulation allowing the brain to include information from more distant parts of the brain within a given thought. But increased myelination may be an effect of higher intelligence rather than a cause. The brain is responsible for a significant part of our overall energy consumption so insulating the neurons might simply help with energy conservation. This is an active area of research. Evolution also gives a clue to the causes of intelligence. Humans, nonhuman primates, and dolphins all share spindle neurons. These spread across the brain and appear to help us coordinate complex actions between the different parts. The high function intelligence that characterizes these disparate species requires a great deal of cooperation between different areas of the brain. Take playing a musical instrument. This uses physical coordination (motor cortex), sound processing (auditory cortex), rhythm (another part of the motor cortex), along with emotional interpretation (amygdala). Humans have more spindle cells than other animals so this might explain our superior ability in performing these complex tasks. However plausible these ideas, they are all hardware arguments. It is like me saying my word processor is better than yours because it has gold plated connectors. That might be true — it might allow the machine to run a little faster without electrical errors creeping in, but we all know HOUSE_OVERSIGHT_015721

32 Are the Androids Dreaming Yet? it's software that matters. A great computer game is great because it is cleverly written and has beautiful graphics. The speed of the hardware might help, but it does not define ‘great’ Can we see these software effects in the brain? No, unfortunately, this is where our imaging technologies fail. They lack sufficient resolution. We would need 100,000 times more resolution to see our thoughts, even assuming we would recognize thought if we saw it. There is no reason to believe the brain lays out thinking in anything resembling the computer software we are accustomed to reading. There is one exceptional group of people that does show a software difference on a large-scale — chess players. It seems Chess Masters use a different part of their brain to process information about chess than you and I. This can be clearly seen on scans of the brain and is such a gross effect it even shows up in old-fashioned EEGs - where electrodes are taped to your head. Interestingly the effect can be used to predict greatness. Players likely to become Grand Masters show they use a different part of their brain from the rest of us at an early age. Chess players possess the only large scale wiring difference we know of, but there is another group with a visible physical difference, London taxi drivers. Their hippocampi are noticeably larger than the rest of ours. The hippocampus does many things, but one of its most significant jobs is to memorize maps. The three years it takes to acquire ‘the knowledge’ and the subsequent years of navigating London's complex streets give cabbies a 30% larger hippocampus than the average London resident. Is Intelligence Static? We've all seen the headline. Every summer public examination results come out and every year is pronounced a record breaker! Year after year, students get better and better grades. This creates a problem. There's is no better grade than an A — and eventually all students get As. Welcome to grade inflation — a problem affecting systems the world over, from British ‘A levels to Harvard grade point averages. Newspapers are awash with stories bemoaning the dumbing down of today’s tests. “Examinations aren't what they used to be.” Grade inflation undoubtedly exists and studies of undergraduate grades show progressive compression into the top grades, most competent students get ‘As, making it difficult to distinguish a good student from a great one. HOUSE_OVERSIGHT_015722

Mind over Computer 33 At first glance, the problem appears to be one of social engineering. Teachers don't want to disappoint, and academic institutions want to improve on last year’s results. The people awarding the grades often have a vested interest in those grades improving. Even a tiny positive bias in the most scrupulously honest teacher is enough for grades to creep up. However, grade inflation might not be purely a matter of over enthusiastic teachers. IQ scores are also rising. Welcome to the Flynn Effect. James Flynn, Emeritus Professor of Political Studies at the University of Otago in Dunedin, New Zealand, reported in 1987 that 1Q scores rise over time throughout the world. All told the population gains about one IQ point every three years, and approximately every ten years IQ tests have to be re-calibrated, so the average student once again receives the average grade. This is a mystery. It is a large effect and cannot be explained by the rote learning of lots of sample questions. The human race is either rapidly getting smarter or the least smart members of society are coming up to the general average fast; either way it means there are fewer dumb people around. The Flynn Effect has recently slowed in western countries, suggesting it might be that intelligence is converging rather than increasing overall. Another interesting fact is people become more intelligent as they age, gaining about one IQ point every ten years. Against the stereotype, it’s not all downhill after forty. There is hope for me yet! Until recently we thought IQ was fixed, but new research contradicts this. Muscles get stronger with exercise, physical skills, such as playing golf and tennis, improve with practice; why not intelligence? Scientists used to believe brains couldn't get smarter; you had the IQ you were born with. You might learn more ‘stuff’ during your life, but the G factor stayed the same. It looks like this is wrong and we were simply not using the right exercises. In 2008, Susanne Jaeggi and Martin Buschkuehl, of the University of Maryland, modified an intelligence test into a game and showed playing the game improved ‘fluid’ intelligence and increases IQ. They believe playing their game helps improve working memory -— the short- term memory we use for storing sums as we do mental arithmetic — or remembering telephone numbers. Previous attempts to improve 1Q through practice had not shown much success as the skills did not transfer between tests, but working memory is such a useful thing it appears to help across the board. These factors argue against intelligence being a hardware feature of our brain. It does not remain static but instead improves with age, time, and education. HOUSE_OVERSIGHT_015723

At the beginning of the chapter, I said Garry Kasparov was once thought to be one of the most intelligent people on the planet. When his IQ was eventually tested — the German magazine Der Spiegel put up the money — he scored 135. That means, in academic terms, he is smart but no genius. Yet, he is undoubtedly a genius by any common sense definition: the best chess player to ever live. These days he involves himself in politics rather than chess and is still uniquely able to concentrate for long periods of time. Concentration seems a very important factor. Einstein was once asked where his genius came from. He replied that he did not consider himself a genius but instead put his success down to his persistence and ability to concentrate on a problem for many years. 1Q tests say nothing of our ability to concentrate over extended periods and nothing about our drive to change the world. The tests are, at best, a useful but dangerous diagnostic tool for educators. One of the worst things IQ can do is pigeonhole people. Would Kasparov have become world champion if he had been given his IQ score of 135 as a teenager rather than late in his thirties after he had conquered the world? i Hole-in-the-Wall Experiment “Education is what is left after what has been learnt has been forgotten.” B.F. Skinner HOUSE_OVERSIGHT_015724

The Learning Brain uman beings are born with an extraordinary ability to learn H through experiencing the world around them. Studies show babies as young as three weeks understand musical ideas, smiling as you play music to them in a major key and frowning at music in a minor key. By six months, babies have learned to distinguish the relationship between objects, and by two, they have a command of language and are beginning to develop a theory of self. They understand how to lie and become adept at playing parents off against each other! Sugata Mitra, of Newcastle University, has run an experiment in India to test minimally invasive education called the “Hole in the Wall Project’. As the name suggests, he cut a hole in the wall of a building in Delhi and put a computer in it. The hole opens out onto a slum district and local children rapidly discovered the computer. Without any formal training they picked up the necessary skills and very soon became adept at searching the Web. Remember, in order to ‘pick up’ this skill they often had to learn the English language as well. Another example showing children’s innate ability to learn is Nicolas Negroponte’s ‘One Laptop per Child’ program, which gives computers to children in remote villages around the world. The laptops are a triumph of cost engineering but are fully functional and can connect to the Internet. The inspiration for the project came from an analysis of the economics of the computer industry. Huge capital investment in the western world is driving most costs down, but one cost that seems to have stuck fast is the access device. Laptops tend to remain at a floor price of around $500, far too high for much of the developing world. At $500, a computer store makes $80 when they sell you a laptop. This is as low as is cost-effective HOUSE_OVERSIGHT_015725

36 Are the Androids Dreaming Yet? Laptops Galore for them to stock the machine, employ someone to tell you about it, and fix it if it goes wrong in the first year. Value for money improvements have all focused on faster processors, more memory, sharper displays and larger hard drives, not lower prices. These improvements are useful if you want to shoot aliens, but overkill if you only want to surf the Internet and learn the ‘3 Rs. So the ‘One Laptop per Child’ project has developed a device for $100. Negroponte is often asked how he deals with the maintenance and repair issues. His answer, “There aren't any.’ The computers are treasured possessions and rarely broken or lost. Children become empowered by the machines and can access knowledge and information far beyond the wildest dreams of their parents’ generation. Stories abound of children checking the spot prices for wheat or coffee on the Chicago Stock Exchange, and advising their parents on the price to accept for their crop. Negroponte estimates there are currently 500,000 children in South America teaching their parents to read! It’s interesting to speculate whether children learn spontaneously or are somehow ‘programmed’ by the adult members of society. In both the ‘Hole in the Wall’ experiment and the ‘One Laptop per Child’ program the children could simply be learning from adults and older children, but there is a novel way to eliminate this influence. Negroponte and Mitra have teamed up to run an experiment to see how children learn for themselves. They are planning to air-drop laptops into remote villages in the Andes. In this scenario, the children can't possibly learn from the HOUSE_OVERSIGHT_015726

Mind over Computer 37 Gy Sa bis : One Laptop per Child adults — the adults have never even seen a computer before. Instead, they must rely entirely on their innate learning ability. At this point, the experiment has only just started; I will put details on my website as the experiment progresses. The 10,000 Hour Club Learning by experience takes humans quite a bit of time. Anders Ericsson, Professor of Psychology at Florida State University, studied musicians in the early 1990s and found they had accumulated a huge number of practice hours by the time they became experts. His research was popularized by Malcolm Gladwell, in the book Outliers, and by Daniel Coyle in The Talent Code. The idea is that humans need around ten thousand hours of practice to become proficient at a skill. The more skilled players seem to have a simply accumulated even more Dan McLaughlin HOUSE_OVERSIGHT_015727

38 Are the Androids Dreaming Yet? practice. A number of people have wondered whether you can take this literally, and if you devote 10,000 hours to practicing something you can become world class. Dan McLaughlin from the USA used to be a professional photographer and decided he might like to become a professional golfer. He quit his job and is now 3,500 hours in. So far, he has achieved a 4 handicap. I also personally got bitten by this bug and am learning the piano. I am about 3,000 hours in and am making good progress. Gladwell’s interpretation of Ericsson's results is not without controversy. Ericsson stresses ‘purposeful practice’ is the important element. Practicing the wrong thing for ten thousand hours will just make you good at doing something wrong. Practicing without concentration and attention will equally have little effect. One illustrative example is the story of Edward Sanford, a supreme court Piano Practice judge, who read the morning prayer aloud every day over a 25 year period. After he retired he was asked if he could recite it from memory. Despite reading it as many as 5000 times during his working life, he was unable to remember it. It seems you must purposefully practice the exact thing you want to do if you wish to learn it, in this case recall. Computers don’t require practice to learn a skill. If their program is right they work correctly, and if it is wrong, they are always wrong. Computers can be programmed to learn but so far this learning has been limited to specific problem domains, such as face recognition. They do not have the general-purpose capability humans enjoy. HOUSE_OVERSIGHT_015728

Astrological Clock at Hampton Court Palace “The die is cast” Shakespeare “How does the water of the brain turn into the wine of consciousness?” David Chalmers HOUSE_OVERSIGHT_015729

HOUSE_OVERSIGHT_015730

Determinism have free will. ] Look... I can choose to type any word I like. Giotto... Many philosophers tell me I am deluded. I was always going to type that word and I have no free will. Everything in my life is predetermined. I’m rather like a character in an enormous video game. The character might think it was free to act, and its actions would appear random. Yet from the moment the player clicked the button to start the game, every action the character takes is determined by a preprogrammed set of rules. This is the free will debate. How can we tell we are free? Would there be any observable effect? One of the big problems is that philosophers codified much of our modern theory of free will in the 19" century, at a time when all the known physical laws were deterministic and reversible. They could not see a way for free will to emerge from such physical laws. There was even a group called the Compatibilists lead by David Hume that thought free will could coexist with determinism. Provided you felt free it did not matter that your actions were inevitable. We all want free will to mean actual freedom to make conscious choices. We would like to affect the world in which we live; not the other way around. I dislike making definitions — I find they take away from the core argument and only result in linguistic jousting, but it seems that two centuries of philosophers have avoided a proper discussion of free will by loosely defining the term. Here is my definition: HOUSE_OVERSIGHT_015731

42 Are the Androids Dreaming Yet? “We consciously, and through the exercise of will, make decisions between different choices without anyone or anything causing the decision in advance. Others can influence decisions — by offering advice or even holding a gun to our head, but we choose: If you can devise a better, stronger definition please email me and I will revise my definition to your better one. I’m searching for the most powerful definition of free will — totally free and born out of the exercise of will. The human mind appears to have free will. At least this is my personal conscious experience. Computers, on the other hand, do not. They run programs that dictate exactly how they will operate in every situation. Could a computer be programmed to have free will? That's hard to do. Let’s see why. Thinking with Clockwork Astronomers have been predicting the motions of the heavens for centuries and to do this they need accurate clocks. The very first clocks were sundials. These suffered the obvious disadvantage of not working at night, but it was also unsatisfactory to use the motion of the sun to predict the motion of the sun. The earliest ‘heaven independent’ clocks used water flowing through small holes in pottery vessels. They were effective over short intervals but plagued by dust, dirt and evaporation. It was the invention of the anchor escapement that enabled the first accurate mechanical clocks. By the sixteenth century clockmakers had gone to town developing astrological clocks with more and more gears, to show all manner of information; the phases of the moon, the motions of planets, even the motion of moons orbiting those planets. These clocks became hugely ornate. The astrological clock at Hampton Court Palace was built for Henry VII circa 1542 and, as well as showing phases of the moon and the signs of the zodiac, it accurately calculated the time of high tide at London Bridge, allowing Henry to travel quickly to the Tower of London. You might also notice it shows the sun orbiting the earth! Copernicus published his book, De revolutionibus orbium coelestium (On the Revolutions of the Celestial Spheres) showing the earth orbited the sun a year later in 1543, and it took centuries before it became accepted fact. Clocks need gears. The humble gear is a simple machine. They work because wheels of different size have different circumferences — the distance around the edge - but one full turn is the same for all wheels. Imagine you have a circular sweet such as a Life Saver — or Polo for HOUSE_OVERSIGHT_015732

Mind over Computer 43 British readers — and you roll it once around the wheel of your car. The small sweet will turn many times. Now put a pencil through the hole in the sweet, jack up your car so the wheel is off the ground, hold the sweet next to the wheel of your car and press the accelerator. The sweet will spin round very fast and probably disintegrate in a shower of minty sugar. This is the principle of gearing. A small circle has to do a lot of work to keep up with a big circle. It’s very predictable. The sweet will turn a set number of times for each rotation of the car wheel, equal to the ratio of the circumferences of the two circles. Gears usually have teeth to lock the wheels together, but this is really just to make sure they can’t slip against one another when they transfer huge forces, such as in racing cars. Some passenger cars have been built with smooth gears; a friend of mine had one at university. If he put his foot down too hard, the gears would slip, heat up and you would get a terrific smell of burning rubber. If you were lucky you could leave the car for a few hours and all would be well. But, if not, you had to replace the rubber belt, which was very expensive. Toothed gears generally win out. Toothed gears also have the enormous benefit of being digital. This is quite important if you want to keep things accurate. Gears can’t move a fraction of a tooth so if a toothed gear has ‘slipped’ forward a small amount, it will be kicked back into position when it meshes with another gear. In a modern mechanical clock, a balance wheel swings back and forth on a spring and moves the main gear one notch forward each time it passes its central position. Gears divide this down to move the hour and minute hands. If I put the hands of a clock at midday and let the clock tick 86,400 times, the clock hands will come back to the same place. Once you understand how a clock works you can play a trick. If you tell me the number of ticks the clock has tocked, I can tell you the exact position the hands will be in. To a small child this might be dressed up as a magicians trick — but, of course, it is simply a matter of dividing the number of ticks by 60 and then 60 again to calculate the amount of time elapsed. This type of precisely predictable behavior is called deterministic behavior. Something is deterministic if you can set it up in a particular way and know the exact state later or, conversely, examine something and trace it back into the past. Modern computers scale up clockwork and make it much more efficient; gears are translated into electronic logic gates and a quartz crystal vibrates at 1000 million ticks per second to give us the clock tick. On each tick, the computer can do a mathematical operation, store and retrieve information, or branch down an avenue in its program. Using HOUSE_OVERSIGHT_015733

44 Are the Androids Dreaming Yet? these simple building blocks the computer allows us to play computer games or process the words of this book as I write. Importantly, all these operations are deterministic; given a set of inputs the computer will always generate the same outputs and that means a computer has no free will. “Ah,” I hear you say, “but my computer plays games with me and is not predictable, otherwise I would always beat it” You are right, but the computer has a clever trick to fake non-deterministic behavior: it uses you! Computers on their own cannot generate random numbers. All a computer can do is generate a pseudo-random number and it does this by working its way through a very long calculation. It could, for example, calculate the first thousand digits of 1 (pi), and then start using the subsequent digits as random numbers. The digits look jumbled up but we know they follow an entirely predictable pattern. The computer appears to behave randomly because when I press the button to kill an alien the computer picks the number it had counted up to at that moment, say the 55,678" digit of 1, and uses that. It is I, the human, who unconsciously picks the precise moment in time and therefore provides the random element. My choice is governed by all sorts of extraneous quantum influences: Did I have coffee this morning? Was it a big mug or a small cup? How hot was it? All these things will be important as they determine the amount of caffeine absorbed across the brain blood barrier and the exact timing of my actions. Humans are not good at consciously generating random numbers. We tend to choose the same numbers too often. If I ask you to pick a number between one and ten, you are likely to choose three or seven. This effect is called social stereotyping; magicians often use it when they pretend to read your mind. The problem arises because we tend to over think the problem. I asked you to pick a random number between one and ten. You won't pick one or ten. Five is too obviously the mid-point. Even numbers don’t feel random. Nine is too large. That just leaves three and seven. So the mind reading magician has you! Humans can unlearn this social programming and become quite good random number generators but normally we tend to conform. There is a way two humans can generate a truly random number without training. Find a friend for this experiment. One of you should pick any number between one and ten and start counting under your breath, when you get to ten just go back to one and keep repeating. The other should wait a while and then shout stop. The number reached should be genuinely random. Please post the results on my website and I’ll tell HOUSE_OVERSIGHT_015734

Mind over Computer 45 you if this crowd-sourced random number generator really works. There should be no way to predict the resulting number as both of you are affected by quantum randomness and, provided you wait a little before shouting stop, any social stereotyping should be overcome. If you want to be scientific, remember the random number you started with and the length of time before your friend shouted stop. There should be an improvement in randomness with the amount of time they wait. In the absence of human interaction another way to give a computer access to a random number is from a quantum device. A lava lamp works well! The Lavarand, developed by Silicon Graphics, is a hardware random number generator which uses images of a lava lamp to seed a random number generator. It is covered by U.S. Patent 5,732,138, fave Lamp titled “Method for seeding a pseudo-random number generator with a cryptographic hash of a digitization of a chaotic system.’ Got that! A computer does not acquire free will just through the injection of randomness. You could simply put an intercept on the link from the lava lamp to the computer and completely predict the computer’s behavior. The system as a whole will certainly do unpredictable things, but the computer did not make a choice; behaving randomly is not exercising free will. Where is the will? Consciousness I remember my first trip to Death Valley in the United States. We were driving along the main east-west highway at the bottom of the valley and a sign said, “Turn off your air conditioning now.’ I did as I was told and to cool down I opened the window. When I put my hand out I felt HOUSE_OVERSIGHT_015735

46 Are the Androids Dreaming Yet? nothing; no wind chill, nothing. The air was so hot the wind carried no heat from my hand. When I imagine hot weather it always brings back this memory. It’s my conscious experience of the world. Humans experience the world through a vivid lens we call consciousness. It allows us to think about the world as we watch it and plan actions. But, it also summons associated memories, something scientists call ‘qualia. Most writers describe consciousness as an internal dialogue with themselves and see it as a consequence of human language. That's probably because most writers are linguists. Non-linguists, perhaps even dyslexic engineers like me, experience consciousness as more of a visual dialogue. It’s hard to pin down consciousness as the difference between humans and computers. Computers do have something that resembles consciousness; they have watchdog functions, they plan and anticipate actions and are aware of their own existence. But they don’t understand or make free choices based on this consciousness. It is an entirely mechanistic affair. A computer might know its CPU is overheating and send a notification message to the administrator, but it does not really appreciate what this means. It does not have our sensation of a near death experience. This self-awareness is the ‘hard question of consciousness. Why, despite the computer knowing it is overheating, does this not translate into the intense experience we have? Philosophers, such as Daniel Dennett, think this lack of consciousness is only a matter of time; once computers live long enough and have sufficient internal complexity they will begin to experience the world the way we do. We are nothing special. The problem with consciousness is it does not seem to have any externally discernible effect. Anesthetics can take it away and brain scanners can see that it has been switched off, but what is it for? I think it comes hand in hand with our faculty of creativity. Consciousness allows us to shape the world — not the other way round. HOUSE_OVERSIGHT_015736

“We can't solve problems by using the same kind of thinking we used when we created them.” Albert Einstein HOUSE_OVERSIGHT_015737

HOUSE_OVERSIGHT_015738

Creative Theories nce I have exercised my free will by getting out of bed in the morning, I often decide to do something creative. Humans seem driven to create. We compose music, draw, paint, and solve mathematical puzzles. Computers are not naturally creative; they spend most of their time doing exactly the opposite — following preset rules. Is this a fundamental limitation distinguishing the computational world from the real world? The Conventional View Most scientists believe pattern-matching algorithms in the brain allow us to be creative. To see how this might work, imagine our brains are chaotic — not hard to do — and process many competing ideas at the same time. The neurons in our brains build millions of useful, and useless, connections based on the patterns in the data we see and hear. Then a selection process goes to work — something akin to natural selection — to sift and prune the connections until something bubbles to the surface and we get that, ‘aha’ feeling. Douglas Hofstadter, Professor of Cognitive Science at Indiana University, famous for the book Gédel Escher Bach, has written a computer program using pattern matching to discover number theorems; things like any number ending in a zero is divisible by 5. The program produces interesting results, even perhaps generating some new theorems. He argues the human brain is essentially a scaled up version of his program. By the way, if you like trivia, his book Fluid Concepts & Creative Analogies was the first book ever sold on Amazon.com. HOUSE_OVERSIGHT_015739

50 Are the Androids Dreaming Yet? The Unconventional View Roger Penrose, Professor of Mathematics at Oxford University, holds a completely different view. He thinks brains operate in a non-algorithmic manner and provides a sketch of the possible mechanism in two books - The Emperors New Mind and Shadows of the Mind. He suggests tubulin molecules, which form the skeleton of our neurons, exploit quantum- gravitational effects to calculate non-computable functions. The scientific community was initially highly skeptical that quantum effects could survive the warm, wet environment of biological systems, but in January of 2014, Edward O'Reilly and others at UCL discovered plants use quantum effects to improve the efficiency of photosynthesis. No prize has yet been awarded for this discovery but it must be a contender for a Nobel Prize at some point. Recently Travis Craddock, now of the Nova Institute in Florida, has submitted a paper showing a very similar geometry of proteins exists within tubulin microtubules in the brain. He believes this is evidence quantum effects may exist there as well. A simple quantum effect in the brain could merely reduce the resistance of the wiring in the brain to help conserve power and avoid overheating. We recognize this is a major problem in building small, powerful conventional computers. Roger Penrose suggests an altogether more radical idea. He proposes our brains are quantum gravity computers capable of calculating non-computable functions. We don’t yet have a theory for quantum gravity so his idea is at the cutting edge of physics — read highly controversial. He raises a deep mathematical question. If the Universe is deterministic and effectively equivalent to a computation, how does ‘creative’ knowledge emerge within it? Lots of knowledge can be manufactured by simply mechanically rearranging data. That’s what happens when I watch a DVD or play a computer game, but, at some point in the past, a director or a programmer had to put in the creative effort to make the movie or write the computer program. How did that happen? Was it baked into the fabric of the Universe at the moment of the Big Bang? Is what we take for a Universe really nothing more complex than putting a DVD in the slot and hitting play? One last piece of trivia links Hofstadter with Penrose: Roger Penrose and his father invented the Penrose Steps, inspiring the never- ending staircase in the Escher prints featured in Hofstadter’s book. For movie buffs, the Penrose steps appear in the film Inception, starring Leonardo DiCaprio. The fact we get pleasure from these trivial links tells me something is going on in our brains that is not so mechanical. HOUSE_OVERSIGHT_015740

Mind over Computer 31 M. C. Escher’s Ascending and Descending (Penrose Steps) HOUSE_OVERSIGHT_015741

HOUSE_OVERSIGHT_015742

Chapter 2 UNDERSTANDING ee 5 ; Coalition Satisfaction with ‘ Funding ntial Services. Time to Develop mand mR x Governance \ ‘ ; ‘ * — 5 Economic Services Investment d ‘ | Time to Develop i | Essential Services r Psvehologic; Operations Support for HN| | | Government ee Neutral Populace — Appropriate Mi fort and a” Perce: | \ We era | Individual Competences, oyun Intelligence Judgment, and Ability |, < \ External Material | toExecute |‘ fe Support . ‘| aN * Time to Develop HN’ | Insurgent Acts ¢ Violence sy, { \ / * ‘ é ‘Total Force Host Natie’ ri’ Security Forces i 4/ Insyrgent to Force \ \ = = ey st “Understanding and "Si Pd geupity Forces / a a ~ Knowledge of Social 2 e wo Structures : yf oo eo The Logical Lines of Operations from FM 3-24 Afghanistan COIN Dynamics “Power corrupts, PowerPoint corrupts absolutely.” Ed Tufte “No battle plan survives contact with the enemy.” Colin Powell Originally, Helmuth von Moltke HOUSE_OVERSIGHT_015743

54 Are the Androids Dreaming Yet? ohn Masters stood up to address General Stanley A. McChrystal and his military staff in Kabul. The topic, of course, the war in Afghanistan. The main war lasted only eight weeks, but this did not end the conflict. A level of tribal violence and insurgent warfare rumbled on for years, killing around 30 people a week. Masters’ job was to explain the dynamics of Afghanistan and provide politicians and military commanders a framework to understand what was going on. Think about your country for a moment. What maintains the fabric of society — police, family, the local charity club, church, newspapers, the broadcast media? All these institutions work to keep us civilized, but what happens if a country loses them? There are institutions in Afghanistan, good and bad: tribes, gangs, corrupt officials, families. Masters had spent a year investigating these interactions, and questioning the returning commanders. He and his team believed that understanding the dynamics of the conflict was the key to bringing peace to Afghanistan. If you live in an industrialized country, you rarely see society without its civilizing web in place. One interesting ‘experiment’ that shows what happens when it fails was the 1976 traffic police strike in Finland. Finland is a fantastically law abiding country where most people obey both the written and unwritten laws. During the strike, this behavior changed. Many people began parking illegally but refrained from blocking the roads. A few took advantage of the absence of police to drive incredibly fast — twice the national limit. These would be labeled as ‘defectors’ in game theory. Without traffic police, a different automotive General Stanley A. McChrystal HOUSE_OVERSIGHT_015744

Understanding 55 morality emerged, a different structure to society. Of course all the other parts of society remained the same. People paid their taxes and went about their lives normally; only the traffic behavior was affected. Afghanistan has had most of its social structures removed over the last forty or so years. First the Soviets, and then the Taliban, took apart much of the fabric until finally the Allied Forces swept the Taliban out, leaving very little behind. There were no police or courts, and few laws — or at least none enforced by the rule of law. The Allied Forces have spent a decade rebuilding these structures. Before we examine Masters’ presentation, let’s look at the daily life of an Afghan farmer. If you are an Afghan farmer you have a dilemma. Your most reliable crop is opium. It grows well in the arid soil, does not require irrigation, and is resistant to most pests. For this crop there is a financial infrastructure to rival the Chicago Commodities Exchange. You get interest free loans secured against the crop, and you can forward sell your product on a futures market. Your investors can ‘add value’ by dealing with the major pest — the US military. They do this through the simple expedience of taking pot shots at them if they get too close to the crop. Since a field of opium is worth $30,000 and a militia wage for the year is $350, you can easily employ a few men to protect your investment. Of course, you are indebted to thugs and criminals, but they are at least reliable thugs and criminals. On the other hand, the traditional products of the Himalayas - walnuts, pomegranates and vines — need years to cultivate. There is no forward market and the timescales over which you must take risks are far greater. If you believe your American protectors will leave before the crops mature, you will be loath to plant and care for them. But, if you make the decision to take this risk, you have a strong incentive to foster stability and reap the rewards of your effort. There is a feedback effect: the balance of power between all the different parties is important to the decisions you make, and the decisions you make affect your desire to invest in future stability. Masters’ team built a slide pack to demonstrate the complex interactions between the groups: farmers, security, stability, markets, military power, and emerging institutions. The COIN - COunter INsurgency — dynamics slide shows just how hard it is to communicate complex topics between human beings. The presentation is beautifully crafted but it was a public relations disaster. At the end of the presentation General McChrystal said jokingly, “When we understand that slide, we will have won the war.” The slide was paraded in the press as, “the most complicated PowerPoint slide in history.” HOUSE_OVERSIGHT_015745

56 Are the Androids Dreaming Yet? If you invest a little time on the slide you will understand it and may even see it as a thing of beauty. But Masters’ audience was obviously expecting something different and, presented with this level of complexity, went into shutdown. Perhaps they wanted a simpler presentation, a high-level summary, a few bullet points. Of course, there is no simple presentation on Afghanistan. The lesson is that context, timing and expectation are often as important to good communication as the elegance of the content, and that information is a complex thing. If you want a lighthearted poke at PowerPoint here is Peter Norvig’s PowerPoint version of the Gettysburg Address. Understanding Next time you are in a business meeting, count the number of times the word ‘understand is used. If you ask the people around you what it means you'll stump many of them. That’s because understanding has two very different meanings. Most people don’t separate these meanings but the distinction is important. Understanding means to decode information, to comprehend — but, more importantly, it also means to absorb and internalize information. That feeling you have when you ‘et it’ If I say, “I understand” I mean I have taken in the question you asked and decoded it into ideas so I can provide an answer. This can be quite a mechanical process and computers routinely understand natural language and answer questions — Apple's digital assistant Siri being a case in point. When I say, “I understand a problem” or “understand a culture” I mean something far less tangible. Somehow the information I have gathered over my life is formed into a matrix within my brain that allows me to ponder and run scenarios. I can predict the effects of my actions before I do them, and often anticipate your responses. That’s clearly a very useful evolutionary adaption, but is there more to it than that? Roger Penrose and David Deutsch think understanding allows us to transfer non-symbolic information from one brain to another. We don't run programs in our brains, nor do we store precise information such as lists and tables. We have, therefore, had to evolve a creative approach to communicating skills and understanding each other. One of the most closely studied areas in the field of communication is when it breaks down in the lead up to a disaster. HOUSE_OVERSIGHT_015746

Understanding 57 “The human mind tends to look for clear linear relationships, we like solutions that are close to the problem in time and space and make sense when we think about it quickly, unfortunately, those simple solutions are usually wrong and come from acting on a complex system as if it was a simple one.” Brett Piersen Gettysburg Cemetery Agenda Dedication = Met on battlefield (great) = Dedicate porti eld - fitting! @ Unfinished work (great tasks) Abraham Lincoln Not on Agenda! Review of Key Objectives & Critical Success Factors = Note or remember what we say Organizational Overview Summary @ New nation d to unfinished work Now Nations le h of freedom overnment not perish Gettysburg Addr HOUSE_OVERSIGHT_015747

Space Shuttle Columbia Crew Photo “For a successful technology, reality must take precedence over public relations, for Nature cannot be fooled.” Richard Feynman HOUSE_OVERSIGHT_015748

Bad Understanding Can Kill () n January 16, 2003, at 3:39pm, the Columbia space shuttle took off from Cape Canaveral. During the launch a small piece of foam insulation broke off the fuel tank and hit the shuttlecraft. The event was recorded on a few low-resolution video frames. They show a tiny white object hitting the shuttle and a plume of dusty material splattering outward. The shuttle made it safely into orbit and for two weeks engineers on the ground debated what to do. In the end, it was decided the risk was minimal and the shuttle could safely return to Earth. On reentry, the shuttle disintegrated, killing seven astronauts. NASA managers had decided the shuttle was undamaged based on a series of presentations by the engineers. One image in particular analyzed the potential damage to the shuttle’s tiles from an impact. Read the slide, look at the key frames, and decide for yourself what action you would have : — taken. . ogi WONELg oy RTV Bond Line \ igirain clan Tile) . :- (sIP to i. ponding _. Re etre 0 T) - Shuttle Tile HOUSE_OVERSIGHT_015749

60 Are the Androids Dreaming Yet? Review of Test Data Indicates Conservatism for Tile Penetration @ The existing SOFI on tile test data used to create Crater was reviewed along with STS-87 Southwest Research data — Crater overpredicted penetration of tile coating significantly + Initial penetration to described by normal velocity * Varies with volume/mass of projectile (e.g., 200ft/sec for 3cu. In) + Significant energy is required for the softer SOFI particle to penetrate the relatively hard tile coating * Test results do show that it is possible at sufficient mass and velocity * Conversely, once tile is penetrated SOF! can cause significant damage * Minor variations in total energy (above penetration level) can cause significant tile damage — Flight condition is significantly outside of test database * Volume of ramp is 1920cu in vs 3 cu in for test {~-Seae@ 2/21/03 6 NASA Internal Slide WHAT DO YOU UNDERSTAND FROM THE SLIDE? Some images of the launch are shown on the right Here is what you should have understood from the slide: tiles are really tough but if the foam dislodged from the fuel tank broke through the outer coating it would cause significant damage. The estimated speed of the foam hitting the tile was 640 times greater than anything previously tested. Worried? Is this a proper understanding of the problem? You have the slide and the images. Take another look and think hard. If you want, you can check a video of a similar launch on YouTube to get a feel for the scale of things, but the still frames shown all the information you need to make your conclusion. HOUSE_OVERSIGHT_015750

Understanding 61 Shower of particles below (-Z) of LH wing after debris struck y | wing | Photographs of the Foam Impact from Video Footage Debris forward of LH wing leading edge Frame Showing Foam Dislodging HOUSE_OVERSIGHT_015751

62 Are the Androids Dreaming Yet? Still from Ground Camera LOOK AT THE IMAGES, WHAT HAPPENED? The truth is you simply don't know. If you are puzzling over the strength of tiles, you have been misdirected. There is video footage of some sort of impact on a wing mostly covered in white tiles, and a slide describing the effect of a benign sounding ‘foany hitting those tiles. But what is the evidence for an impact on a tile? The shuttle is certainly not made entirely from tiles; I can see a window in the picture. You should instead be asking more questions, “What happened?” “What hit what?” and “How bad is that?” It was bad. The foam, a very tough material, had hit the leading edge of the wing, a weak point, punching a hole through it. The wing failed on reentry and tore the shuttle apart. Clearly, a full discussion of the possibilities did not occur amongst the shuttle team, or perhaps it only happened amongst the engineers in private. Once the analysis was tidied up and presented to ‘management’ it was a one-way communication of the conclusions, not a discussion of the underlying ambiguous thought process. The result: people passively listened to the information rather HOUSE_OVERSIGHT_015752

Understanding 63 than interactively understanding it and agreed on the recommendation that it was safe to return. Clearly they did not understand the ambiguity otherwise they would have realized they did not have enough information to form a conclusion. This is the tragedy of lack of understanding. If they had known how little they knew, they could have deployed a spy satellite to take pictures of the damage - one was available nearby and would have taken a few hours to re-task — but they did not. Ed Tufte served on the second shuttle disaster commission and provided an analysis of the disaster. He views slides as a poor medium for communicating complex problems and thinks documents are far better. The danger with slides is they force you to simplify information in a way that destroys the essence of the information. His analysis of the failure of communication at NASA formed a major part of the final report on the disaster. Later he coined the paraphrase “All Power corrupts; PowerPoint corrupts absolutely” Good communication benefits from stories and narrative, not bullet points and graphic fluff. Instead of using bullet points, speak! After all, we have evolved for 250,000 years to understand language, but only 25 to read PowerPoints. ’ If you write presentations, Ed Tufte’s book The Cognitive Style of PowerPoint is compulsory reading. He argues that much of the information you want to communicate is complex and interconnected. PowerPoint or any similar presentation software encourages you to simplify it into hierarchical bullets. The format implies simple causal relationships where none exists. This is dangerous. Communication should convey understanding — which is very important — and not just information. What, you ask, is the difference? HOUSE_OVERSIGHT_015753

/ ) > i m™ “The hardest thing to understand is why we can understand anything at all” Searle’s Chinese Room Albert Einstein HOUSE_OVERSIGHT_015754

The Imitation Game s an experiment, I am going to ask a student to spend a week in A: locked room. The room is perfectly nice; it has a bed, a light, a desk, some reading matter, oh, and we'll give him some washing facilities too! Every now and then I post some food under the door to keep him going, Pop-tarts and pizza (thin-crust) work well. On the first evening a note is pushed under his door with a symbol on it. The student puzzles for a while, then opens the book sitting on the desk. The book says, “If you get a piece of paper with symbols on it look them up and follow the instructions.” He looks up the symbols and the entry in the book says, “Go to page 44, write down the third symbol on a piece of paper then post it back under the door.” He follows the instruction and is rewarded with another piece of paper, this time with a larger set of symbols on it. Again he follows the instructions in the book and posts his answer back under the door. This goes on for several days. He is somewhat bemused, but it passes the time, and he diligently looks up the symbols and performs all the complicated actions as instructed. Meanwhile, I meet our new Chinese graduate student and explain to her she needs to interview a potential translator for the department. He has just come in from Hong Kong and there is a health scare, so we have quarantined him in the lab room. He is bored and I have some paper for writing messages. She writes “hello” in Chinese on a piece of paper and posts it under the door. HOUSE_OVERSIGHT_015755

66 Are the Androids Dreaming Yet? The exchange of notes goes on for a few days and the two seem to be getting on well. There is even a little romance in the air. When the week is over I open the door and the two meet. The graduate student says, “Hello. It’s nice to finally meet you in person.’ The man is puzzled because, of course, she has spoken to him in Chinese. He knows no Chinese. “Tm terribly sorry, but I don’t speak Chinese,” he says. She is puzzled, “But I spoke with you this last week!” “No, I really don't speak it,” he says. And, of course, he is telling the truth. The book he has been using contains the rules for answering questions in Chinese, but he has absolutely no knowledge of the language. I'll leave to your imagination whether the two strike up a real relationship and live happily ever after. This is the Story of the Chinese Room. The setup is able to fool someone into believing there is a Chinese speaking person in the room, yet there is not. Where does the understanding of Chinese lie? The man definitely does not understand Chinese. And the book clearly does not understand Chinese because it is an inanimate object. Yet the person outside the room is convinced she is communicating with a Chinese speaker. The analogy to a computer is clear. The book is software and the man blindly following instructions is the hardware. John Searle, who devised the thought experiment uses it to show computers can never understand because there is no place in a mechanistic system for understanding to exist. The Chinese Room has sparked huge argument in philosophical circles; let me boil it down to its simplest form. First, let's refute Searle's position with the ‘System Argument. The man plus the book form a system. Systems understand; their individual components do not. My blood does not understand. My brain without blood would not understand — it would be dead! Plug my brain into a good supply of blood; add a dash of glucose, and it will understand the most complex of things. The systems argument is elegant and most scientists think this is the definitive argument against Searle, but Searle has a neat way to counter it. “Imagine”, he says, “that the man memorizes the book and leaves the room. Now there is no system, there is just the man, but the man still does not understand Chinese; he is just parroting rote-memorized words and rules.” Computers, Searle argues, process syntax — the rules of language; humans understand semantics — the contextual meaning of language. Artificial Intelligence (AI) proponents hate the Searle argument. They believe the memorization of a set of words and rules is exactly what gives us knowledge of Chinese. That is why we go to school! HOUSE_OVERSIGHT_015756

Understanding 67 A key problem posed by Searle's Chinese Room is whether you can know everything about a situation from just looking at the inputs and outputs. This is very similar to the restriction posed by the Turing Test. In that case if we were to trace the wire from our computer terminal to the other room we would either find a human typing messages or a large box covered in flashing lights. This would definitively answer the question whether we were talking to a man or a machine. Similarly, if we opened the door to the Chinese Room we would immediately know whether there was a real Chinese speaker in there or not. But opening the door on both tests misses the point. The question asks, “if the inputs and outputs are the same does it matter what is really going on inside a closed system?” Black Boxes Experiments involving closed systems are known as Black Box experiments. They presume you can learn everything about the inner workings of a box simply by probing it from the outside. Young electronic engineers are often given black boxes as a test. Electronic components hidden in the box are connected to three external terminals on the outside. The student is asked to deduce what is in the box using only an electric meter to probe those terminals. Here are a few examples of the possible contents of a black box. They would all show up identically on the student’s meter. Although internally different they are externally identical. Even my ‘silly fourth choice with a cat in the box does not give 170 Black Box Equivalence HOUSE_OVERSIGHT_015757

68 Are the Androids Dreaming Yet? itself away if all you have to go on are electrical readings. (I dare say the cat would make its displeasure know if left in there for any time.) The contents are, therefore, said to be black box equivalent. The reason for teaching engineers about black boxes is to help them understand how to simplify things. We could construct option four, with a cat and some food, but it would cost a great deal of money. Option 1 is functionally identical from an electrical point of view, but for a fraction of the cost. Steve Wozniak and Steve Jobs were so successful when they started Apple because Wozniak was brilliant at simplifying logic circuits. He could take a design with thirty chips and come back with a black box equivalent solution using only five. It was a fraction of the cost and far more reliable. Scientists put great store in black box equivalence because of a principle called Occam’s Razor. William of Occam was an English Franciscan friar living in the fourteenth century. He proposed the idea of minimal explanation. It states that, ‘among competing hypotheses, the hypothesis with the fewest assumptions should be selected? When trying to explain the workings of a black box, the more complicated inner workings should be discarded, as they have no externally verifiable effect over the simpler mechanism. Our extraneous animal must be eliminated! Sorry. Ironically, given his calling, Occam’s Razor is sometimes wheeled out as a disproof of the existence of God. Surely God is a complication unnecessary to the explanation of our Universe. The argument is illustrated beautifully in Carl Sagan’s book Contact and the film of the same name. God gets the last laugh in Sagan’s book when the difficulty with Occam’s Razor is brought into sharp focus. Occam's Razor contains an inherent paradox. At any moment in time we only have evidence to support the simplest of explanations, yet we know many of these simple explanations are incomplete. We regularly discover new phenomenon — dark matter and dark energy being some recent examples. If we stopped discovering new things, Occam's Razor would be a good way to simplify our thoughts. Occam’s Razor is a useful intellectual tool to prevent us over complicating explanations, but there will often be explanations that are correct, but for which there is not yet any observed effect. If we go back to our black box example, we see the flaw in concluding the boxes are identical from examining only their inputs and outputs. Opening them would clearly show they are not identical! But, how would this fact reveal itself if they remain closed? The answer is: over time. If something in the box has memory or understanding, it could present one set of results for a while and a completely different set of results later. HOUSE_OVERSIGHT_015758

Understanding 69 In my trivial example, the cat could eat a wire and change the operation of the black box. Now there is an open circuit where none existed before. If this happened, the output would change and we would need a new theory to explain it. If the circuit was attached to a missile control system or a life support system, you would really want a full understanding without waiting. It’s humans nature to try to open black boxes. This is what MRI scans, X-rays, particle accelerators and all our other tools of scientific investigation are for. We want to open all the black boxes of nature and see what is going on inside: simply waiting to see what happens is not acceptable. In a sense, we live in a black box. We experience the world through our senses, seeing with our eyes and feeling with our hands. The brain never directly experiences anything; it only infers the likelihood of ae et ~ Scene from The Miracle Worker. Helen Keller pictured at the moment she understood language. HOUSE_OVERSIGHT_015759

70 Are the Androids Dreaming Yet? something from the signals it receives. This is similar to our engineer probing the terminals of the circuit of a black box. How can we know our experience of the world is real? Understanding the World The French philosopher Descartes gave us an explanation for this paradox. He spent a long time looking skeptically at everything we perceive. For example, when we poke a stick into a pond, the surface of the water bends light and the stick appears to have a kink in it. Our eyes tell us the stick is bent, but our brain ‘knows’ the stick is straight: it’s an illusion. Descartes wondered if something so simple could be an illusion, perhaps the whole of our experience is too. His eventual solution underpins much of modern philosophy - ‘T think therefore I am, cogito ergo sum. Even if we doubt everything else, we cannot doubt we are thinking about this doubt. At least we can rely upon the existence of this ‘thought’ as some reality. Descartes built up from this bedrock the real world we live in. We can be sure we experience things and can apply logic and use thought. We can use this intellectual faculty to tell a great deal about our Universe. True Understanding In the QED lecture series, The Strange Thing about Light and Matter, Richard Feynman relates the story of the ancient Mayan astronomers. 3000 years ago they were able to predict the motion of Venus in the sky using only pebbles. They had a simple system that could predict when the planet would rise over the horizon. Put a stone in the jar every day, take out a stone once a week, add a stone at every new moon. If the number of stones in the jar is divisible by 23, Venus will rise. ’m making up the details but you see the idea... It’s a very simple algorithm. What should we conclude if the Mayans had perfected their calculations to predict the motion of Venus and it proved reliable over a whole century? Would this constitute understanding? Feynman would say no: the Mayan understanding was not complete. It was only black box equivalent to our modern understanding over a limited period. We known that once the Sun begins to run out of fuel it will swell to a red giant and explode, destroying Venus and the Earth. Their model could not predict this catastrophic failure. Our modern deeper understanding of the workings of the solar system allows HOUSE_OVERSIGHT_015760

Understanding 71 us to predict this future even though there is no clue from the motion of Venus today. Understanding allows us to predict discontinuous events: a system changing its state or a star running out of fuel. We see the same predicament in stock markets. Stock markets normally behave in a linear fashion but, when they go wrong; they go very wrong. Recent recessions have been made much worse by the failure of hedging systems to handle market disruption. Some even think the crises were caused by the automatic trading strategies of these hedging systems. The quants - as mathematicians in banks are called — spend considerable effort modeling financial instruments to show that if one stock goes down, another will go up at the same time. If the stocks are held together your investment is safe because, on average they will remain constant. The problem with these correlations, which often hold reliably for many years, is that when trouble hits they fall apart. Historical correlations don't give us understanding of the future: something that was only meant to happen once in a million years has happened within six months. As they say on your investment papers, past performance is no predictor of future results. Do Computers Understand? Today’s computers don't have our general-purpose ability to understand. Watson was thrown off by badly formatted English. The human contestants, by contrast, had no problem with this. Just how good would Watson have to be, to call it - or should I say ‘him - intelligent? How could I judge this had happened? Alan Turing proposed an ingenious test in his 1950 paper Computing Machinery and Intelligence using “The Imitation Game? We now call the Turing Test. If we ask a series of questions to a computer and we cannot tell its responses from those a human would give, then the computer is, for all practical purposes, the same as a human. Since we are intelligent — or at least we hope we are — the computer must also be intelligent. QED. That’s all there is to the Turing Test. Puzzled? Let’s pick his argument apart. Imagine you are chatting away on Facebook with someone you don't know. They may have posted a photograph so you can see what they look like. The photo might be a fake; you have no real way to tell. What question would you ask the other ‘person’ to prove they were human and not a computer? There are obviously some giveaway questions. Please multiply the numbers 342,321 and 23,294 and give me the answer. This HOUSE_OVERSIGHT_015761

72 Are the Androids Dreaming Yet? would be very hard for a human but easy for a computer. If you got a very quick answer; the computer would have given itself away. But, the computer has been programmed not to give itself away, and it is free to give the answer slowly or even reply that the calculation is too hard. Our computer can say anything it likes, including lying to pass the test! If the computer can fool a questioner into believing it is a human then Turing argued the computer has shown it is at least as intelligent as we are. It used to be assumed that the field of broad general knowledge would be hard for a computer, but Watson has shown this is not so. With enough storage and a reasonable algorithm, winning a pub quiz is well within the capability of a modern computer. The really difficult questions for a computer are philosophical ones, novel questions and things that don’t fall into a pattern. For example, “Are you happy?” “What do you think of Shakespeare’s Hamlet?” “Ts there life after death?” “How went it?” “Think Differ...” If a computer could plausibly answer this sort of questioning for an extended period, say fifteen minutes, should we conclude it is intelligent, or do we need more time to be certain? Turing’s approach to certainty was simple. Just ask lots of questions. As you ask more and more questions, you will become increasingly certain you are talking to an intelligent being. He characterized it as a linear process; after 15 minutes of questioning you might be 99% certain and after a few hours 99.9% certain and after a few days completely certain. The problem with this approach is it does not flush out discontinuities. What if the questioning suddenly stopped without warning or explanation? A human responder is likely to worry that the questioner has had a heart attack and do something to find out what is going on including leaving the room. Humans can make creative leaps, solve non-computable puzzles or come up with a clever new joke. A humans could even announce the test is a waste of time and walk off. They just exercised free will! A computer cannot do these things. Each year a group of scientists enters a competition run by Cambridge University to win the Loebner prize, a competition to see how close a machine can come to passing the Turing Test. If you can beat the test you win $100,000. So far no one has come close and scientists are beginning to realize just how hard it is. HOUSE_OVERSIGHT_015762

Understanding 73 “On the Internet, nobody knows you're a dog.” New Yorker Cartoon With the anonymity the Internet provides we can imagine all sorts of strange scenarios if the Turing test could be passed. You would have no way of knowing what you were talking to. The New Yorker ran a cartoon back in 2000. “On the Internet no one knows you are a dog?” We come across a similar problem the other way around when we encounter bad customer support. A few years ago, while trying to get an answer to a computer problem, I became convinced the thing responding to my emails was a machine. The company did use machine responder technology so it could well have been. I asked it to prove it was human by putting the word marmalade into an English sentence and fixing my HOUSE_OVERSIGHT_015763

74 Are the Androids Dreaming Yet? problem. The human pretending to be a machine saw the joke, fixed my problem and replied “Marmalade is served with butter and toast.” The test worked! HOUSE_OVERSIGHT_015764

Understanding 75 ‘The sister test in robotics is equally hard. The goal is to simulate the physical human form, its movements and mannerisms. It’s easy to get close, but close is not good enough. The term “Uncanny Valley’ has been coined to describe the discomfort humans have with something that tries to simulate a human being but does not quite get there. I think it is part of the reason Madam Tussaud’s waxworks are so fascinating. Humans have a love-hate relationship with facsimiles of themselves. They love the flattery but feel a sense of revulsion at anything that comes too close. Searle and Turing In the Turing Test, we limited our senses to the purely symbolic: using only typed words on a screen. I could break the lock on the door and go into the room to see what was there. “Aha!” I would say. “I can see you're a computer, I, therefore, know you'll be good at sums and bad at creativity” But Turing wants us to see if the difference is given away purely through intellect. He argues there is no way to tell. But if you follow my argument from chapter 1, there is one way: ask the computer to find a non-computable solution to a mathematical puzzle. This is, in practice, a difficult test to pose because it might take a very long time. Twenty- five billion people have lived on planet Earth during the last 350 years, and about 5 million of them were mathematicians. None of them was able to solve the problem posed by Pierre de Fermat until Andrew Wiles turned up but this is a clear difference between humans and computers. However long you give a computer it would never be able to solve the problem. This creativity test would take centuries to run if non-computable thought was rare, but I think we see it often — on display even when we tell jokes. In which case computers and humans should be easy to tell apart: humans are the funny ones. I am not saying you can't build a brain; our brains are physical devices, after all. I just believe a computer or a mechanistic machine, cannot think like a human being. I like the Searle argument but qualitative arguments are insufficient. We need a quantitative argument. In the forthcoming chapters, I am going to look at the mathematical argument underlying the difference between human intelligence and computer processing. Before we do this let’s take one last look at a qualitative difference; the way computers and humans communicate. HOUSE_OVERSIGHT_015765

HOUSE_OVERSIGHT_015766

Chapter 3 BODY LANGUAGE & BANTER Body Language HOUSE_OVERSIGHT_015767

“England and America are two countries separated by a common language.” George Bernard Shaw “I speak two languages, Body and English.” Mae West “The body never lies.” Martha Graham HOUSE_OVERSIGHT_015768

in person for their second negotiation session, this time at the H6fdi House in Reykjavik. For five days, the leaders talked alone except for interpreters. Reagan badly wanted to develop the Strategic Defense Initiative; known by its nickname, ‘Star Wars. The idea was to put smart weaponry in space that could destroy ballistic missiles before they reentered the atmosphere. Reagan believed this would remove the threat of imminent destruction that had hung over the world since 1945. Gorbachev, on the other hand, felt this was just another escalation in the Cold War, and the Soviet Union would be forced to build yet more weapons to overcome the American defenses. He wanted Reagan's plans shelved, arguing that it broke the Anti-Ballistic Missile Treaty. He was probably right. The leaders talked back and forth, unable to overcome the impasse. At the end of the summit there was a mad scramble to announce some sort of deal, but this proved difficult. In the last moments before they had to conclude a communiqué, Reagan suggested they abolish ai/ nuclear weapons. Reagan's negotiating team was horrified and shut the door. For decades, the American strategy had been to use nuclear weapons as a deterrent against the apparent numerical advantage of the Soviets. In all the potential scenarios analyzed by the Pentagon, Russian forces ended up overrunning American forward positions — otherwise known as Western Europe! The only way to stop them was through a release of nuclear weapons, which, inevitably escalated to all-out nuclear at. [: the summer of 1986 Ronald Reagan and Mikael Gorbachev met + X Ronald Reagan and Mikael Gorbachev HOUSE_OVERSIGHT_015769

80 Are the Androids Dreaming Yet? war. It was assumed this inevitable progression deterred the aggression in the first place, and the threat of mutually assured destruction kept the world peaceful. Giving up this tenet of defense strategy was something the American military just could not contemplate. Many people did not think it a rational defense strategy; it seemed appropriate the acronym for mutually assured destruction is MAD, but this was the status quo. We now know our worry over Russian superiority was groundless. The West’s technological advantage, founded on the invention of computing and sophisticated materials technology, gave us a huge advantage. In the only battle to be fought in the 20" century between Russian and Western tanks, during the first Iraq war, most of the Russian tanks were destroyed with no losses to American tanks. We know this now, but we are talking of a time when paranoia over the Soviet advantage was the common view. There is speculation that Reagan had muddled intercontinental ballistic missiles with all nuclear weapons. I do not think this is true. Reagan was a man of vision, quite comfortable with using his folksy way to convey sincere belief, and I think abolishing all nuclear weapons was in his mind. It would have been a breathtaking moment. In the end a rather feeble communiqué was put together and the talks declared a technical failure. But, both leaders had seen eye-to-eye; both were prepared to make major concessions and both wanted an end to the old strategy of mutually assured destruction. Wiping each other out was no longer considered a successful outcome! The meeting, and H6féi House in Reykjavik HOUSE_OVERSIGHT_015770

Body Language & Banter 81 the fundamental thawing of relations between East and West, was to lead to the Intermediate-Range Nuclear Forces Treaty and the end of the Cold War. Face-to-Face Communication What really happened between these two leaders when they met and talked? Was it a mechanical process of offer and counter-offer, as easily executed by fax, or is human interaction more complex than this? Reagan, as a young man, had been a liberal, sympathetic to socialist ideals until a painful strike in California caused him to lose faith in the politics of the left. Gorbachev, a lifelong Communist, was desperate to reform the Soviet economy and make it more competitive. He, also, had come to see the hypocrisies that could emerge in far left-wing ideology. I don’t believe this common experience could have been communicated by fax or email. Indeed, I am sure these specific points were never made, but the nonverbal communication must have conveyed something of their common background and purpose. When we phone someone or exchange emails, the interaction is factual, there is no body language, and we rarely laugh. When we travel to meet someone, we spend a great deal of time with them. The average length of a phone call is two and a half minutes, but meetings, especially when one party has travelled to see the other, can be hours long. When humans meet they greet each other, shake hands, sit in the same room, talk at length, and laugh. Body language is important; people mirror each other’s postures, adopt open and receptive stances, and make eye contact. You can see this in the picture of Reagan and Gorbachev above. Body language allows us to convey qualitatively different things, such as trust and happiness. It is very expressive; you can see the more guarded postures of Yasser Arafat and Shimon Pérez below, just after they negotiated a landmark peace deal. Can you tell if the leaders smiles are false? Communication Communication is one of mankind’s greatest expenditures. The US telephone system is arguably the largest machine on the planet, while the world’s mobile phone networks have a capital value of $2.5 trillion, greater by an order of magnitude than all the steel plants in the world put HOUSE_OVERSIGHT_015771

82 Are the Androids Dreaming Yet? Yasser Arafat and Shimon Pérez together. This lifeblood of our existence — long-distance communication between human beings — turns out to be amazingly difficult, even with all our clever technology. In recent years the Internet has, in theory, allowed each and every person to communicate freely with any other person on the planet. In some of the most distant parts of the world mobile phones, and projects such as; ‘One Laptop per Child’ are rapidly bringing unlimited communication to all. This communication can be personal, one-to-one, or broadcast: I can talk to people interested in a particular topic directly. As we watch the Arab world democratize, catalyzed by the Internet, there is no question that digital communication has now become a major force in the world. Yet, people don’t communicate over the Internet as much as you would expect; they often use the Internet to set up phone calls during which they arrange meetings! This is odd. We have a fantastic phone system and sophisticated communication technologies; email, video and instant messaging. Yet, we still choose to travel when we want to communicate. On the face of it, there should be no difference between a phone call and a meeting. In principle the same information can be conveyed. Yet when we want to really understand someone, we always go to meet HOUSE_OVERSIGHT_015772

Body Language & Banter 83 Smiles Fake or Real in person. No great treaty or big industrial contract has been negotiated without a face-to-face meeting. We see this daily: people talking on the phone get to a certain point, give up, and arrange to meet in person. The consequence is that we spend $550 billion annually, flying around the globe to meet each other. Each day the world’s population takes three million plane flights. Around 80% of these are business flights. Some are people emigrating or going to do specific manual tasks, but most are to have meetings. We have always assumed that this is because the parties are unable to reach a sufficient level of trust over the phone and need face-to-face interaction to build that trust, but it may be that the parties are not able to convey sufficient information to fully understand each other. Face-to-face meeting may convey much more information than we think. HOUSE_OVERSIGHT_015773

84 Are the Androids Dreaming Yet? Smiles When we smile naturally we use a full set of facial muscles, including the muscles around our eyes. When the smile is forced those eye muscles remain passive and the smile, although superficially the same, is missing something. You can’t put your finger on it, but the look is insincere. A study of marriages in the USA analyzed smiles in wedding photographs. The couples with false smiles divorced much earlier than the genuinely happy couples. Similarly for high school photos; people with genuine smiles at 18 years of age were happier later in life and in more stable relationships. Smiling is really important. It is good to be around people who smile, they are more successful — and nicer. There is also a curious reverse effect. The link between our minds and bodies is much more fundamental than we thought. If you grasp a pencil between your teeth, it forces you to smile. Try it. The mere act of smiling is found to make you happier, it causes the release of the chemicals called endorphins which improve your feeling of well-being. Micro-expression Analysis Since the involuntary movements of the muscles around our eyes give away genuine happiness, a whole science has evolved looking for other biological cues to mood. The two most interested groups are the FBI, trying to detect lies, and poker players, trying to make money! Much has been written on the topic, including a few best sellers, but the evidence for micro expressions is mixed. Regardless of whether involuntary actions give away our emotions, humans voluntarily use a great deal of body language when talking. Body Language A study by Albert Mehrabian is often cited to say 93% of the information in a conversation comes through nonverbal cues. This is misquoted. The study really stated 93% of the emotional content is nonverbal. That's more believable. And further studies have shown when there is doubt, nonverbal cues win over verbal information every time. The rule is sometimes laid out as the 7%-38%-55% rule — 7% words, 38% tone of voice and 55% body language. Remember this is emotional content, your conviction and sincerity. You will still have to get over the factual information you want to convey. HOUSE_OVERSIGHT_015774