Richard Feynman

by John Gribbin

  The operation, performed at the local hospital in Pasadena, seemed to be a success, but after it Feynman, although still cheerful, was not only physically affected by the damage done to him by the cancer, but also knew that he was now living on borrowed time. Of course, all of us are under sentence of death, in the long term. But in Feynman’s case, the future had begun to close in. Subscribing to his own philosophy of ‘never fool yourself’, he treated the cancer as an interesting case study, and looked up all he could about it, observing the changes going on in his own body like a scientist watching an experiment. The cancer, a so-called liposarcoma, was malignant, and although the huge lump had been removed, the textbooks said that there was essentially zero chance of his surviving for another ten years.

  Meanwhile, life continued as normal, including his physics, his teaching, drawing, holidays, intermittent efforts to get the Tuva project under way and drumming with Ralph Leighton. The drumming sessions, though, became as much storytelling sessions as anything else, as the tape of the ‘Safecracker Suite’ highlights.26 Feynman had always been an inveterate storyteller, but now he seems to some extent to have taken stock of his life by pouring out the anecdotes that became the two books on which Leighton collaborated with him. Leighton feels that he was simply the right person in the right place when Feynman was ready to make his stories available to a wider audience. He recalls an occasion at dinner when Feynman mentioned that he had been being interviewed about his scientific work, but that whenever he got on to ‘the good stories’ the interviewer would shut his tape recorder off. ‘Feynman was kind of complaining about that,’ says Leighton, ‘so I piped up that those were my favourite stories, so let’s see if we couldn’t write them down in an organized way.’

  All things considered, it’s hardly surprising that Feynman produced little in the way of new ideas in physics after his sixtieth birthday. But fate was to give him one last great opportunity to demonstrate to the world the way that a top scientist thinks, and how the scientific method should be applied to solving problems. This opportunity, reluctantly taken, also made Richard Feynman even more famous than he had been already; above all, though, it was to highlight a damning example of what can happen to organizations, as well as to people, when they start fooling themselves by believing what they want to be the truth, rather than what really is the truth.

  Notes

  1. Davies has told the story of this trip in his contribution to No Ordinary Genius.

  2. See note 1.

  3. Weiner interview with Feynman, quoted by Gleick.

  4. Surely You’re Joking.

  5. See note 3.

  6. Leighton, interview with JG, April 1995.

  7. Feynman, The Character of Physical Law.

  8. Interview with JG, April 1995; see also Most of the Good Stuff.

  9. Helen Tuck, interview with JG, April 1995; the biography she is referring to is the one by Gleick.

  10. Mehra.

  11. Conversation with JG, April 1995.

  12. Norman Dombey, interview by JG, October 1995. Dombey also says that the only occasion on which he was present at a Caltech seminar and Feynman didn’t tear the speaker’s arguments to shreds was when Oppenheimer was the speaker. But he couldn’t say if that was down to Oppenheimer’s brilliance, or Feynman’s respect for his boss from the Los Alamos days.

  13. See note 12.

  14. This point was made in an interview with JG, in October 1995, by Norman Dombey, who completed his PhD at Caltech under Murray Gell-Mann in 1961, and worked there in 1961–2 as a postdoctoral researcher.

  15. See Sherman’s contribution to No Ordinary Genius.

  16. The former student, now an eminent physicist in his own right, asked to remain anonymous in order not to suffer any backlash from Gell-Mann about this remark. The same physicist told us that ‘Murray always had to be right, even when he wasn’t, but Feynman wasn’t afraid to admit when he’d made a mistake.’

  17. Correspondence with JG, January 1996.

  18. Interviewed by JG in April 1995, Helen Tuck recalled that Richard and Carl Feynman once had a private meeting with Uri Geller, and found that none of Geller’s claimed powers operated under their careful scrutiny.

  19. Surely You’re Joking.

  20. Quoted by Mehra.

  21. Carl’s teacher called Gweneth one day, expressing concern that Carl, who seemed a bright child, had scored only 129 in an IQ test; Gweneth told the teacher that this wasn’t too bad, since Richard’s IQ was only 125.

  22. Leighton, Tuva or Bust!

  23. Gweneth Feynman, as told to Gleick.

  24. Leighton, Tuva or Bust! There is some confusion about the mass of the tumour, which Gleick says weighed six pounds; but there is no doubt about the damage it caused to Feynman’s internal organs. Freeman Dyson, usually a reliable witness, also quotes a figure of six pounds, in his book From Eros to Gaia.

  25. David Goodstein, interview with JG, April 1995; see also No Ordinary Genius.

  26. See Bibliography.

  12 The last challenge

  The event that would make Feynman known to a wider audience than ever before was the explosion of the space shuttle Challenger in 1986. But although his work for the Challenger inquiry became the best known of his activities in the final decade of his life, it was far from being his only piece of technical work after he turned 60. Although Feynman made no major contributions to theoretical physics in the 1980s, he did have an absorbing scientific interest – one that harked back to his childhood fascination with solving mathematical problems, and to his work at Los Alamos in charge of the Theoretical Computations Group. With his son Carl (whose interest had, happily for Feynman, switched from philosophy to computing), he became involved in the development of the next ‘big idea’ in computers, parallel processing.

  Carl studied at MIT, where his father introduced him to Marvin Minsky, one of the pioneers of research into the possibility of creating artificial intelligence. Through Minsky, Carl met Danny Hillis, a graduate student who had a crazy ambition, to build a giant computer. ‘Well,’ said Carl, ‘what did I know? I was seventeen years old, and I thought it would work – nobody else did.’1 So Carl became one of the undergraduates helping Hillis out with his thesis project.

  The idea behind the plan to build a giant computer was that instead of having one huge machine (one ‘central processor’, in computer jargon) working on a single huge problem, you would break the problem down into smaller bits and feed each of the pieces to a smaller processor, with all the small computers linked together so that they could cooperate in taking the various calculations through to their logical conclusion. This is parallel processing, which has begun to become an important practical possibility in the 1990s. It is, of course, exactly what Feynman did at Los Alamos in the 1940s, only then his computers (the parallel processors) were human beings operating calculating machines, every person solving their own tiny bit of the problems involved in making the first atomic bombs. The dream Hillis had was of a million computers working together in this way – a million processors operating in parallel. As his dream began to look like becoming a reality in the early 1980s, he had to lower his sights a little and settle for 64,000 processors working together, sixteen of them on each single computer chip, with 4,000 computer chips wired together and programmed in the right way to do the problem solving. Anybody who knew Feynman could have guessed that once he heard about the project he would have to get involved.

  It’s no coincidence that Feynman was acquainted with Marvin Minsky He had maintained his interest in computation, off and on, since his work at Los Alamos, and by the end of the 1970s that interest extended to the theoretical limitations of computers, as well as to the practical aspects of building them and making them work. As a result of a question posed by the head of the computer department at Caltech, Feynman had tried to discover the minimum amount of energy required, in theory, to carry out a computation, and was intrigued to discover that there is no lower limit. No mat
ter how little energy was available, an ideal computer would still be able to carry out its work.

  At a meeting on computation at MIT, Feynman was pleased to discover that a real computer expert, Charles Bennett, had reached the same conclusion. This led to discussions about the limits set by the rules of quantum physics, a puzzle worried over by several physicists, and involving a visit by Bennett to Caltech. Once again, the surprising conclusion was that there are no limits, except for the physical ones, like size. The smallest, fastest computer possible would store numbers on individual atoms, as a string of binary digits (zeros and ones) indicated by some property such as the spin of the atom (up or down), and carry out computations using those strings of numbers.

  Feynman was also intrigued by the way in which the workings of artificial computers differ from the workings of the human mind:

  I found it amusing that the things I consider myself smart at – for instance, when I was young I was good at calculus, playing chess, and other logical things – could be done by computers … Mathematical and logical thinking, which we were always so proud of, that they can do. It’s illogical thinking that … we do immediately, easily, as the eye jumps from one part of the scene to another and integrates the whole picture into a room with chairs and furniture and everything that we see, that’s difficult [for computers]. It’s very interesting. Altogether, computers are fascinating and the problems that they can do are fascinating.2

  It’s actually slightly more subtle, and fascinating, than even this example indicates. Even the things that computers do well, like playing chess, they do not necessarily do in the same way that people do them. A good computer chess program works by considering a large number of possible moves (perhaps every move it can make), looking ahead to every possible response to each move, then at each possible next move, and so on (down to a ‘depth’ decided by the power of the computer and the amount of memory that it has available) to decide which is the best move to play. A good human chess player looks at the whole pattern of the pieces on the board, developing a feel for the balance of power, and often deciding on a particular plan of campaign (or, just as important, rejecting an alternative plan) because it fits (or does not fit) the overall ‘feel’ of the game.

  In spite of what Feynman said about the things he used to be good at himself, what made Feynman a great scientist was not his ability to think logically and carefully like a machine. His great achievements – for example, QED itself – came about as much through intuition as anything else, through having a ‘feel’ for physics, knowing instinctively (which means, as a result of this subconscious process that he talks about) what is the right approach. He never did develop a completely logical version of the path integral approach to QED and Feynman diagrams; to this day, the great successes of this approach are built upon making inspired guesses to develop a description of what goes on in some interaction, and then tinkering with the resulting diagrams and equations to make the guesses agree more and more with the real world of experiment. Feynman seemed to understand how nature must respond in different circumstances, in the same way that nature herself understands. A ball following a curved trajectory through a window doesn’t have to calculate a complicated mathematical equation in order to follow the path required by the Principle of Least Action, and Feynman didn’t have to invent a rigorous mathematical proof in order to know that his version of QED worked. He was, indeed, a magician, not an ordinary genius.

  Feynman was also attracted by crazy ideas. If everybody worked in the same safe areas of conventional research, after all, progress would be very slow. He always encouraged people to try out wacky ideas, because although the chance of any one of the ideas being fruitful might be small, the potential rewards for anyone who did hit the jackpot would be enormous (of course, you had to know where to draw the line, and Feynman did not encourage people to pursue wacky ideas that disagreed with experiment; this was not an endorsement of spoonbending or ESP). So when, in the spring of 1983, Hillis told Feynman that he was planning to leave the MIT Artificial Intelligence Lab and start a company to build a computer using a million parallel processors, the reaction he got – ‘That is positively the dopiest idea I ever heard’3 – was actually a ringing endorsement of the plan. Over lunch, Feynman agreed (perhaps ‘insisted’ would be a better description) that he would spend his summers working for the company, as yet unnamed, that Hillis planned to set up. Apart from the fun of new problems to solve, it would give him more time with Carl.

  Although delighted to have a Nobel laureate on his letterhead (when he got around to having a letterhead), Hillis had no real idea what to do with Feynman. When Dick arrived in Boston that summer to start work, the company had only just been incorporated, and was largely staffed by young people who had not yet formally graduated from MIT, although they had finished their courses there. When he asked them what his job was, after some discussion they told him that he could advise them on the application of parallel processing to scientific problems. He was having none of that. ‘Give me something real to do’, he said.4 So they sent him out to buy some office supplies, and when he got back they told him that he could analyse the way in which the individual processors would communicate with each other – a system known as a router, which would be responsible for finding a way for each communication between individual processors to travel along the wires linking them into one machine, without interfering with other messages travelling along the wires.

  Feynman focused intently on the problem, but also found time to help out in wiring up the machine, setting up the machine shop and shaking hands with investors in the project. He also made a major contribution to setting up the structure of the company, encouraging Hillis to set up different teams, each under a group leader, working on specific tasks, just the way things had been done at Los Alamos (itself, in effect, a form of parallel processing). Just about every facet of his lifetime of experience turned out to be relevant to something that was going on in the project.

  By the time he had completed his main task, analysing the requirements of the router, the company had a name – Thinking Machines Corporation – and so did the machine the Connection Machine. Feynman’s analysis showed that in order to work efficiently, each of the chips in the Connection Machine would require a minimum of five buffers for its communications with the rest of the machine, to prevent a logjam of messages piling up. Conventional computer wisdom had it that they would require seven buffers per chip, and in order to play safe the team decided to go with the conventional wisdom. But when it became time to make the chips, it turned out that they were too big to be manufactured using standard technology. With five buffers on each chip instead of seven, the manufacturing would be straightforward. Hoping that Feynman was right, they went ahead with the smaller design. It worked, and the first program was successfully run on the Connection Machine in April 1985.

  By then, Feynman had made many more contributions to the project. He showed the young team the importance of cutting out jargon and explaining their work clearly, using everyday language wherever possible, when describing it to other people (including those investors). He soldered circuit boards, and helped paint the walls. Meanwhile, at Caltech, a conventional computer was being built to carry out computations simulating what happens when quarks interact with one another, and Feynman wondered whether the Connection Machine (which had not yet been completed) could do that. He made up a computer program that could tackle the job using the principles of parallel processing, and then worked through some of the steps in the calculations that would be involved on paper, to see how much processing power would be needed to do the real job, and how long it would take. He was, in fact, simulating with his paper and pencil the operation of a computer simulating the interactions between quarks using the rules of quantum chromodynamics. He found that it would work – the Connection Machine, when completed, would be able to carry out calculations involving QCD faster than the conventional machine being built at Caltech specifically to
carry out calculations in QCD! ‘Hey, Danny!’, he yelled. ‘You’re not gonna believe this, but that machine of yours can actually do something useful!’5

  In Most of the Good Stuff, Hillis describes the last piece of work he did with Feynman, a simulation of the way in which populations of living creatures evolve, in accordance with the Darwinian principle of Natural Selection. Hillis had been surprised to discover that in computer simulations populations seemed to stay fairly stable for many generations, and then to evolve suddenly into new forms. This echoes the appearance of many features of the fossil record, which has led to a variation on the Darwinian theme known as punctuated equilibrium. Together with Feynman, Hillis worked out a theory to explain this, a mathematical model of evolution at work. Then he discovered that it had all been done before and that biologists already knew about it. Disappointed, he called Feynman to pass on the bad news. But Feynman was elated. ‘Hey, we got it right! Not bad for amateurs!’ As ever, what mattered to him was the pleasure of solving the problem himself. He didn’t care whether someone else had solved it first.

  Feynman was the ideal person to work on the Connection Machine, a father figure to the team, because, as Hillis says, ‘he was always searching for patterns, for connections, for a new way of looking at something’. But ‘the act of discovery was not complete for him until he had taught it to someone else’.

  By the middle of the 1980s, Feynman was seriously ill again, and his friends knew that he could not have long to live. But he was to have one last opportunity to find a new way of looking at something, to make connections, and, best of all, to explain his discovery to a large audience. Sadly, though, the opportunity came about as a result of a human tragedy which stunned the entire country.