George Dyson: There's Plenty of Room at the Top: Long-term Thinking About Large-scale Computing

### Long-term thinking about large-scale computing Ever since his 1997 breakthrough book, _Darwin Among the Machines_ , Dyson has become regarded as a leading historian and interpreter of computer science, bringing a rigorous and unconventional perspective. Thus his willingness to examine the long-term prospects for mega-scale computing. Most computer people are averse to discussing seriously any future beyond ten years. With the Dyson seminar, our series begins to get down to specific cases of applying long-term thinking. "Your now is only as long as you remember. To way to understand the future is give the present more depth," said George Dyson at the start of his Seminar at Long Now last Thursday. Then he proceeded to help us remember what really went on at the very beginning of computerdom, back in 1947 at Princeton's Institute for Advanced Study (IAS) when John von Neumann and colleagues were inventing the digital world. Thanks to his physicist/mathematician father Freeman, George grew up at the IAS, along with his older sister Esther. Last year he went back to spend an entire year burrowing through the ancient files and logbooks of the von Neumann period. A skilled photographer, he has hundreds of richly evocative photos of those records to show. In them the past comes to life. Nion McEvoy said George's artifacts are deep and consequential, like relics from earliest Christianity. The artifacts showed that at the beginning the code was small and sound and the machines (vacuum tubes and oscilloscope tubes) were maddenly flaky. By the end of that period the machines had become reliable, and most of the problems were in the ever-growing code. Von Neumann wrote a landmark paper, "Reliable Organizations of Unreliable Elements." Strings of bits (then still called binary digits) come in two forms, Dyson reminded us---sequential in time and structural in space. The sequences make things happen; stored as structure they are memory. Building an architecture that would manage the transitions adroitly and also scale up radically for decades was a profound achievement. Engineers were never particularly welcome at the IAS. One letter shows an administrator complaining that the computer people from the basement were using excessive quantities of sugar and tea, and doing so at all hours of the night. Programmers evidently have been caffeine and sugar swilling nocturnal troglodytes from the very start. Like Eniac, the IAS computer was mainly used for atomic bomb studies---that's where the money came from. The researchers also worked on weather prediction (with approaches still in use, only now they're fast enough to predict the future instead of the past), and von Neumann assumed we would eventually use them to control the weather---an assumption George shares. By 1953 at the IAS Nils Barricelli was working on "Symbiogenetic Evolution Processes Realized by Artificial Methods," running the first artificial life programs. (Their source code is still around and could come to life again.) Danny Hillis, who has also done a-life experiments, noted that there are no seriously long-lived computations going on (max is a year, for prime numbers and the like). Maybe Long Now should start some. George ended by looking at the larger picture. He said that computation at present is still parasitic on us, but it probably is in the process of becoming part of life itself---it's just another system of genetics. Evolution, after all, is massively parallel computation, and now the world of code is evolving, perhaps to reshape the tree of life itself. As von Neumann's famous work on Theory of Games showed, coalitions are the most important element. "The great collective organism we're becoming part of will have a completely different sense of time."