Hackers

by Steven Levy

  And they formed an exclusively male culture. The sad fact was that there never was a star-quality female hacker. No one knows why. There were women programmers and some of them were good, but none seemed to take hacking as a holy calling the way Greenblatt, Gosper, and the others did. Even the substantial cultural bias against women getting into serious computing does not explain the utter lack of female hackers. “Cultural things are strong, but not that strong,” Gosper would later conclude, attributing the phenomenon to genetic, or “hardware,” differences.

  In any case, only rarely were women in attendance at the Chinese restaurant excursions or the sessions at the Tool Room next door to TMRC. So naturally, one did not have to look one’s best. Greenblatt, perhaps, took this to an extreme. He worked on several mammoth projects in the mid-sixties, and would often get so wrapped up in them that his personal habits became a matter of some concern to his fellow hackers.

  After he dropped out of school, Greenblatt had taken a job at a firm called Charles Adams Associates, which was in the process of buying and setting up a PDP-1. Greenblatt would work at their offices near Boston’s “Technology Highway” outside the city during the day and drive thirty miles back to MIT after work for some all-night hacking. Originally he moved from the dorms to the Cambridge YMCA, but they booted him out because he wouldn’t keep his room clean. After his stint at Adams, he got rehired at the AI Lab, and though he had a stable living situation—as a boarder in a Belmont house owned by a retired dentist and his wife—he would often sleep on a cot on the ninth floor. Cleanliness was apparently a low priority, since tales abounded of his noticeable grunginess. (Later Greenblatt would insist that he was no worse than some of the others.) Some hackers recall that one of the things Greenblatt’s hacking precluded was regular bathing, and the result was a powerful odor. The joke around the AI lab was that there was a new scientific olfactory measure called a milliblatt. One or two milliblatts was extremely powerful, and one full blatt was just about inconceivable. To decrease the milliblatts, the story goes, hackers maneuvered Greenblatt to a place in the hallway of Building 20 where there was an emergency shower for cases of accidental exposure to chemicals, and let it rip.

  Gosper would sometimes tweak Greenblatt for his personal habits, and was particularly bothered at Greenblatt’s habit of rubbing his hands together, which resulted in little pieces of dirt falling out. Gosper called these blattlies. When Greenblatt worked on Gosper’s desk and left blattlies behind, Gosper would make a point of washing the area with ammonia. Gosper would also sometimes kid Greenblatt about his awkward speech patterns, his frequent coughing, his poor spelling, his mumbling—even though many of Greenblatt’s expressions became integrated into the specific vernacular which all the hackers used to some degree. For instance, it was probably Greenblatt who popularized the practice of doubling words for emphasis—like the times he’d get revved up explaining something to Gosper, Kotok, and Samson, and the words would get tangled up, and he’d sigh, saying, “Oh, lose-y lose-y” and begin over. Gosper and the others would laugh—but, like the way a family will take on a baby’s speech patterns and cute malapropisms, the community adopted many Greenblattisms.

  Despite these odd personal traits, the hackers held Greenblatt in awe. He was the way he was because of conscious priorities: he was a hacker, not a socialite, and there was nothing more useful than hacking. It so consumed him that he sometimes would go six months without finding time to pick up his MIT paycheck. “If he randomly sat around and tried to articulate what he was thinking and doing all the time, he wouldn’t have gotten anything done,” Gosper would later say. “If he worried how to spell things, he wouldn’t have gotten anything written. He did what he was good at. He was a complete pragmatist. What people thought, be damned. If anyone thought he was stupid or nerdly, that was their problem. Some people did, and they were wrong.”

  Gosper could appreciate Greenblatt’s single-mindedness because his own insistence on graduating (which he did in 1965) had led him to trouble. It was not that his final year at MIT was an academic disaster, because he managed to fulfill the graduation requirements by a slim margin. The problem was a pact he had made with the United States Navy. Before he entered MIT, he’d taken a civil service exam and placed high enough to be included in an exclusive student engineering development program. He worked summers for the Navy, which paid half his tuition and required him to work there for three years after graduation. When Gosper signed up, there had been an escape clause that allowed you to postpone your commitment if you went to graduate school; and if you could get a corporation to pay off the Navy’s three-thousand dollar investment after that, you’d no longer be obligated. But during Gosper’s senior year the graduate school loophole closed. Only a buyout would save him, and he didn’t have the money.

  The prospect of going into the Navy was hideous. During his summer employment stints he had been exposed to a pathetic system that was antithetical to the Hacker Ethic. Programmers were kept in a room totally separated from the machine; sometimes, as a reward for years of service, they would let a particularly obedient worker venture into the computer room and actually see his program run. (One woman, the story goes, was allowed this privilege, and the sight of the lights flashing and disks whirring caused her to faint.) In addition, Gosper’s Navy boss was a man who could not understand why the logarithm of the sums in a given equation was not the sum of the logarithms. There was no way in hell Bill Gosper was going to work under a man who did not know why the logarithm of the sum was not the sum of the logarithms.

  Then there was Gosper’s perception that the Navy was in bed with Univac. He considered the Univac machine a grotesque parody of a computer, a Hulking Giant. The Navy had to know it was a basically phony computer, he figured, but used it anyway—it was a classic example of the inevitably warped outcome of Outside World bureaucracy. Living with that machine would be immersion in hell. Gosper used computers to seek things that no one had ever found before, and it was essential that the computer he used be optimal in every way. The PDP-6 was the best thing he had found so far, and he was determined not to leave it, especially for a dog like the Univac. “If I see a machine has some incredibly stupid thing wrong with it, some error in its design or whatever, it just irritates the hell out of me,” Gosper would later explain. “Whereas the PDP-6 always seemed like an infinitely perfectible machine. If there was something wrong, you would change it. In some sense, we lived inside the damn machine. It was part of our environment. There was almost a society in there . . . I couldn’t imagine being without a PDP-6.”

  Gosper was determined to find the money to pay back the Navy, and to earn it while working for a company with a PDP-6. He fulfilled these rigid criteria by landing a job with the firm that Greenblatt had worked for that past year, Charles Adams. The fact that the Adams company never quite got their PDP-6 working right (Greenblatt insists that he did his part of the preparation adequately) did not seem to upset Gosper: what freaked him was the fact that Charles Adams scrapped the project and bought a carbon copy of the same Hulking Giant Univac that the Navy had.

  But by that time more funding for Project MAC had come through, and Bill Gosper found his way onto the payroll. He hardly had to change his habits, since during his whole stint at Adams he had been working on the PDP-6 on the ninth floor every night.

  • • • • • • • •

  By then, Greenblatt was in full hacking swing. One of the first projects he worked with on the PDP-6 was a LISP compiler, to allow the machine to run the latest and most nifty version of Jobn McCarthy’s artificial intelligence language. Young Peter Deutsch had written a LISP for the PDP-1, but it was not too effective, since the One had less memory; and LISP, which works with symbols and not numbers easily translated to binary, consumes an incredible amount of memory.

  Some people, notably Gosper, thought that LISP would be a waste of time on the PDP-6 as well. Gosper was always concerned with what he considered the atrocious lack of comput
er power in those days, and later would marvel at how ignorant they all were in the AI lab, trying impossible tasks and blaming their failures not on the piddling machines they had, but on themselves. In his senior year, Gosper had been put to work by Minsky on a display that would test whether a certain visual phenomenon was binocular or monocular. Gosper did manage to come close with a clever, clover-leaf shape, which at least displayed the phenomenon, but generally was banging his head against the wall trying to make the machine do more than it could do. One of the tasks that Gosper considered impossible was a useful LISP on a PDP-6—it might be nice as a symbol evaluator, but not to do anything. He considered it one of Minsky’s follies that Greenblatt and the others had been tricked into implementing.

  But Greenblatt saw more. Though he realized that LISP on the PDP-6 would be to some extent a hack, not fully pragmatic, he did see the need to move toward it. It was a powerful language that would help the field of artificial intelligence move forward: it was the language by which computers would do extremely difficult tasks, by which they could actually learn. Greenblatt was just starting then to have a certain vision of the future, an inkling of a technical implementation of the hacker dream. So he and some others—even Kotok came down from DEC—began implementing LISP on the PDP-6. They filled the blackboards of TMRC with layers and layers of code, and finally got it going on the machine.

  The crucial sections were written by Greenblatt and another hacker. Two or three people on a project were considered The Right Thing—far fewer than IBM’s so-called “human wave” style of throwing dozens of programmers at a problem and winding up with junk. And it was better to rely on two or three people than on a single crusader—so that when one person was at the end of his thirty-hour phase, someone else could come in and keep hacking. Kind of a tag team project.

  With PDP-6 MacLISP (named for Project MAC), the hackers began integrating that computer language into their programs, and even into their conversation. The LISP convention of using the letter “p” as a predicate, for instance, was the inspiration for a common hacker style of asking a question. When someone said “Food-P?” any hacker knew he was being asked if he wanted to get something to eat. The LISP terms “T” and “nil” came to stand, respectively, for “yes” and “no.” LISP’s acceptance did not diminish the hacker love for assembly language, particularly the elegant PDP-6 instruction set. But as Greenblatt and even Gosper later realized, LISP was a powerful system builder that fit neatly into the hands-on Hacker Ethic.

  DEC had shown an interest in MacLISP, and Kotok arranged for Greenblatt and the others to go to Maynard late at night to work on the program, type in their code, and debug it. It was all part of the easy arrangement between MIT and DEC, and no one questioned it. The Right Thing to do was to make sure that any good program got the fullest exposure possible, because information was free and the world would only be improved by its accelerated flow.

  After working on MacLISP, Greenblatt was perhaps the most authoritative of the systems hackers on the PDP-6. The new administrator of the AI lab, a young man from the Southwest named Russell Noftsker, had hired Greenblatt mainly to maintain and improve the organic creation that is a computer operating system. But Greenblatt’s vision did not stop at systems; he was intensely drawn by the concepts of artificial intelligence. He decided to use the system to actually do something in that realm, and, since he had been a chess player all his life, it was only logical that he work on a chess program that would go far beyond Kotok’s effort and beyond the other AI chess projects that had been attempted at various labs around the country.

  Like any good hacker, no sooner did he decide to do something than he began work on it. No one asked him for a proposal. He didn’t bother to notify his superiors. Minsky did not have to ponder the relative virtues of the project. There were no channels to go through because in the mid-sixties, in those early days of the AI lab, the hackers themselves were the channels. It was the Hacker Ethic put to work, and Greenblatt made the most of it.

  He’d seen a game played by the Kotok program and thought it was crap. Basically, those guys did not know how to play chess: swayed by the romance of a computer making moves, they had somehow forgotten the idea that the name of the game was to take the other guy’s pieces. Greenblatt’s program used sophisticated artificial intelligence techniques to try and figure out moves in accordance with certain criteria that he considered good chess. Working with a couple of other hackers, Greenblatt went on a coding blitz. He’d manage to get four hours of PDP-6 time a day, and he’d keep writing offline when he wasn’t on the machine. He got the program actually playing chess in one week. The program was debugged, given features, and generally juiced up over the next few months. (Greenblatt was eventually offered an MIT degree if he would write a thesis about his chess program; he never got around to it.)

  Circulating around MIT around 1965 was a notorious Rand Corporation memo called "Alchemy and Artificial Intelligence.” Its author, an academic named Herbert Dreyfus, lambasted the field and its practitioners. To hackers, his criticism was particularly noxious, since the computer was their implicit model of behavior, at least in their theories of information, fairness, and action. Dreyfus focused on the computer’s ridiculously limited structure (compared to the structure of the human brain). His coup de grace was the blunt assertion that no computer program would be able to play a good enough game of chess to beat a ten-year-old.

  After Greenblatt finished his chess program, called MacHack, MIT invited Dreyfus to play the PDP-6. The hackers gathered round to watch the computer surrogate of Richard Greenblatt play this cocky, thin, red-headed, bespectacled anticomputer opponent. Artificial intelligence pioneer Herbert Simon, who watched the match, later was quoted as saying that it was

  . . . a real cliffhanger. It’s two woodpushers . . . fighting each other . . . Dreyfus was being beaten fairly badly and then he found a move which could’ve captured the opponent’s queen. And the only way the opponent could get out of this was to keep Dreyfus in check with his own queen until he could fork the queen and king and exchange them. And the program proceeded to do exactly that. As soon as it had done that, Dreyfus’ game fell to pieces, and then it checkmated him right in the middle of the board.

  Peter Samson later recalled the scene immediately following Dreyfus’ loss: the defeated critic looked around at the assembled MIT professors and hackers, including a victorious Greenblatt, with a look of puzzlement. Why weren’t they cheering, applauding, rubbing it in? Because they knew. Dreyfus was part of that “real world” that couldn’t possibly comprehend the amazing nature of computers, or what it was like working with computers so closely that a PDP-6 could actually become your environment. This was something which Dreyfus would never know. Even Minsky, who never really immersed himself in the thirty-hour-day, seven-day-week assembly-language baptistery, had not experienced what the hackers had. The hackers, the Greenblatts and the Gospers, were secure in having been there, knowing what it was like, and going back there—producing, finding things out, making their world different and better. As for convincing skeptics, bringing the outside world into the secret, proselytizing for the Hacker Ethic—all that was not nearly as interesting as living it.

  Chapter 5. The Midnight Computer Wiring Society

  Greenblatt was hacker of systems and visionary of application; Gosper was metaphysical explorer and handyman of the esoteric. Together they were two legs of a techno-cultural triangle which would serve as the Hacker Ethic’s foundation in its rise to cultural supremacy at MIT in the coming years. The third leg of the triangle arrived in the fall of 1963, and his name was Stewart Nelson.

  Not long after his arrival, Stew Nelson displayed his curiosity and ability to get into uncharted electronic realms, traits which indicated his potential to become a master magician in service to the Hacker Ethic. As was the custom, Nelson had come a week early for Freshman Rush. He was a short kid, generally taciturn, with curly hair, darting brown eyes, and a large overbit
e, which gave him the restlessly curious look of a small rodent. Indeed, Stewart Nelson was sniffing out sophisticated electronics equipment that he could play on, and it did not take him long to find what he wanted at MIT.

  It began at WTBS, the campus radio station. Bob Clements, a student worker at the station who would later do some PDP-6 hacking, was showing a group of freshmen the control rooms when he opened a door that opened to the complex machinery—and found Stew Nelson, “a weaselly little kid.” he later remembered, “who had his fingers on the guts of our phone lines and our East Campus radio transmitter.”

  Eventually, he found his way to the PDP-1 in the Kluge Room. The machine got Stewart Nelson very excited. He saw this friendly computer which you could put your hands on, and with a confidence that came from what Greenblatt might call born hackerism he got to work. He noticed immediately how the One’s outside speaker was hooked to the computer, and how Peter Samson’s music program could control that speaker. So one night, very late, when John McKenzie and the people tending the TX-0 next door were asleep in their homes, Stewart Nelson set about learning to program the PDP-1, and it did not take him long to teach the PDP-1 some new tricks. He had programmed some appropriate tones to come out of the speaker and into the open receiver of the campus phone that sat in the Kluge Room. These tones made the phone system come to attention, so to speak, and dance. Dance, phone lines, dance!

  And the signals did dance. They danced from one place on the MIT tie-line system to the next and then to the Haystack Observatory (connected to MIT’s system), where they danced to an open line—and, thus liberated, danced out into the world. There was no stopping them, because the particular tones which Stew Nelson had generated on the PDP-1 were the exact tones which the phone company used to send its internal calls around the world, and Stew Nelson knew that they would enable him to go all around the marvelous system which was the phone company—without paying a penny.