1945, 1968, 1975-1990
We didn’t so much replace one form of computing for another insomuch as we added forms on top of each other. — Ben Thompson, “The State of Consumer Technology at the End of 2014”
The Computer Came Much Later
The machine later realized as the PC — a screen sitting on a desk that an educated person controlled with a keyboard to browse and transform information — was designed half a century before the Web became popular. That initial design did not have a computer powering its information processing. It came before before Turing’s vision of universality was made real in circuits, and years before Hopper invented the computational processing of words.
Vannevar Bush designed the PC in 1945. Though his design lacked universal computation, it illustrated the basic functionality that PC users rely on today, seven decades later. Bush asked us to imagine a researcher sitting at his desk. Machinery built into the desk presented an easy way to find an enormous variety of personal and published writings, photographs, sound recordings, and so on, and read them on a screen. The system included a keyboard for writing, a microphone for recording speech, and so on. An important and difficult problem inspired Bush’s work. The thousands of researchers he oversaw produced so much new knowledge that no one could keep track of it. A scientist himself, Bush considered how we manage information. He looked at the tools then available for moving words, images, and numbers around. He imagined linking them together in a machine that held a single researcher’s knowledge in one place. As the researcher worked: moving from one book to another book, writing up his thoughts on the books, perhaps watching a short movie, Bush’s device would automatically record what he saw, heard, wrote, said, and did:
any item may be caused at will to select immediately and automatically another. This is the essential feature[:] the process of tying two items together … when one [item] is in view, the other can be instantly recalled. … when numerous items have been thus joined together to form a trail, they can be reviewed in turn … It is exactly as though the physical items had been gathered together from widely separated sources and bound together to form a new book.
Bush wanted his design to reach many people. He published it to a general audience in a famous essay, “As We May Think”. He named the design “Memex”. It bore fruit only after his death, but on a wider scale than he probably imagined: both the PC on which I have typed these words and the World Wide Web on which you are reading them descend directly from Memex. A better name for Memex might be “research paper connector” or “idea tracker”. The engineers who built on Bush’s work returned repeatedly to the idea of a page, so we’ll offer
to replace “personal computer”. As versatile information machines, PCs have many uses, but Bush’s model of connecting pages remains central to what they are for today:
Wholly new forms of encyclopedias will appear, ready made with a mesh of associative trails running through them, ready to be dropped into the memex and there amplified.
That description foretells Wikipedia, perhaps the world’s single most important web site. Bush imagined Wikipedia not just decades before its birth, but years before anyone had used a keyboard to type into a computer, or a screen to read words from one. Since Bush designed the tool we call a personal computer without a computer in it, we should not think of it as being a computer.
From Code to Knowledge
Engineer Douglas Engelbart read Bush’s article while sitting on an island in the western Pacific Ocean in 1945. He devoted his career from the late 1950s through the early 1970s to constructing a descendant of Memex around a code runner. You can watch the famous 1968 demonstration of his invention here:
Engelbart focused on Bush’s conjecture that connecting knowledge should be the primary function of a versatile information machine. His invention, called Online System, made Memex’s associative indexing into an action performed by the user. Using Online System you moved between pages of information using a pointer controlled by a mouse. The page connectors we use today work the same way, and the people connectors work similarly. Online System also connected workers to each other by video conferencing, allowing them to collaborate on individual documents, anticipating the shared online presence that so many of us enjoy every moment. Engelbart’s invention, computer augmented knowledge work, defines the days of productive professionals half a century later. Online System was an extraordinary accomplishment.
As with the early bespoke contraptions, Online System was also the product of an organization, operated by engineers fully devoted to the machine for its own sake1, rather than to what the machine produced. Watching Engelbart’s triumphant demonstration, you see Online System produce … Online System, its documentation and its code. Online System wonderfully brought the power of Von Neumann computers to Bush’s vision of tools for managing knowledge. But while it matched the bespoke contraptions in the human and financial expense it demanded of its builder-users, unlike them Online System produced no important knowledge beyond itself.
The first Von Neumann computers served experts whose work consisted of numerical recipes, once they translated those recipes into machine language. That community had to make their work fit the computer. Hopper’s code made the computer perform some of the fitting, saving labor and expanding the community of numeric recipe experts who could use computers. Engelbart strove to empower professionals whose skills were not focused on calculation, such as those for whom Bush had designed Memex. Where Hopper’s colleagues computed results, Bush’s wrote prose and crafted images. Engelbart discovered how to implement the core mechanism of Bush’s information machine — associative indexing — with a code runner. He enlarged Bush’s design with online presence, a wonderful addition to civilization’s toolkit that billions of us enjoy each day. But Online System wasn’t actually useful for normal scientists, let alone normal people. Like Turing, Engelbart produced enormously important knowledge of what information machines could do without directly empowering a wider community to put that knowledge to practical use.
It took about a decade for Engelbart’s colleagues and successors working at the Palo Alto Research Center to extract from his expensive, opaque prototype a design useful to less esoterically trained workers with less munificent budgets. Recall that Bush designed Memex as a mechanized desk with keyboard and display. Engelbart put Von Neumann circuitry inside the desk: Online System was computerized furniture2. Given the training it required, Online System represented two steps forward from Memex but also one step back: real where Memex was merely a concept, but restricted to specialists where Memex was meant for a wide range of knowledge workers
Faced with the task of powering the work of desks and papers with a computer, PARC in effect swapped the choice that Memex and Online System had made. Rather than turn a knowledge worker’s desk into a complex computer terminal, they took a computer terminal and wrapped it in simplified pictures of a desk, files, papers, and other equipment with which millions of professionals employed in offices were familiar. Those workers already knew what a memorandum was, what a file folder was, what it meant to file a memorandum in the folder holding correspondence with a particular external organization, and so on. Rather than take over the physical space in which they did that work, PARC offered them an abstract but familar space inside a computer display in which to do it:
Xerox Star Desktop, courtesy of the amazing Digibarn computer history collection. CC BY-NC 3.0
They populated the space with little cartoons of office tools, especially papers and the folders used to store and organize them. Think of the Xerox Star as a
and of the little cartoons as
By reminding knowledge workers of the physical tools they had been using since before Bush’s time, these pictures hinted how to do the same work with PARC’s information machine, the Star. Online System, like code runners, required its users to learn and maintain a complex model of the machine’s state and possibilities in their heads. That training and ongoing effort was the price its users paid to to perform the associative indexing, or linking, at the heart of Bush’s vision. Bush did not specify how the links would be constructed in his design for Memex — if he had gone on to build a prototype, he would have had to grapple with the challenge. Engelbart’s made linking real at the cost of demanding expertise. PARC simply dropped linking from its design, and expanded upon the physical metaphors in Online System. The Star’s reminding pictures developed a full-fledged cartoon desktop to move the working of keeping track of the device’s information space out of our heads, instead leveraging our spatial memory: “Where is that letter I wrote? Oh, right, I put it ’inside’ the ’folder’ on the left … now I ’see’ it.” Bush’s researchers worked with letters, folders, and desks — the tools of 20th century offices — but Bush rightly defined their work around the fundamental capacities of the human mind. Star kept the surface of Memex, while discarding the heart.
A decade after Engelbart’s performance, the electronic circuits powering code runners had grown cheap enough that they sold for the price of a car. Steve Jobs, co-founder of a company that sold an inexpensive code runner, saw the Star and decided to build his own version. It took five years and two tries to create Macintosh. Macintosh directly adopted Star’s reminding pictures:
Macintosh leveraged the reminding pictures to show the few million people who had the underlying memories, the most affluent and literate members of the world’s most affluent countries, how to write their own documents. Its core use became crafting visually-oriented pages. This field, “desktop publishing” or “graphic design”, ironically reversed the focus of its grandparent Memex. Memex focused on freeing knowledge work from the constraints of paper, the better to further research. Macintosh focused on improving the appeal of printed3 graphics, to further marketing.
Macintosh stumbled onto marketing in search of, well, a market. Turing, Von Neumann, Hopper, Bush, and Engelbart all worked for society as a whole doing government-funded research. Jobs did not. His predecessors created astonishingly novel tools. Jobs did not. As a tool, Macintosh was a modest variation on Star4. It mattered because it brought computer-powered versatile information machinery to a much wider audience than its predecessors. To make that shift it had to pay its own way, which meant finding functions for which millions of people would spend their own money5. The code runner inside it made Macintosh a versatile information machine, but PARC and Apple had built it to work with pages. STAR had cost as much as a small home. Macintosh cost less, closer to the price of a cheap car, but that remained more money than most people would spend to manufacture pages. Advertising professionals comprised a concentrated minority of knowledge workers who became Macintosh’s core market, without which it would have failed commercially. Macintosh persuaded them to invest in it as a tool for crafting glossy promotions to run in magazines and newspapers Note the two-sided nature of the market here: marketers make information to persuade the many to spend their money; Macintosh persuaded marketers to buy it so they could sell more effectively6.
From that commercial foothold Macintosh extended itself into other markets in which its affluent customers purchased information goods: photography, video, and especially music. Page machines are versatile information transformers, and their versatility freed them to follow the money: first serving the most well-heeled creators of saleable information goods, then gradually becoming consumption tools as costs came down. We call these information goods “mass media”, and Macintosh called its versatility in working with them “multi-media”. Computing mattered only under the covers. In some countries today, customers prefer iPhone knock-offs that come with a built-in FM radio. They buy information machines for listening to music and news, not Von Neumann computers for running numerical recipes.
Early versatile information machines did rely on a market for numerical recipes, but a much more general market than Hopper had served. A few years before Macintosh, the falling price of circuits made code runners available for purchase by individuals. The first were bought by hobbyists — people who worked with the machine for its own sake. They immediately began crafting logic recipes, some of which turned out to be broadly useful. In 1979 Dan Bricklin released Visicalc, a pioneering tool for working with accounting-style tables of numbers. Bricklin’s spreadsheet quickly became a primary information machine for businesses, who as we have seen would shape the PC market, and therefore PCs themselves. By wrapping code runners in a tool for computing, Visicalc and its successors such as Excel made the PC a computer in the old, pre-Turing sense of a worker who does calculations. The computing an accountant performs with his spreadsheet does not look like the operation of a bespoke contraption. Rather than calculating the flight of an artillery shell or the formation of rain clouds, he creates a little world of quantified possibilities: how many gadgets do I have to sell at $9.99 to pay off a manufacturing run of 50,000? If I hire a distributor to persuade stores to stock them in lots of 20, how successful does she have to be to earn back her salary? Call Visicalc and Excel examples of
a family of information machines that provide a mental space for understanding arithmetical relationships. They have proven themselves the most popular tool we have for working with structured knowledge. Yet even with that quantitative focus, their name, “spreadsheet”, refers to a piece of paper.
When we use a keyboard to put numbers and letters into an information machine, the machine asks us to imagine that we are writing a page. The pages we create with these machines have become much of the world’s knowledge.
1. A lovely description of this relationship between tool and user comes from the creator of the Web: “At this time, a computer was still a sort of shrine to which scientists and engineers made pilgrimage.” Tim Berners-Lee, Weaving the Web, p.8
3. To this day you tell your email program to include another recipient by putting their address in the Cc: field. “Cc” stands for “carbon copy.” The literal meaning of that command is ”print these words on an additional piece of paper and carry it to that desk,” which is perhaps not the best analogy for understanding what is going on when you send an email. A prominent technology leader and author of many well-received books recently explained to me in conversation that messaging was not a meaningful advance over email and smartphones were not different in kind from laptop computers. This person currently leads an effort to update an important aspect of how the Internet is used by ordinary people.
4. Interestingly, PARC’s institutional status as a research/commerce hybrid mirrored Star’s status as a saleable prototype. Business-oriented writers are fond of mocking PARC for failing to commercialize page machines as Jobs did. I think this perspective is backwards: they should honor PARC for closing 90% of the technical gap between Online System and widespread use of versatile information machines. Of the remaining commercialization challenge, much was addressed by falling hardware prices between the late 1970s and mid 1980s.
6. Two-sided markets pop up repeatedly in computer history: eBay, Google, Facebook, and Uber led successive waves of the economic expansion of the Internet by creating marketplaces that brought millions of sellers together with millions of buyers. The structural parallels among computers, information theory, and markets run deep. While contributing to the development of the first digital computers, John Von Neumann co-wrote Theory of Games and Economic Behavior, which provided mathematical models that have been extensively applied to markets. In the second half of the 20th century economists working to mathematize the intellectual tradition of their field theorized that the price of information played a crucial role in the potential “efficiency” of markets. Their assertion is called The (First) Fundamental Theorem of Welfare Economics. For a brief discussion of the many roles information plays in economies, see Joseph Stiglitz’s survey “Information” in The Concise Encyclopedia of Economics. Insofar as markets and economies are built of information, information machines promise to remake them continuously and dramatically.