Labs!
I started reading about labs because I was interested in the history of the personal computer. And I kept reading, to immerse myself in that environment where everything was in flux and everything was exciting. Scientists, engineers, and businessmen built on each other to push the bounds of reality.
There are leaps in technology that define history: the telephone, the Internet, the computer, generative AI. In all these cases, technological brilliance played a role, as well as political and economic factors. Yet what I’m most interested in are the human ones — how labs became such wildly generative and satisfying places for work.
But first, some history :). Labs in the 20th century were institutions that conducted basic science research to create speculative technology ahead of their time. It all started with Bell Labs. AT&T formed an agreement with the government, which gave them a monopoly in the phone market. Flush with cash, in 1925, they established an R&D lab to further advance their product line. Prior to that time, new products had come from pure engineering. For example, Thomas Edison did not care about the math behind why his inventions worked. He was an engineer, who used trial and error to isolate useful materials and then combine them together. The same is true for the inventions of the sewing machine and barbed wire.
Enter Mervin Kelly and Clinton Davisson. Kelly would become the head manager of Bell Labs. He brought along his best friend, Clinton Davisson. Davisson worked slowly, refused to work on anything except basic science, and was often tired. But Davisson’s office was the most frequented at Bell Labs. He answered engineers’ questions of how telephone transmission worked, down to the electrons. For this, he is called the “Father of Basic Research”. More significantly, this was the start of marrying scientific research and engineering to produce products inside of a company.
With AT&T money, Kelly had his pick of researchers from top grad schools. One of these, William Shockley, would go on to invent the transistor. His advisor? Clinton Davisson.
The transistor enabled the multi-function chip, which enabled the birth of the PC, a computer that could sit in a person’s lap. Prior to that time, memory was stored in magnetic cores. These cores were very reliable, but they were both expensive and big. With the transistor came the Intel 1103 chip, used in the first version of the PC developed at Xerox PARC.
In 1970, the Xerox Corporation, then the biggest player in the photocopier market, established an R&D lab to ensure the continuation of their dominance. This lab was called Xerox PARC, and similar to Bell Labs, it had several labs within it. The two most notable were the Computer Science Lab, led by Bob Taylor, and the Systems Science Lab, informally dominated by Alan Kay. Taylor’s big idea, influenced by J.C.R. Licklider, was that computers should be interactive mediums for communication, not just tools for math. Noticing that humans processed the world through their eyes, Taylor believed this human-computer interactivity required enhancing the computer display. This was a contested idea back in the day! Similarly, when Alan Kay proposed making a portable computer so simple that kids could use it, he was laughed out of conferences. But Xerox PARC gave them the funds and latitude to develop their visions. The first version of what we would recognize as a modern PC was called the Alto. It came about because Chuck Thacker bet another employee he could create Taylor’s PC in 3 months and proceeded to do it. It used a bit per pixel. Two-thirds of its processor cycles and three-fourths of its memory went to the display. The other crucial first feature of the computer was the “character generator”, developed by Butler Lampson. This allowed users to type memos, characters on a screen. In parallel, also in the Computer Science Lab, Bob Metcalfe and David Boggs invented Ethernet. In the Systems Science Lab, Gary Starkweather invented the laser printer. At this time, customers were not interested in any of these products on their own (well, a few were interested in the PC but definitely none were interested in Ethernet). But, developed in parallel and put together, the suite of products was marketable. Business customers could now write a memo, see it on a display, and print it with the push of a button.
The end of Xerox PARC was the beginning of Microsoft and Apple. The PARC products failed because they were too expensive. With the increasing availability of computer kits and clubs (such as The Homebrew Computer Club), hobbyists could make computers for a third or a fourth of the price of PARC computers. They didn’t have the GUI or mouse capabilities of PARC’s computers, but they could sell them to the general public. Observing this trend, the main developer of PARC’s text editor, Charles Simonyi, went to Microsoft. Larry Tester, the co-creator of SmallTalk (the language used in PARC computers and the first object-oriented language), went to Apple. More and more scientists from Xerox PARC followed them to those two companies, incorporating what they had learned in the labs into the PCs we use today.
The condensed story of these labs makes everything seem planned, but the reality was more chaotic. And now we return to my original thrust — the wildly fulfilling culture of the labs and its paradoxical nature. In the labs, researchers went deep on their interests, they pursued what they wanted when they wanted to — and somehow it still mattered.
Their work mattered because there was loose direction set by managers of the labs, who still had to answer to a corporation. It was Mervin Kelly, the head of Bell Labs, who initiated Shockley’s research into the transistor. He dropped by Shockley’s office one day to make an offhand comment that it would be great if they could “get all of the relays that make contacts in the telephone exchange out of the telephone exchange and replace them with something electronic so they’d have less trouble.” This sparked something in Shockley, and he would think about it for years (starting before WWII, during the War, and after the War) before arriving at the solution of the transistor. It was an existing Labs project (to figure out long-distance communication) that excited Claude Shannon, who only worked on problems he found interesting. Shannon realized that all communication, from a conversation in the lunch room, to radio, TV, letters, and the phone, transfers signal from a transmitter to a receiver with noise in between. (And because this was the telephone, they literally meant noise!, and our phrases around data being noisy or finding the signal in the noise come from Claude Shannon). All communication transfers information. And all information can be encoded into bits. (The popularization of the word “bit” also comes from Shannon). Through digitally encoding phone communication, Shannon put his ideas into practice and cut the noise from long-distance telephone calls.
At Xerox PARC, Bob Taylor had a slightly different tactic. He hired scientists and engineers who were already working on the problems he thought were important for the computer revolution. He took Thacker and Lampson, who had been building computers at Berkeley Computer Corporation. He took Gary Starkweather, who had been ideating the laser printer in New Jersey, and made sure the rest of his lab welcomed and helped him out. He took Bob Metcalfe, who was finishing his thesis on an information packet transmission system, from Harvard. There’s a funny story about this, where Harvard rejected Metcalfe’s thesis because it was too practical and not “theoretical enough” for a PhD. Under Taylor, he ended up implementing his thesis at PARC as Ethernet. However, even when Taylor got them all together, his lab proceeded to first work on what they wanted (a time-sharing machine) and not what Taylor wanted. Taylor waited. He had hired these people for a reason, and knew they could not be coerced into building something. The waiting paid off. After the CSL Lab realized they could not build a decent display on a shared computer, they begun working on Taylor’s vision, a personal computer.
Former employees loved that their work mattered. John Pierce noted, “People cared about everything.” Once he became a professor at a university, he reflected, “no one can tell a professor what to do, on the one hand. But in any deep sense, nobody cares what he’s doing, either.” And “The great laboratories of the twentieth century had a clear purpose: Someone depended on them for something, and was anxious to get it. They were really needed, and they rose to the need.”
However, as partnership with a corporation gave their work meaning, separation from the corporation gave them freedom. This separation was enforced by managers of the labs, who fought valiantly (and sometimes threatened resignation) for their researchers. Since the labs were 5-10 years ahead of where the company was, scientists were allowed to explore research on a topic for years. Additionally, the drive to develop quickly came internally (from their own passions) rather than externally (since they were not addressing urgent customer needs). Alan Kay wrote that the two principles underlying Xerox PARC were “visions not goals”, and “people not projects”. PARC encouraged people who simply had to create, paid or not. And to create the future, the lab tolerated a high percentage of failures and even projects that tackled the same thing in two seemingly opposing ways. There was no centralized command structure. Kay writes it was
“Out of control’ because artists have to do what they have to do. “Extremely productive” because a great vision acts like a magnetic field from the future that aligns all the little iron particle artists to point to “North” without having to see it.
In this way, the structure of the lab, separated but inside of a corporation, gave researchers not only money and meaning, but also freedom. It created the conditions for satisfying work.
What about the culture cultivated inside the lab itself? After studying labs to form the Arc Institute (and also giving me a great book list), Patrick Collison theorized that it was transmission of culture by individual researchers that created progress. It was the culture of the lab that mattered, more than the resources they had or the educational attainment of the personnel or the population growth or the laws present at the time. I agree to an extent — but I wouldn’t say Bell Labs or Xerox PARC were culturally perfect. There were a lot of cultural practices that engaged scientists and engineers, but there were practices that disengaged them as well.
The real story of the transistor is that, after WWII, Mervin Kelly appointed William Shockley as manager of a lab to find a solid-state solution for vacuum amplifiers. At first, Shockley drove a great culture where people across the lab would, at all hours, experiment and share and give input on each others’ ideas. In this lab, John Bardeen (a theorist) and Walter Brattain (an engineer) worked closely together for over a year. Bardeen would think of a new material to try, Brattain would implement it. And they would iterate again and again. Finally, they did it. They created a point-contact transistor. Shockley proceeded to pull both into his office separately. He told them, “Sometimes the people who do the work don’t get the credit for it”. Shockley then worked for three days to build his own. In a subsequent meeting, as he watched another researcher start to reach his conclusions, Shockley leapt out of his chair, interrupted him, and started presenting his junction transistor to get there first. The team disintegrated after that. There were a lot of big egos in the labs of the 20th century. There was a lot of dismissal of anyone who could not keep up, which prevented good ideas from incorporation. Lynn Conway, who would go on to co-develop the chip powering cars and washing machines, recalled, “I always had a hard time dealing with Butler…He had this complete photographic memory of all theory that ever existed about anything, but sometimes that can be kind of a mental block to being creative. You can be so confrontational and challenging about how smart you are that you can’t always see that somebody else has got this cool idea.”
A generative aspect of the culture of the labs was small teams that built on each other. Many of the teams were two researchers working together who had different strengths (a theorist and an experimentalist, a visionary and a pragmatist, an extravert and introvert). Today, Bluesky (a Twitter competitor built in less than a year on a completely new protocol) also develops in teams of two. In the lab, each of these pairings was surrounded by others in their lab, who offered advice and opinions and built other related features in parallel. They were also privy to people dropping in from the nearby university and the exchange of ideas that flowed across the country as more and more communication channels were built.
Another interesting part was how much debate and disagreement was in the culture of 20th century labs. Oppenheimer, in the Los Alamos Lab, got Groves to allow a weekly colloquia where ideas across labs could be debated. In Bell Labs, the patent department conducted a study to figure out what made some employees win more patents than others. They found that the employees with the most patents all had a weekly meal with an engineer named Harry Nyquist. While Harry did not give them answers, he would ask them questions and get them to reconsider or further their work. In Xerox PARC, every Tuesday at 11AM was the Dealer Meeting. A researcher would talk about his work for an hour in front of the lab. Then they would go at him, guns blazing. Taylor’s lab loved nothing more than debate. To keep the conversation productive, Taylor would at times prompt the critic to state the opposing view — not to believe in it but fully understand it and align on assumptions. Once everyone had full knowledge and perspective, they would usually end up agreeing.
There were also huge advantages to having people from different disciplines work together. Hans Bethe, from the Los Alamos Labs, recalled, “Very often a problem discussed in one of those meetings would intrigue a scientist in a completely different branch of the laboratory, and he would come up with an unexpected solution”. In both Bell Labs and Xerox PARC, there was a lab for basic scientific research and a lab for applying that research to a system — yet most discoveries were due to unexpected cross-pollination of these labs. A researcher would come across a colleague’s work and become fascinated. They would realize they needed an engineer, or someone in a different field, ask around, and pull them in. And then…magic. Similarly, today at OpenAI, research and product engineering are in two separate orgs. Yet they credit their success to the tight integration between the two, having researchers be part of product engineering teams. After all, a lot of product questions are research questions (e.g. “How can ChatGPT produce more concise outputs?” or “How can ChatGPT produce more accurate answers?”). In addition, researchers (similar to the researchers in the labs of old) are often tinkerers themselves, and some of the prototypes they make can easily make their way into production.
What happened to the labs of the 20th century? We no longer have the great wars to keep some running. But what about the industrial labs? Why haven’t we had more explosions in technology in the 21st century despite having more scientists and more engineers? Well, maybe it’s as one venture capitalist at Kleiner Perkins put it, “We don’t fund science experiments.” Or maybe they’re yet to come. For good or worse, more companies are investing in basic AI research. My partner is in quantum computing, another burgeoning field. But more than the technology, there is an allure to the labs themselves. That they were places where ideas mattered, where creation ebbed and flowed, where you could always find someone ready to build or debate or listen to your latest thought. Sometimes, they were unpleasant places. Products didn’t work out, corporate management disagreed with lab management, egos got in the way. But the researchers that have been interviewed largely viewed their time as happy. They were scientists and artists and creatives, and they were given a vision, resources, and each other. Their work mattered and it has changed our world forever.