Thursday, August 20, 2009

Over the Wall

The Bell Labs I came to in 1985 was a mammoth enterprise, spanning activities from truly basic research to advanced development, while other parts of the company (then AT&T) made actual products.

Each of these activities was centered in distinct parts of the organization. One of the amusing consequences was the idea that once basic researchers found something useful, they figured they just had to "throw it over the wall." Ideally, it would hit some developer on the head. After rubbing the bruise, he (rarely she) would notice the shiny new object and figure out to make tons of money.

That was the theory.

Toward the end of my career, I spent a lot more time with developers, for example at the factory in Orlando where they perfected the processes that in a few years would create the next generation of integrated circuits. It was rather embarrassing to realize that these folks were not hanging out next to the wall, waiting to be hit on the head. In fact, they were far across the field, barely visible, working on the problems that really mattered to manufacturing. Back at the wall lay hundreds of rusting ideas, with no one paying the least attention.

I'm afraid this is a common problem among academic scientists, as well. They get grants for research related to some economically important field, but often they never take the time to learn the real issues, and talk to the people that understand them. So they devise elegant solutions for exciting, unimportant problems, often the same problems over and over, while engineers do the real work. (The analogous problem in biomedicine is the need for "translational" research.)

A perennial example is ultrasmall transistors. Of course Moore's law demands that transistors get ever smaller. But smallness is not the end goal. Smallness is related to low power and to high speed, but these don't come for free; they must be directly addressed. Moreover, a small transistor is useful only if it can be packed very densely with other small transistors. A single, really small transistor every millimeter or so is not solving any important problem. Finally, making anything like a modern chip requires billions of transistors with virtually none failing. For all these reasons, making a single, isolated, slow, wasteful transistor is just irrelevant.

Of course, every so often one of those crazy, disconnected ideas changes the world.


  1. Don, you're absolutely right but also half wrong.

    As you say, any real breakthrough technology has hundreds of constraints that people in basic research don't understand. There is little hope that a brilliant idea will escape from a research lab and revolutionize any important technology by itself.

    But, who ever thought that science and technology really worked that way? That belief was merely a collective fiction to justify the existance of a research lab within a corporation.

    The reality is that anything that goes "over the wall" will need a lot of engineering, probably at least a decade's worth, before it becomes technology. Bell Labs in its later days didn't have time.

    The other reality is that many of the ideas that research labs come up with are really solutions waiting for problems. Until you find a problem that they can solve, they are useless. Only after you find the problem, they can be seen to be important.

    Looking at research as a way to get short-term advantage isn't a good idea. It's a long-term process and the benefits often turn up somewhere completely different from where the research was done.

    So, yes, the ideas went over the wall and the engineers couldn't use them. But they don't rust. They're seeds, not screwdrivers. Some of them will germinate, and then 30 years later an engineer will wander by and say "Wow, that's an odd fruit. I wonder if it's edible?"

    The metaphor should be that basic research is like Johnny Appleseed. He didn't make a profit, but he's remembered for his contributions.

  2. Thanks, Greg, good comment.

    I largely agree, and I think that spreading seeds is a good thing for some people to do. But I do think they should be clear about how their research fits in.

  3. Hi Don,
    I think that Bell Labs never figured out how to transfer technology...we had different time horizons, different languages and different reward systems then our developer friends. Other place were able to take some of the technologies and make a profit pretty quickly. We just had the wrong corporate culture to be nimble. - Karrie

  4. Thanks, Karrie.
    It is true that AT&T at the time was big enough that some developers had a "not invented here" attitude toward ideas from research in their own company. The Bell System was famous for its ponderous adoption of new technology. But when it was finally adopted, it did work!
    I'm sure in some technologies that are less mature than integrated circuits it's easier to make big innovations without knowing as much about how things are done today. But I wonder what examples you had in mind?

  5. There's an interesting book, "The Constraints of Corporate Tradition". It talks about how limited corporations are by what they think their goal or mission is.