Software - Interviews with Experts

What should software firms be doing?

Cusumano: Software developers are special users with lots of technical expertise. Companies need to set up usability labs where teams that include ordinary people try to imagine what goes on in normal users’ heads. In the North American market, the automatic transmission made cars accessible to the masses. We need something like that in the software industry.

Are users overwhelmed by the increasing number of PC functions?

Cusumano: Software developers are special users with lots of technical expertise. Companies need to set up usability labs where teams that include ordinary people try to imagine what goes on in normal users’ heads. In the North American market, the automatic transmission made cars accessible to the masses. We need something like that in the software industry.

Devices like cell phones and MP3 players contain more and more software. Don’t their developers need new concepts?

Cusumano: Most of these devices have no keyboard and small memories, so they have to function autonomously. But we don’t really need new concepts for this embedded software. Take speech recognition. Some cell phones recognize spoken language. But speech recognition programs were ultimately developed for PCs.

When do you think we’ll have really intelligent computers?

Cusumano: I think we'll need a century to develop a computer that is even half as intelligent as the human brain. None of the intelligent software I've seen so far has impressed me. You still need a tremendous amount of memory and computing power to recognize even simple patterns, and neural networks also need large databases. A century is perhaps too long an estimate, but on the other hand, some of my colleagues believe we’ll never create a computer that imitates the human brain.

What about autonomous programs that write themselves?

Cusumano: The only thing that’s really impressed me recently is a self-diagnosis program that finds out why a computer has crashed. There are also intelligent agents based on autonomous software—but nothing really revolutionary. Do we need revolutions?

Cusumano: We recently had one—the Internet. And some people still call wireless technology or peer-to-peer computing revolutionary. Swap shops could also trigger a small revolution.

In your book The Business of Software, you wrote: "The driving force behind software development is not so much technology as business." What’s your view of the evolution of the Microsoft monopoly?

Cusumano: Don’t forget that this monopoly has made PCs cheaper and accessible to millions. But a monopoly is usually only the second-best solution. Nonetheless, we're dependent on Microsoft because of the many applications it offers.text

Cusumano: In the U.S. we have an enormous market driving developments. But software development is a global business today and will continue to be one. You only have to design software as modules and synchronize the development worldwide. IBM developed an operating system in eight different locations all over the world as early as the 1960s. Back then, you had to transport copies of tapes by plane. Today, the Internet is making exchange much easier.

One of the main reasons for the malfunction of the electronic toll collection system in Germany was the involvement of so many different software companies. Combining the modules simply became too complicated.

Cusumano: That could have been due to synchronization or architectural problems. Of course, a system developed at different locations is never as good as one developed by a single team. Development is always a trade-off between costs and manpower, or quality and speed.

Interview conducted by Jeanne Rubner