(FORTUNE Magazine) – On May 1, Nobel laureate and Bell Labs chief scientist Arno Penzias retired after 37 years with AT&T and Lucent. The day before he signed off, he joined a group of FORTUNE editors for lunch and gave them a taste of the wit and keen intelligence that helped make Bell Labs preeminent.

In retirement, Penzias, 65, will continue to work out of his home office on Telegraph Hill in San Francisco, where he and his wife enjoy a view stretching from Lombard Street in the west to Alcatraz in the north to the Bay Bridge in the east. He'll scout new companies and technologies for Lucent and for New Enterprise Associates, a venture capital firm (FORTUNE columnist Stewart Alsop is a partner). Penzias--who won the Nobel in physics in 1978 for his discovery with Robert Wilson of background radiation emanating from the big bang--says he wants to be a "tech grandfather" to up-and-coming startups looking to turn a concept into a business. "I don't yet have that Silicon Valley sense of what's hot," he says, "but I can go in and look under the smoke and see what they really have." As the lunchtime interview that follows amply demonstrates, startups would be wise to pay attention.

How has Bell Labs' culture changed under Lucent?

Bell Labs' culture started changing in the late '80s. In the old Bell System, AT&T's research was an organization that was supposed to be the crown jewel. It was aimed at a company that was doing this almost as a kind of a public service. I came to see that this was untenable, so we changed and gave people technology responsibility in management--managers now have to start looking sideways, not just looking up and down--and we tightened our connection to the business units. A lot of the resulting products are coming to market now; Lucent TrueWave fiber is one example. We're a clear market leader in fiber now.

What do you think about the way telephone networks and data networks that rely on IP (Internet protocol) are merging?

Right now IP has got a big problem because you can't do real-time network management. That's the unfortunate thing. The typical router today can give you some information about what's going through it, but it can't help you figure out what's gone wrong until you have some sort of stoppage.

The new "router switches" that are coming handle cells, frames, and packets as a mix. Not only that, they will tell you what your customers are doing, and they won't charge you a lot of machine overhead for it.

Owning a telephone network is a license to learn about your customers. The important thing is the management of information rather than the movement of the bits. Walter Wriston once said something like this: "The time will come when the information derived from a financial transaction will be more valuable than the execution of the actual transaction itself." The same thing is true in our business. I keep talking about two telephone companies: One has the world's best network and provides service; the second knows what the customer wants and hires the first one to provide just that. Who's going to make the money?

The shortcoming in the IP business is that, in a so-called dumb network, nothing knows anything about anything else, and the only way you find out anything is wrong is if it crashes. Well, this is going to be replaced by a new and more robust protocol system, one where IP becomes more predictable.

So will tons of people start making their phone calls over the Net?

It's a little, cute business. This is the breakdown: There are people who will spend time to save money, and folks who will spend money to save time. Those are two different markets.

I remember years ago when I was a graduate student, we used to get Consumer Reports. This one guy wrote in to the magazine saying he took umbrage with the fact that they said that razor blades should only be used 20 times. What he did every time he shaved was, he'd take his double-edged razor out of the shaver and carefully wash it in hot water. He would then run it carefully on the ball of his thumb on both sides, invert it and put it back in the razor. And he was able to get 70 shaves per blade.

The editors wrote that in Mr. Jones' case, a penny saved is really a penny earned. I'm sure if Mr. Jones is still alive, whether he's shaving himself or not, I'm sure he's using the Internet for his telephone service.

There is a certain market, but I think it's very small. If it isn't, why has the Internet fax not taken over? Fifty percent of all the overseas call units are faxes. This is tens of billions of dollars. Lucent Technologies makes a neat machine that you can put next to your PBX, and it will fool a fax machine into thinking that it's talking to another fax machine. It's slick as a whistle, a product that can save $10 billion a year, and no one's using it.

What is the biggest development now in computing in general?

Right now, it's probably the emergence of networking. We're moving into an era where the computer is beginning to disappear as an entity. You have computers that will cost no more than postage stamps very soon and will be just as ubiquitous.

Today we're in a trend where networking will begin to have a negative cost of quality. What that means is that it will become cheaper for a manufacturer to make something that networks well than to make a bad stand-alone product.

An example I gave in my last book is a watch motor. A watch drive has a battery, a battery holder, a calibrated quartz crystal, an oscillator circuit; it's got an [analog-to-digital converter]; it's got a counter, a divider, a [digital-to-analog converter], an amplifier, a stepping motor, reduction gears, case, and assembly--the whole thing for a buck and a half. If it wasn't working perfectly, some engineer down the line would have to go and tweak it, and the cost would explode. It's only because absolutely everything works that it can come together.

In the same way, adding networking to an appliance reduces its price. The Pilot doesn't need anything more than AAA batteries, because you back it up on a PC. That makes it cheaper to the manufacturer. It also doesn't need a keyboard or fancy display because the fancy keyboard and display and all the other stuff is off on another machine.

What about network computers?

I think you'll see the network computer and the PC in different spaces. Java is not a substitute for bloatware. For order entry, for those kinds of things, I think a network PC could be very cost-effective.

We're going to see a very healthy ecology of networked devices. It's the opposite of a golf green, which is the sickest of all the possible ecologies because one thing and one thing only is allowed to live there--and a few token earthworms underneath, as long as they don't stick their heads up and ruin somebody's putt.

If we asked you to look out about 20 years into the 21st century, what interesting things are going to happen that we're not seeing now?

We will live in a networked world, so, for example, I think that a lot of the things that we would normally do--waiting in line, for example--will be changed. So let's say you're carrying the communicator you carry on your body all the time. On the way to Starbucks, you'll let them know you're coming, so the coffee will be ready by the time you get there. That's a trivial example.

A much better example would be transportation. Within a few years, vans will take over a large part of public commuting. If I wanted to start a business and had enough money, I'd buy one of these shuttle companies and just put a GPS receiver in each van and link it to a dispatching system. Then you could say something like, "Hello, Mr. Matthews. You'll be home in 22 minutes, or your ride is free."

The scary impact of all of this is the disappearance of privacy. I think anonymity will disappear.

I now go into a local dry-cleaner, and I like the fact that they know who I am. But what they could do even now, although it would cost them hundreds of thousands of dollars, is to have a video camera every day and just take a picture of everybody that comes in. Then a simple imaging recognition system could identify a few things about each customer--like gender, height, hair color--and sort them.

Let's say this gets cheap. Then what happens is, you go in, the recognition system compares your image against the ones it stored earlier, and picks the most likely matches. A screen behind the counter shows the 15 faces most like yours. The clerk picks the right one and says, "Hello, Mr. Matthews." Or if he can't pick exactly the one, he could say, "I don't remember your name, but were you wearing a great-looking gray sweater the last time?"

What else do you look forward to?

Another thing that's going to happen is that we're going to do a lot better in medicine than people suspect. The biotechnology is going to become a lot better. Today, the best molecular design technology can simulate molecules atom per atom. You can take a molecule, scan it, and look at the exact shape you want, and if it's smaller than 100 atoms, you can actually design it. You might, for example, want something that works like nicotine. There is a nicotine-like molecule that a viper in Sumatra uses in its venom. Nicotine locks up a gate that would otherwise allow pain signals to go from one nerve cell to another. So you feel better, but the molecule only shuts the gate for a while. The viper has a slightly better locking mechanism. If a few of its molecules get in your body, nothing works, and you shut down totally. So what a designer really wants is something in between [the viper and nicotine]. Something so people feel just as good as with nicotine without having to get nicotine from a cigarette fix. That can lead to all kinds of feel-good drugs. That can lead to solving diseases.

So that raises the possibility of life-extension products as well?

That's beyond my headlights.