Oral-History:Robert Lucky

About Robert Lucky

Robert W. Lucky

Lucky joined Bell Labs in 1961. He worked on the T1 system, then, to his most enduring fame, invented automatic adaptive equalizers in 1965-66. In the 1970s he worked on programmable modems and network games, though these were ahead of their time. In the late 1970s he did significant research on digital switching, which AT&T finally adopted in the early 1980s. He testified at the court hearings of the AT&T divestiture, worked at Bell Labs for eight years after the 1983 break-up, then switched to BellCore. He discusses the people who worked at Bell Labs in the 1960s, the fields the Labs was investigating, mistakes he has made in his career, the importance to the field of the emergence of ARPANET (not associated with Bell) in the 1970s, and the shift from pure to applied research in the field following the AT&T break up, due to commercial pressures. He discusses his membership and positions in the IEEE and the IEEE Communications Society, and the organizations’ importance to the field. He hesitantly predicts the future of the field, with a due sense of historical contingency and the futility of prediction.

About the Interview

ROBERT LUCKY: An Interview Conducted by David Hochfelder, IEEE History Center, 10 September 1999

Interview # 361 for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, 39 Union Street, New Brunswick, NJ 08901-8538 USA. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:
Robert Lucky, an oral history conducted in 1999 by David Hochfelder, IEEE History Center, New Brunswick, NJ, USA.

Interview

Interview: Robert Lucky
Interviewer: David Hochfelder
Date: 10 September 1999
Place: Red Bank, New Jersey

Bell Labs, 1960s

Lucky:
I joined Bell Labs in October of 1961. When I first joined, the millimeter wave guide was going to be the big thing that transmitted information. It was a hollow pipe about 2½” in diameter that carried microwave signals. They were going to lay straight pipes along roads and rights-of-way. A lot of transmission people were working on that, and at that time they believed that that was the way things were headed. The Picture phone got going a little later than that. When I joined, I began working on data communication, which at that time was modems.

I went to Bell Labs because there were so many famous people there, and the environment there in the 1960s was very special. Shannon was supposed to be there, but I learned later on that he was not. He was listed in the phone book and his secretary would answer and say, “He’s not in today,” but in truth he was at MIT and never came around Bell Labs as far as I know. There were a lot of people that to me were really famous, like Steve Rice who was famous for his work on the statistical theory of noise. He inspired me. He was a little guy, meek, modest, one of the nicest people you could imagine. He always wore a sweater even on the hot days of summer. John Pierce headed a lot of the research at that time and was a real famous guy.

Shortly after I joined, the first satellite experiments were done, and of course Pierce was very involved in that. First they put up a balloon, and then it was the ECHO satellite, and then TELSTAR. Signals were sent, right up the street here at the Garden State Parkway, at the point you pass mile 115 at Telegraph Hill. It’s a famous place in communications history. At least three important things happened there. In 1927 Karl Jansky discovered radio astronomy there. That was the place where noise signals were first received from outside the earth. Then the first experiments with satellites happened there. Eisenhower’s recorded conversation was sent over ECHO from there. Then Penzias and Wilson won the Nobel Prize for the Big Bang experiment with their horn antenna perched on that same location.

Hochfelder:
Penzias and Wilson were part of Bell Labs. Is that right?

Lucky:
All those things were a part of Bell Labs, and that’s why they were right here. When I joined Bell Labs, it was primarily at Murray Hill and they were just building the facility at Holmdel at that time. I was one of the first people to occupy the Holmdel Building. I remember the day they burned down the original laboratory in early 1962. We sat at the windows of the new glass building and watched the old wooden building across the road burning down. The Bell Labs building here was supposedly the first all-glass industrial building ever built. It was designed by Eero Sarinen, who also designed the TWA terminal at JFK airport.

Satellites were just coming along and modems were just beginning to be built. The millimeter waveguide was the way long distance transmission was going to be done, and the primary traffic that was going to ride on this was going to be for Picture phone. That was a sort of picture communications technology at that time. The T1 carrier was also an important project in 1961. John Mayo and Bob Aaron were involved in that, and I’ve forgotten who else. Mayo was the project head on T1. That was a very important thing.

T1 carrier, digital transmissions

Lucky:

Of all the developments I just named, that was probably the most important – the T1 carrier.

Hochfelder:
The advent of digital transmissions?

Lucky:
Right.

Hochfelder:
Could you talk about that and explain why it was important?

Lucky:
I didn’t really understand digital even after I got a Ph.D. The idea is something you really have to intuit for a while to understand. PCM was invented in 1939 by Reeves of ITT, but it wasn’t something that was expected to become a primary way to carry telephone traffic because voice was intrinsically analog. I was at a meeting just a year ago when the head of a big telecommunications laboratory said, “Putting voice on packets is an unnatural act.” I think the same thing applies to putting voice on digits. It seems like an unnatural thing, because it expands the bandwidth by something like a factor of ten. You change a 3-kilohertz audio signal into a 64-kilobit digital stream, and expand the bandwidth by a factor of ten or twenty. You wonder why the hell you should do that, and the answer is because instead of having to go 3,000 miles you only have to go a mile and a half – because then you regenerate it.

Hochfelder:
Oh, right.

Lucky:
You can regenerate the digits; you cannot regenerate the analog. Instead of having to preserve that analog signal in the face of noise and distortion over 3,000 miles, you only have to preserve it for approximately a mile and a half. Therefore you can put more voice channels in the same bandwidth, because you don’t have to go very far.

Hochfelder:
All you have to do is recover the bits.

Lucky:
Right. All this is mired in old history. The old N-carrier had six voice channels on a pair of wires. The T1 carrier put twenty-four voice channels on it. Even though each channel consumed a lot more bandwidth, you didn’t have to go very far before you regenerated. The beauty of digital was its robustness, and its perfection was the fact that you could put it back together again periodically so that you didn’t accumulate noise and distortion. However, I don’t think people really grasped that at that time. The digital revolution hadn’t happened in our minds yet, and we didn’t have integrated circuits yet. We didn’t appreciate it. In 1961 it had been about fourteen years since the transistor had been invented, but it was still almost a vacuum tube world. Jack Kilby was doing his first experiments at Texas Instruments on integrated circuits at that time, and that was going to be the big thing. It took about twenty years for people to understand Moore’s Law and the economics of the digital way of doing things.

Hochfelder:
Were there any other significant developments in addition to digital transmission?

Modems; satellites

Lucky:

<flashmp3>361_-_lucky_-_clip_1.mp3</flashmp3>

Looking at things that were happening at Bell Labs in 1961, we didn’t realize modems would still be around forty years later. That’s funny, because many times we thought that modems also were an unnatural act.

Hochfelder:
Can you explain that?

Lucky:
Well, in the opposite way – taking bits and turning them into an analog thing also seems like an unnatural act. You are taking digital signals and turning them into voice-like signals so that they can put into a telephone network that then turns them back into digits. It’s just the opposite. Nevertheless they work, and work everywhere, so they’re still around. I never would have dreamed that forty years later everyone would have these in their houses and carry them on trips. Modems were something then that would be used for big computer centers. They weren’t personal things. I never thought I would actually own a modem of my own. And modems were what I worked on at that time.

The millimeter wave guide went away and the Picturephone died a horrible death. Satellites were another important thing happening at that time. In those days Bell Labs had the idea of low-flying satellites and thought geosynchronous satellites would never work. It’s funny after all these years to see that the low earth orbit satellites have come back, although they were probably a bad idea at that time. In the 1960s the most important thing certainly was the digital carrier.

ARPANET, Picturephone, and corporate histories

Hochfelder:
What about in the ‘70s and ‘80s?

Lucky:
In the ‘70s the most important thing was happening outside of Bell Labs, and that was ARPANET. It received very little press and very little notice at Bell Labs, which is painful. The way ARPANET was financed and organized and led, it was primarily a military-academic thing and lay outside the industrial research combines. There was also a bit of a feeling that it wasn’t terribly important. It certainly escaped the notice of most of the management at Bell Labs. There was a guy on National Public Radio earlier this year, supposedly some kind of an historian, named Jack Schwartz. He said that the government offered to sell the Internet to AT&T in the ‘70s, but that Bob Lucky turned it down.

Hochfelder:
Is that true?

Lucky:
Absolutely not. I tried to track it down, and I asked Vint Cerf and Larry Roberts and the people who were in the government running the Internet, and they have no recollection of anything like this. I don’t know where he got it.

Hochfelder:
Maybe there was another Bob Lucky working for AT&T at the time?

Lucky:
Not a chance. I don’t think the event happened, let alone my involvement.

Hochfelder:
You’re saying that the most significant development of the ‘70s was ARPANET.

Lucky:
Looking back, yes. In 1979, the answer may have been different. History is always interpreted retrospectively. Corporations are particularly bad about rewriting history. First, they have no sense of history, and second, a warped sense of how things actually happened. I was involved in editing part of the Bell Labs history, which was something like seven volumes. I edited one of the volumes, and I can assure you it’s complete bullshit. The corporate view of how things happen is absolutely myopic. The corporations like to think things were planned and charted and went according to management plans, but of course that’s not true at all.

Hochfelder:
A case in point you’ve written about in one of your books, I think. Was it Silicon Dreams? About the Picture phone.

Lucky:
There’s something about the Picturephone in that, but I’ve written about that often in other places as well. It would be very interesting to do a true history of some of these things. How was it that the Picturephone became a major development? It’s an interesting question not only from the historical standpoint, but also from a government policy standpoint. The way the world worked in those days, the cost of Bell Labs could be written off by the operating telephone companies and essentially charged to the American public through the regulatory authority, so there was a cost allowance for Bell Labs, a so-called license contract fee.

Hochfelder:
Is that the one percent?

Lucky:
Yes, exactly. In a sense there was taxation, but the way to spend it was decided by Bell Labs management. They decided to go to a Picturephone. Later on the question was asked – though no big deal was made of it – who authorized the taxing of the American public to develop the Picturephone. How did something like that happen? It’s the regulatory agencies, the FCC and the state agencies who had to authorize this, but it was obviously a loose kind of thing. How did they decide whether this was a good thing to do? The question of why the Picturephone, and why people thought it was a good idea was a very murky thing, and to this day its shadow hangs across all of video communication. Historically I think it’s a very interesting thing. I was a manager in those days, but I didn’t have anything to do with the Picturephone. Bill Cagle was the engineer on that, but he would probably say he was told to do it, and it would be very hard to trace the history of how this came about.

Communications theory; adaptive equalizers

Lucky:

The things that I am most proud of at Bell Labs go back to the ‘60s. It didn’t have anything to do with things we’ve been talking about, but the theoretical work, the foundations of communication. A lot of that work is still good today.

Hochfelder:
Could you give some examples?

Lucky:
Of course Shannon’s papers, which are still being interpreted fifty years later, are incredible. Whole theories of error correcting codes and data compression were created then. I did my work on automatic adaptive equalizers in 1965-66.

Hochfelder:
What is an adaptive equalizer?

Lucky:
An adaptive equalizer corrects the distortion of a telephone channel and enables you to send data faster. When I joined Bell Labs in 1961 the highest modem speeds were 2400 bits per second, and when I invented the adaptive equalizer it enabled a speed of 9600 bits per second. All modems today use adaptive equalizers. They’re built in now, and so sophisticated that very few people even know how they work. When you send a pulse over any kind of telephone channel it comes out distorted. You have intersymbol interference; that is, it interferes with pulses carrying other bits of information. Unless you pass this through some kind of adaptive filter that learns the channel characteristics and corrects for them you are limiting speed – because you have the self-noise of pulses interfering with each other. Therefore you adapt yourself to the channel characteristics and correct for them. I invented the first really workable adaptive equalizers back then.

Retrospectively, interpreting my own work, I made a couple mistakes, though I really started the field out, and it was certainly the best technical contribution that I made. One important thing I did not push hard enough in my patents was the idea of decision feedback. It was the idea that you could use your own decisions to improve correction ability. It sounds impossible. When you start transmitting data and the channel is very corrupted the bits you’re getting have a lot of mistakes in them. Nonetheless, you say, “Let’s assume these bits are correct and reconstruct what the perfect signal would have looked like with these bits. Compared to what the signal actually was, we’ll just try to make them look alike with our adaptive filter.” You’re basing this model on corrupted bits on the outside, so that the question is, “Is this ever going to converge?” The answer is yes, and it works damn well. What happens is that as you twiddle knobs trying to make your actual signal look like the ideal one, the data itself gets a little better. The bits start getting more accurate on the output, and what you are assuming to be ideal becomes closer to being ideal. The idea of using your own decisions as something to improve your performance – in spite of the fact that your own decisions may be bad – is really powerful, and it’s used today in lots of things. I should have paid more credence to that, particularly my patents and the way we did that.

My other big mistake, which I made as a young engineer, was that I was much too captivated by the mathematics. I thought I had a very clever solution to what was then called zero forcing equalizers. A zero forcing equalizer would twiddle the knobs on the adaptive filter and make the pulse shape have regularly spaced zeros in it – zeros at the times when the other symbols were to occur. Each pulse then ideally would not interfere with any other pulse, because its shape would be adjusted to pass zero when you expected to receive other pulses. I had a brilliant mathematical solution to optimizing this and doing it adaptively. It was cute as hell. However in my exuberance about the mathematics of this I gave little credence to mean square error criteria. The pulse did not necessarily pass through zero at the times when other pulses were expected to appear, but that its total mean square deviation from that was minimized. This turned out to be something more robust and easier to do – probably a better criterion to use. That’s the main thing that is done today.

The world was a different place then, a much smaller place, and it was easy to do things that were more significant. I used to have a Bell Labs phone directory from 1956, and when I left Bell Labs I willed it to my successor Arun Netravali, so I don’t have it anymore. Those old phone books are hard to come by and I’m not even sure that they actually exist anymore. They’re very interesting because if you look at the directory of research in Bell Labs, essentially everybody was famous or became famous. It was a small research area, and from the distance of twenty-five or thirty years you can see that this person won a Nobel Prize, this person won the Medal of Honor, this person became president of this university, another of that university, this person made a famous invention, and so on. Everybody knew everybody else, and if you could find a good problem like the equalization problem you could make your mark a lot easier than you can today.

<flashmp3>361_-_lucky_-_clip_2.mp3</flashmp3>

I made a good invention driving home one night while waiting at a red light in Red Bank, New Jersey. Sometimes I think about that when I’m stopped at that red light, because the idea just came to me in a flash. I don’t think very many inventions really happen that way, but this one did. I had the problem in my head of looking for a way to adapt this filter in a way to eliminate or minimize intersymbol interference. I didn’t have any algorithm where you could automatically adjust the filter to do this. It was a very difficult problem. The mathematics were very difficult, and I needed to find some algorithm that would automatically adjust an adaptive filter to do that minimization. The two problems, the mathematical problem and the problem of adjustment algorithms, came together in my mind for a hill climbing approach – to continue to adjust the knobs automatically to try to climb this hill to get to the top of the hill where it would be minimized. It all came together with the realization that looking at the problem in a certain way there was only one hill. You didn’t have to think there was a surface with a lot of little hills and that if you climbed one of them it would be the wrong one. In this particular problem there was only one hill, so if you adjusted the knobs to always go uphill you would eventually get to the top.

Hochfelder:
Okay. That makes sense.

Lucky:
This is what came to me while I was sitting at a red light. I also saw a really easy way of doing this adjustment so that you would always go up the hill – incredibly easy. I went home that night, and I just couldn’t wait to go to work the next morning. I sat up all night just waiting to run in to work and tell people how to do this. The next morning I went in to see Floyd Becker, who headed the development group that had been given the assignment of creating an adaptive equalizer. I described it to him on a blackboard, and he shook his head calmly and said, “Yes, that’ll work.” I don’t think he appreciated the beauty of it. He started making them and they were a great hit. They worked, and everything worked out really nicely after that. We made those first adaptive equalizers out of relays. We had hundreds of relays on a big rack. Digital logic was not cheap.

Hochfelder:
You’re talking about mechanical relays?

Lucky:
Yes, mechanical relays. It was great, because we had a rack of equipment about six feet high that was the adaptive filter and made out of about a hundred relays. You’d turn on the modem and hear all these relays click-click. What would happen is that when you first turned it on the clicks would be very sporadic, click, click, click-click, click, click, click-click-click, click. This was because the equalizer would be wandering around in parameter space with no idea where it was going. Then suddenly it would get an idea and get hot on the track.

Hochfelder:
It would find the hill.

Lucky:
Yes. All of a sudden you would hear click-click, click-click-click, click-click-click-click-click-click-click, and it would race up the hill. Then it would get to the top of the hill and go into an adaptive mode where it would just circle around the top of the hill trying to stay there, and the clicks would become sporadic again. It gave you a real feel for what was going on.

Programmable modem; inventions and timing

Hochfelder:
Are there any other technical contributions for which you feel pride?

Lucky:
I’ve only had a few really good ideas in my life. Certainly my most significant contribution was the equalizer. I had some other very creative ideas that were badly timed. I think engineers in general have a very warped sense of credit, in that engineers tend to give credit to the first person who thinks of an idea. In my experience, it is who thinks of an idea at the right time that really matters. For example, should we give Leonardo da Vinci credit for inventing the submarine? I don’t think so. I started a group back in the 1970s working on a programmable modem. My plan was that we would build a programmable single second order filter section, build it digitally, re-circulate the data through that, continually change its coefficients, and do everything that’s needed in a modem. I started people building that, and we actually built things that worked. It was a brilliant idea, because that’s the way all modems are done today, with digital signal processing (DSP) chips. But it was a stupid idea at that time, because technology wasn’t ready for it. Another example is that I became a great proponent of network games, and Hal Alles and I set about working on cooperative games where you saw the action from your own standpoint. That was way the hell ahead of its time. Now they are doing that kind of thing, but this was a dumb idea for the late ‘70s.

Most of us just sort of chisel away at things, and somehow the great events arrive out of all this chiseling, so that it’s not too clear actually who really created this great thing – like the Internet today. There are a lot of people trying to claim credit for it. There are a few people who played key roles, but you get a very murky story there. In my estimation, the Internet is almost a social invention. When ARPANET was created and when TCP/IP was written in 1973, PCs didn’t even exist. PCs had to exist and be sort of ubiquitous in order for the Internet to take off. What you really have is a concatenation of events. Other kinds of data networking paradigms could have triumphed, because there isn’t anything terribly special about TCP/IP or ARPANET, but somehow everyone got together and agreed, and it just sort of emerged out of a cloud and became the thing.

Optical fibers and millimeter wave guide

Lucky:

We heard one day at Bell Labs that Corning had made optical fibers that had 20 dB per kilometer loss. When the word went around the halls the response was, “Fibers? But we’ve got the millimeter wave guide.”

Hochfelder:
Is 20 dB per kilometer a lot of loss?

Lucky:
It is a lot of loss, but it was regarded as a benchmark loss. If you could get it at that, you could start to do things. Don’t forget that the T1 carrier has a regenerator something like every mile and a half. If you could get pulses over an optical fiber for a mile or so, you could regenerate them. That’s the whole thing about digital. With 20 dB per kilometer you started to think, “Hey, this is almost there, because I could go a couple miles and still detect that signal and then regenerate and start it out again.” Suddenly at 20 dB per kilometer it became a nearly practical system. The dam was broken when they did that. Now of course it’s a couple orders of magnitude less. When everyone started working on the fibers the millimeter wave guide died a pretty quick death. It was a very fast development between Corning making those fibers and putting in the first commercial fiber optic system, which was the Northeast Corridor from Boston to Washington. It was all multimodal fiber then.

Hochfelder:
Is that around 1980?

Lucky:
Yes, 1980 or 1981. It was done pretty quickly, and that can be what happens when people do really recognize something – they can rush in and do things rather quickly.

Digital switching, 1970s

Lucky:

Going back to the 1970s, I transferred from where I was in Bells Labs development over to research in 1976, and in my first job over there I was given responsibility for a research group that was doing digital switching. There had been a personnel blowup up in the Murray Hill Lab where the two people doing this hated each other and it came to blows. I was the new guy on the block in research and I was directed to take over the group and straighten it out. I didn’t know anything about digital switching, so I went to Amos Joel, and he said, “You shouldn’t take this job. Anything you could do digitally we can do better and cheaper with analog.” At the time what he said was true, but it wouldn’t remain so for very long, and that’s where a lot of people really missed the boat.

At that time the analog switches were made from remreed relays. They were little glass vials that had a couple of little metal leaves inside them, and an external magnet could open or close the leaves. They made them for pennies. There was a big factory outside Chicago that made these things. They switched the whole bandwidth. How could you possibly compete when the relays were made so cheaply? With digital switching you had to do all this processing, and every single line going in had to be transferred and changed from analog to digital just so you could screw around with the bits. We built a prototype digital switch. Hal Alles was the principal designer, and about 1976 we had a meeting with Bill Fleckenstein. He headed the development of the next generation switch at the laboratory outside Chicago. We gave him the story and he nodded his head and said, “Thank you,” went away, and that was the end of it. They went ahead with their analog switch. I do not blame Fleckenstein for that, although I think he did eventually take responsibility.

However, all of us failed to see Moore’s Law and understand that although analog may have been a bad way to do it in 1976, in only a couple of years the economics changed drastically. This gave an opportunity for a couple of AT&T’s competitors to build switches in their basement that could compete by being digital. It made AT&T mad for awhile, because AT&T would be saying, “They’re not as good,” and the competitors were saying, “Yes, but they’re digital.”

Hochfelder:
It’s like when products from the ‘60s or ‘70s were all solid state.

Lucky:
Yes, exactly. With some griping and disgust, AT&T had to go to digital too. That happened only a year or two later, and then switching went digital. First a metropolitan area went digital with a T1 carrier in the early ‘60s, and then in the late ‘70s switching – or at least the fabric of the switch – became digital. Then in the early ‘80s the long haul plant went digital because of fiber optics. The speed of the events was incredible. Optical fiber was made by Corning, and the next thing you know the Northeast Corridor was put in about 1980-81. Sprint began their “pin drop ad” on TV, and they made a big play out of, “We have an all-digital, all-optic network.” AT&T said, “This is sour grapes. We invented all this, and now they’re making a PR play out of it.” The very next year, around 1983, AT&T wrote off its entire analog plan.

AT&T antitrust trial, divestiture

Hochfelder:
Let’s talk about the breakup of AT&T in the ‘80s. What effect did that have on the research Bell Labs had been doing, on research and development in the industry, and on your own career?

Lucky:
One of the most memorable points in my career was the AT&T antitrust trial. I was on the stand for a couple of days. It was very interesting. The issue of Bell Labs didn’t receive a lot of press during the antitrust trial. My involvement was almost coincidental. One of the issues at the trial was anti-competitive behavior of AT&T, the Bell System, in failing to open the network up to foreign attachments. There had been several court cases. The Hush-a-phone case was the first one I think, and the Carter Phone case, where competing companies built products they wanted to attach to the phone network and the Bell System had forbidden them to do that. They went to court and both of those cases were won by the competitors. The Hush-a-phone was some kind of an acoustic thing that went over your telephone to give you more privacy in your conversation, and the Carter Phone was an electrical attachment. After the Carter Phone case, the Bell System was ordered by the court to make it possible for so-called foreign attachments to be put onto the network. You couldn’t connect a foreign modem into the Bell System; you had to go through these connector blocks. It was analogous to a fuse box. A fuse box just keeps you from sucking too much power out of the network, and the telephone adapter just kept you from putting too much power into the network. It was a little more sophisticated than that, but it cost something like $300 – and you didn’t need this if you used a Bell System modem. You could trust the Bell System, but since you couldn’t trust other companies to make modems that would not violate the standards of the network you had to go through this adapter which cost as much as the modem. This was ruled to be anti-competitive behavior.

The Carter Phone decision happened in the ‘60s. I don’t remember exactly what year. After this, Chuck Elmendorf was the vice president of engineering at AT&T and really in charge of all engineering for the Bell System. Elmendorf took up a charter or two to see how foreign attachments would be allowed to attach to the network, and formed a small task force of people to look at these things. I was a very junior person at that time, and was the person who was volunteered from the data area to go do this. It was early in my career, and I had very little influence on what they did. Elmendorf was ahead of his time in his thinking, and he was actually against the kind of couplers that they ended up using. He was for a much more liberal policy. I was on his little committee that looked at that. In 1983 when the antitrust trial took place most of the people that had been on that committee were retired, while I was still in mid-career, and I was asked whether I could defend what happened. Saunders was the chief lawyer defending AT&T in the antitrust trial. He was a very famous lawyer from Chicago hired from the outside by AT&T to head the defense of the case. I went in and met with Saunders in a hotel suite in Washington, and he had a legion of other lawyers with him. He started talking to me, and turned to his assistant lawyers and said, “We’re going to put him on the stand.” The assistant lawyers were saying, “You can’t do that.” Everyone was arguing about me, most saying that they couldn’t put me on the stand because it would be too dangerous. Saunders said, “No, I’m going to do it. I’ll take the responsibility. We’re going to wing this.”

As it turned out, I was the only witness that did not have a prepared written testimony in the whole antitrust trial. Saunders did wing it. He put me on the stand and for two days we talked, with Judge Greene joining in. We talked about Bell Labs and the role that Bell Labs filled in science, and so on. We started out talking about this coupler thing, this anti-competitive behavior, and Green got really mad. That was a scary place to be at that time, because Judge Green had such power, and he was a tyrant. About the data coupler he banged his desk and said, “Just like an engineer. You want to build a box to protect against everything.” He said, “That’s why we have laws.” He was absolutely right. It’s a tough thing. Basically what you need to do is trust people and say, “All right, don’t put too much power into the network.” Then every now and then we’ll catch somebody putting in too much power and we’ll hang ‘em. The fact is, most modems are made by reputable firms, and they won’t do that. This behavior happens in a lot of situations today. Everybody who has a satellite receiver has to have a decrypter. That expense is passed on to everyone for the few people that would otherwise rip it off. That kind of thing is often done in engineering, but Green might say there too, “You pass laws that people are not allowed to steal this content. Why put the expense on every law abiding citizen to protect against that?” The time I spent on the stand was a rewarding experience for me. The case itself never went to judgment. It was settled out of court by the Consent Decree. Later I asked Ian Ross, who was then president of Bell Labs, why we alone were allowed to keep the name Bell Labs when the divestiture happened. I said, “Obviously Greene had a lot of respect for Bell Labs, because he allowed us to keep the name,” and Ross said, “Don’t kid yourself. He just didn’t care.” They did create Bellcore at that time in order to give a Bell Labs presence to the operating telephone companies. I stayed with the original Bell Labs at AT&T and came over here about eight years later when they offered me this job.

That started a time of great change in AT&T, the Bell Companies, and in general in telecommunications. In the telecommunications industry, it was the singular event. What happened was that it created a tremendous competition, and stimulated technology and jobs. To this day my wife says, “Wasn’t it a terrible thing that Judge Greene did?” and I always say, “No, they really did the right thing by opening it up to competition.” There were times when I was angered about MCI being allowed to come in and skim cream. However in the end I think there was great wisdom in spoon-feeding the competition and allowing it to come about. Today the whole world is copying what we did. I recently heard one of the best thinkers in Europe talking about why Europe doesn’t have a Silicon Valley. There is a great tendency in Europe to protect jobs, but by protecting them they are losing out on new jobs that could be created. By protecting the jobs at British Telecom and keeping the monopoly, they failed to create more jobs outside. A great thing about Bell Labs is that everywhere in the telecommunications industry today, wherever you go you’ll find old Bell Labs people. There is a fabric that connects all of the industry together, and it’s the old Bell Labs people – the tentacles reaching out and the people networking in such a way, which will never exist again. It seeded the entire industry.

Hochfelder:
Bell Labs had a wonderful and deserved reputation for several generations in doing basic and applied research in communications. Do you think that level of research and development is going on in the industry now?

Lucky:
Absolutely not. The 1983-84 Consent Decree started revolutions in everything. One of the revolutions was that technology was given a tremendous spur through competition. Innovation really blossomed because of that, but at the same time a trend started that continues today, and that is the commercialization of research. With the passing of the old Bell Labs model the one percent license contract fee disappeared, and with it the blue sky thinking. That blue sky environment just doesn’t exist anywhere in the industry anymore. When the old timers get together now, the ‘60s are regarded as the golden years at Bell Labs. This may be partly because people who were there in the ‘50s are not around so much anymore. In the ‘60s, there was an atmosphere that science was everything and that we would try to do the right thing. When asked how research was managed Bill Baker once replied, “We hire the best people and let them do their thing.” That was what was done. The people who were famous there were respected for their scientific accomplishments. Whether or not they had any effect on commercial practices was never a consideration. There was respect for innovation, brilliance, mathematics, and there were artists in residence. Shannon was a prototypical person of that time period. He rode a unicyle in the halls, and after him Alan Berlekamp did the same thing. Ron Graham was a juggler and had a net in the top of his office so he could pull it down, put it around his waist and juggle balls without hitting the floor. In Silicon Dreams I wrote about the great contests with the mind reading machines. That kind of thing went on. You don’t find that today. All of that probably would have disappeared anyway, but the singular event that cracked that open was the antitrust trial, because that did away with the license contract fee and it changed Bell Labs into a commercial enterprise. Today it continues to move ever more in that direction, so that research in the old sense simply does not exist anymore. Is that a good thing or a bad thing? I don’t know.

IEEE and Communications Society

Hochfelder:
Let’s move on to the IEEE and your relationship with the IEEE and with the Communications Society.

Lucky:
I was president of the Communications Society at one point. The Communications Society has always been a kind of fraternity house for communication engineers. The amazing thing about it to me – and I appreciate this more as I approach retirement myself – is I see retired people like Amos Joel continuing to come to the conferences of the Communications Society. I often marvel at that, that people have these associations. I think it happens in technical careers that you make your friends and your connections early in your career, and those serve you throughout your career regardless of what directions you move on. It’s very hard to get into a different community. You may start out in communications and then decide to move to solid state or something else, but your friends will always be the communications people, and your own conference will be those CommSoc things or whatever. I think it serves a very great social and networking function, and I’ve always been very glad and proud to be a part of that. When I go to these meetings – although I don’t go to them all anymore – I feel this affinity with the people and the crowd. That’s a very nice thing, a sense of continuity there. The purpose of the Communications Society is really three things: they serve as a networking place, they publish, and they run conferences. Those became sort of big machines as time went along, at least the two main things of running conferences and publications. We created different journals as time went along. Probably the biggest thing was creating the magazine. I’ve forgotten exactly when that was done, but Communications Magazine started out as more the fraternity magazine whereas the s were basically – I didn’t appreciate it at the time, but I appreciate it more in retrospect – unreadable.

Hochfelder:
Very theoretical?

Lucky:
Yes. Since I used to be sort of theoretical, I was snobbish and looked down on people who said they didn’t read it, but now I’ve become one of them. I think many people don’t read it. The magazine started out more as a newsletter trying to promote the social part of this. Then the magazine started to carry articles, and then the articles started to get more content. It turned into a different kind of technical journal vehicle that was probably more suitable. It became more academic, and other newsletters were spawned. I was once quite involved, but not so much in the last decade. I went off and did IEEE things, and I was an executive vice president of the IEEE. I was vice president for publications of the IEEE two different times and editor of the Proceedings of the IEEE. I did a lot of IEEE things, and then I got into other things. I went through a National Academy phase and did things with the Academy, and then I went through a big Air Force phase and became chairman of the Air Force Advisory Board. When I got involved in these different things, the CommSoc went on the back burner for me for a long time. People like Steve Weinstein and Don Schilling were threaded in more consistently throughout all these years with CommSoc, and Mischa Schwartz to a somewhat lesser degree, since he wasn’t as political. Mischa was always the grand old man, even when he wasn’t so old.

Hochfelder:
Could you talk a little bit about the role that CommSoc played in helping develop communications technologies? You talked about the journal being very theoretical. Did you, in the course of your work at Bell Labs or Bellcore, find your association with the Communication Society very useful in your career and the technologies you worked on?

Lucky:
It was an essential part of the environment, the professional part of the environment. I think that may be changing now because of the worldwide web. In those days you didn’t have many ways to get information and meet people outside your own company other than the professional societies. That was the way you found out what people were doing around the world and you met people that served as a network. Today I think that the web does the same thing in other ways. Everything is a lot more visible, and there are more connections, contacts, networks outside. If you go back even just ten years ago, CommSoc was basically the main way of finding out what was going on. There always were other ways, because there always were other journals and other kinds of meetings where you would meet people, but CommSoc was fairly unique in its maintenance of a neutral and professional meeting ground. It was essential. As to the role CommSoc played in the development of communications, it was a catalyst and no more. People have said that CommSoc led this or that development, but I don’t agree. It’s performed a very important catalytic action, but the events had to come from elsewhere. People at CommSoc don’t sit around a table planning and designing things. Standards do play a very important role in the development of technology, but CommSoc was never really big in that.

Information theory and communications engineering

Hochfelder:
Okay. That makes sense. In Silicon Dreams you talked a lot about information theory. I’m curious about the connection between information theory and communications engineering.

Lucky:
Information theory is a beautiful ideal. Shannon’s capacity equation hangs over much of communication theory work. There has always been a dichotomy between the Information Theory Society and the Communication Society. They tend to divide into two distinct groups socially. I was chairman for one information theory annual symposia. There was a different crowd of people at the information theory meetings than at the communications meetings. In fact, that’s probably even more true today. While there are industrial people that contribute to the Information Theory group, particularly at Bell Labs, it’s largely an academic field. Certain disciplines come along and have to find their natural home. Error correcting codes came along and their natural home ended up in information theory. When optics came along physics was its natural home, and some of it went in a physics way and some of it went over to communications. Things like data compression that could go either way, but the home usually turns out to be where the main people first publish. That becomes the place where other people want to do their thing, so that really determines it.

The CommSoc really lost out on data networking. They never were the place to be for the Internet things, and they got to it very late. If you go back even three years, in the major conferences of CommSoc you won’t find anything about the Internet. Not only did big companies miss this, but CommSoc missed it. What happened was that the early workers were not part of that community. They published elsewhere, in the computer magazines. The people who have threaded their way through CommSoc for many years are running sessions on things they know about, and there is a self-perpetuation. It is sometimes hard to bring in a new group of people. You can’t just invite the Internet people in and say, “Come to our conference.” They’re not going to do it, because they are going to go where their friends are and where people are going to listen to what they have. CommSoc would go off on other things. For instance, ATM was theirs.

Hochfelder:
ATM being asynchronous transfer mode?

Lucky:
Yes, although ATM is ATM. It is asynchronous transfer mode, but nobody knows that, and it’s so irrelevant that ATM is just ATM. It’s a cell-based transmission protocol, and it’s called ATM. To say it’s asynchronous transfer mode, you know less than you did before. That is not at all descriptive of anything. ATM and ISDN were creatures of CommSoc, but TCP/IP was a creature of the computer people.

There are only a few people that go back and forth between the Information Theory Society and the Communication Society. There have been some big crossovers where information theory has really affected communications in a very practical way. There have been times that the fruits of what started out as information theory became adopted by the communications people and really been taken to heart. An outstanding example is the Trellis code of modulation. It’s now basically used for everything, and was started by Godfried Ungerboeck, who should be on your list. Godfried has made a very brilliant singular discovery of what came to be known as Trellis code modulation, and it is a tremendous improvement.

Hochfelder:
What is Trellis code modulation?

Lucky:
When you send digital information, people talk about having signal constellations. When you have a modem it has a fixed constellation of signals that it can send. Each point of the constellation represents an amplitude and a phase and it represents a couple bits of information. Basically today’s high speed modems, such as a 28.8 modem, send points in amplitude phase space. These points are in a constellation and they are on concentric circles. Ungerboeck’s idea was that as you went along for consecutive signals that the constellation itself would change, and that there would be a coding that would be superimposed on the constellation. I don’t know a good way to describe this, but basically what it did was it dramatically achieved about a 6 dB gain. This was an incredible thing, because everyone thought there wasn’t a tenth of a dB left to be gotten – and he found 6 dB or so just lying around. It really made a tremendous improvement, and everybody uses that. There have been a number of things, which have moved over from information theory into communications, which were not always understood at first by communications people. Someone will make that transition and push them over and then it will be taken up.

Hochfelder:
Translate the ideas?

Lucky:
Yes.

Technological innovations and social impact

Hochfelder:
What do you think the future of communications will hold? You wrote a little bit about that in your October 1997 article in Proceedings.

Lucky:
I don’t like to predict the future these days, because I think it’s unknowable. We’ve screwed up so many times and gone totally bankrupt by trying to do this. Starting with the Picture phone, ISDN, home information systems and video-on-demand, every vision of the future that has compelled the industry has been wrong. I think there’s a chaotic phenomenon at play here. A lot of people disagree with me, but I’m old enough I can say what I think. There’s a chaotic element at play that makes some of the very important things unpredictable. I do think that technological advances are predictable, because they follow their own Moore’s Laws. Optics is following a Moore’s Law, wireless is following a Moore’s Law, and all those are amazingly predictable, but the big things – the worldwide web and what people do with communication – are social discoveries that can go one way or the other. A butterfly flaps his wings in Brazil and you get a web – or not.

Hochfelder:
There is a certain randomness, a contingency of chance.

Lucky:
Yes, because the raw technology offers all kinds of potential for what people and businesses can do with it. Some businesses will flourish and some will die. I heard a speaker recently, one of the top managing executives at Microsoft, saying that the Internet industry today is a copycat industry, and I think he’s right. He said, “Nobody knows what works and what doesn’t work, but they all look around and see what different companies are doing and then when someone achieves a success they say, ‘Hey, this works. Let’s all do this.’” eBay.com comes along and, “That works, so let’s go do that.” The big things going on right now are wireless and the Internet, a sort of a combination of those things. In the telephone industry there is the Next Generation Network (NGN) thing going on which is really the packetizing of the telephone network. Those are the biggest things that are happening. I just made my own list of the ten most interesting things happening. At the moment I have appliance networks at number one.

Hochfelder:
Appliance networks?

Lucky:
Yes. Everything being connected to everything else. Everything online – all kinds of sensors and cameras and gadgets. This has nothing to do with what’s important, but the most interesting things happening right now. Wireless data is high on the list, and NGN is going down on my list as being not quite so interesting as it once was. Virtual private networks for corporations is high on my list. Self help for customers barely made the list at number ten. Home networking is becoming a very big thing. It will be interesting to see what happens when you put local area networks in homes. Broadband access is slipping down the list because it’s already starting to happen. Those are some things now, but in a year it will be a different list.

Hochfelder:
Certainly in five years.

Lucky:
Yes. Things that are already happening are wireless data, broadband access, the merging of the telephone network with the Internet and embedded connectivity. What they will mean in a global context remains to be seen.

Hochfelder:
It sounds like you’re saying it’s easier to predict the hardware and how the hardware will develop than to predict the essential impact of the hardware.

Lucky:
I don’t mean just the hardware. I would include the technology itself, which includes software and protocols and architectures and things. Speech recognition gets better every year, and biometrics are coming along. All these things are predictable and they are problems that we’ll never completely master but will get better and better. We just don’t know the social uses of these developments. We know that we can do video telephony, but we can’t anticipate what people will do with it.

Hochfelder:
Thank you very much.