ERIC HASELTINE: I'd like to get started by asking each of you a two-part question: What do computers do today that engineers used to do by hand? And what's the most exciting thing going on in this field?
TERRY HENG: I'd have to go back about 20 years, when I was part of the team that worked on 4-bit microprocessor design at Motorola. In those days, it took us almost 12 months just to lay out that processor. Today you could do that in less than 10 minutes. So the tedious task of having to do all the drawings, all the layouts, all the mess, has been superseded by computers. The most exciting thing going on in the field of computer design, for us, is that we're in the midst of trying to revolutionize the automobile. I think the cars you're going to see tomorrow will be much smarter. A BMW in about five years will have something like 150 microprocessors.
MICHAEL HAWLEY: For me, it's movies. Regardless of what you think of the latest Star Wars movie, every frame is a visual work of art. But it's more than just art. It's a window into an ungodly amount of computation and engineering innovation and talent. All that stuff used to be done by hand, laboriously. I think what's most exciting about it is that probably more than any other driver of demand, motion pictures are forcing machines to have eyes to understand sound and image. They're pushing those two sensory aspects into mainstream computing.
BOB LUCKY: I'd like to ask Mike a question. You recently won the Van Cliburn amateur piano competition using your fingers. Is a computer going to replace your fingers? Why do we have to worry about people being able to move their fingers like you can, very dexterously, to produce great music?
HAWLEY: I'm actually the wrong guy to ask about that because I'm the one who once tried to enter a computerized player piano in the Tchaikovsky Competition, because you could apply by a tape audition.
LUCKY: And what happened?
HAWLEY: I didn't make the cut. But you know, Bach used to say that it's not so hard to play the organ; you just have to push the right button at the right time. It's tantalizingly computable, and yet . . .
LAWRENCE BERNSTEIN: I think the biggest thing that has affected our lives over the last 20 years is the use of computers in structural design. In Japan you can actually stay in a building and not worry about earthquakes. Their design of building safety is that good, thanks to computers. I think the ability to explore DNA, RNA, and proteins with computers is exciting. I think we're on the verge of another explosion in bioengineering, which is the future for the next 10 years.AL AHO: What is the first thing you do when you get a computer today? You connect it to a network. And what makes computers interesting is that they talk to one another. I think the biggest change in computing is that we're turning the world into a global village. We can have real-time conversations over the computer networks of today. I think the marriage of computers and communication is the biggest thing I've seen in the last couple of decades. As to the future, one of the biggest questions to be answered is whether people are biological computers. If they are, then someday people can be simulated by Turing machines. If the Church-Turing hypothesis holds, maybe we'll be living in a world in which people and machines are interchangeable.
HASELTINE: You mean, not only will computers replace engineers as engineers but as people?
AHO: Yes. Sometime in the next 20 or 30 years, we're going to have, because of Moore's law, machines that will have the computational power and memory of humans. But we don't know how to program them yet to interact naturally with people. So it's all a software problem. There's one other aspect of computers that is worrisome: the large amount of faulty software that gets produced every year. Let's assume there are 10 million programmers in the world, and they're producing 10,000 lines of software each. That's a lot of software. But we don't know how to produce defect-free software. And therefore it has to be maintained. Who will maintain the billions of lines of new software generated each year?
JEFF HARROW: Well, I'm going to take a contrarian attitude. I don't think computers are anywhere close to replacing engineers. Computers do the scut work, the large-scale heavy lifting. Computers are good at memorizing, doing repetitive things. But they don't have creativity. Somebody has to instruct them as to what to do, where, and when.
HAWLEY: I don't know. Maybe it's easier to get away with creative murder in the arts business, but we've had synthetic music accompanists, software composition algorithms, software performers . . .
LUCKY: You know, John Pierce wrote an article for Playboy magazine many years ago entitled "Portrait of the Machine as a Young Artist." He postulated that in the future, computers would produce all the art and music but that someone would have to pore through it and decide what was good and what wasn't, and that person would eventually become known as The Artist.
HASELTINE: What is creativity, Jeff?
HARROW: Creativity is coming up with something that is not evident from what's in front of you.
LUCKY: Arthur Koestler wrote a book entitled The Act of Creation in which he got into this. It's the unexpected thing that comes out of seeing something in a different light, the intersection of two disparate fields or two disparate thoughts. Jokes are like that. There's a punch line. There's a twist. But I think there's a stage that comes before creativity. Doug Lenat, of Cycorp, has a Web site about that. He has been trying to develop rules for commonsense reasoning, so that computers will have those rules, like: You have to carry a glass of water with the open side up. Everybody knows that, but computers don't. So he's got about 60 people just reading newspapers and taking every sentence and trying to figure out what you have to know to interpret a sentence. They've developed something like 3 million basic rules. You can go to their Web site for it, and they have a whole grammar of these rules for common sense. I hope one day—they've been working on it for 15 years—they have all the rules. But when you don't know how to carry a glass of water, it's a lot to be thinking about creativity.
HASELTINE: Jeff, you didn't get to tell us what you think is the most exciting thing happening in computer design today.
HARROW: I think it's the use of computers outside the computer field. A great example, I think, is computerized surgery. We have machines now that not only give a doctor fine control by scaling down his movements but also filter out tremors, so that a surgeon can work at a finer level of detail than he could otherwise. This also lengthens a surgeon's ability to help others long after nature has unsteadied her fingers.
HASELTINE: Nick, what are some things today that computers are used for in design that only engineers could once do?NICHOLAS DONOFRIO: Number one, you can't build computers today without computers. You can't design the systems, you can't design the components, and you fundamentally can't design the software. So it's a productivity tool. I think in the future, the computers that we see are going to continue to learn from other spaces and other places. They're going to learn from us. They're going to learn from life sciences and biological sciences: How we put these things together, how we organize them, how we automate them, how we make them autonomic—these are important challenges. The industry has built such an incredibly complicated mishmash of things that we can barely manage them.
HASELTINE: Are you talking about what Neal Stephenson was getting at in his books on nanotechnology, where you take a simple seed of something and the computer grows and programs itself?
DONOFRIO: Even before you get there. We need to build computer systems that are literally self-installing, self-diagnosing, self-healing. If we don't do that, then we'll just bog down. Technology will continue to explode, and we won't be able to provide real economic value for anybody. The things I enjoy watching computers replace, in terms of tasks, are right at that interface between the digital world and the real world—simulations. We can build simulated cars, and we can test them without ever having to actually build them or crash them into real walls. I like this whole idea of simulation for nuclear nonproliferation. I know it's got a million issues, and there are a million people in favor of it as well as a million people against it. But the fact of the matter is, scientists can now think of their experiment as something that can take place in the computer instead of in the lab.
LUCKY: I'll be a contrarian. People have talked about how computers are pulling us up the steps, so let me talk about how they're pulling us down. What's really different is, 20 years ago, when I wanted to do a graph for a presentation, I just went to the art department. Now I can do my own graphs, and I can do my own typing, and I can do all my own travel reservations. So my secretary has moved up and is now doing real thought work, and I've moved down to doing all the secretarial work. That's changing the way engineers think and talk. We're starting to think in bullet points, to speak in PowerPoint. How much of the day now am I really just spending typing and figuring fancy ways to use PowerPoint? That is pulling us down.
HASELTINE: But there's another aspect to this. I remember about eight years ago, when I first came to Disney, my boss said, "Design this lens." I had never designed a lens before. So I got this program called Zemax, and I basically said, "Well, here's what I want as an output." I had no idea about Seidel aberrations or third-order theory, let alone Fourier optics or any of that stuff at the time. I just knew I wanted to get so big an exit pupil and so much field of view. And I push a button and poof, the lens was designed. All of the specifications were there. And I didn't really know anything about what I was doing. Those things are now all over Disney's parks. The guests don't know that a computer designed those lenses. And that says to me that you are going to get to a point where you really don't need engineers, just someone who has a very general idea of what they want to do.
BERNSTEIN: Don't kid yourself. The machine wasn't an engineer, and neither were you. You were being a technician.
HARROW: I suggest that if we ever get to the point where there's nobody who understands what that computer did, we're in deep trouble, because then we'll never be able to make any additional moves forward. Eric was able to effectively use a tool to build a lens, but he can't build the next new type of lens or the next new optical system with the programs stored in a computer. We have to have people who understand the absolute basics from the bottom up. Engineers are absolutely necessary.
HENG: These comments bother me a little bit. They presuppose that the world is hinged on humans. You look through the history of humans—we've gone through wars, we've bumbled through life. Maybe computers can take off in a different direction, with a different intelligence that is more predictable and more peaceful. We have allowed humans to evolve over thousands of years. I think this is part of human arrogance. We are constraining computers.
AHO: Bob, you were starting to say that we're moving engineering up to higher and higher levels of abstraction, but we're also moving it closer to the real world. I think that engineering is becoming broader in scope, higher in its levels of abstraction, and it requires people who can absorb these new ideas in real time. You just can't learn everything in an undergraduate or a graduate program. So it requires people of breadth as well as depth for the future, and people who can learn on a continuing basis.
HAWLEY: To me, the most intriguing thing about computers is that they bring everything down to the same lowest common denominator—bits. That means people who are working on biotech can share code with people who are designing automobiles, because it all moves through the same sort of digital soup strainer. I think that's amazing. Look, 25 years ago, when Lucas was inventing Star Wars, he didn't have a word processor. He had two fingers and an Underwood typewriter. And now he's got a movie that's almost entirely synthetic. You don't have to hire real actors. It's hard to project how much further we're going to go in the next 25 years, except it's going to be a much bigger jump than the last 25.HASELTINE: But let's suppose for the moment that what you're saying is already true, that we're designing things that are too complicated for anybody to understand. What are the implications of that?
HAWLEY: That you need more machine understanding.
HASELTINE: We shouldn't care that there is no single person who knows what's going on under the hood?
BERNSTEIN: That has been true for 100 years, 200 years, 500 years. I mean, if you look back at it, nobody ever understood everything. The way the Romans made sure their bridges worked is what we should do with software engineers. They put the designer under the bridge, and then they marched over it. That's why their bridges still stand in Rome.
HASELTINE: Today you see computers being designed by nondeterministic, non-closed-form-type algorithms. For example, genetic programming or neural nets, that sort of thing, where literally we may not understand what we're doing. And now we're talking about a new generation of computers—quantum computers—where we don't even know what intermediate calculations are going on, because we can't know. So let's just extrapolate a little bit ahead into the future. If we have a system that is an emergent property of some algorithm, aren't we starting to border on the kind of chaoplexity that Michael Crichton talked about in Jurassic Park, where we don't know what it is we've got and where it's going to go?
AHO: I don't know how many people here have read Steve Wolfram's new book, A New Kind of Science. He makes the point that very simple programs can exhibit very complex behaviors. Those of us in computer science have known this for several decades. Who knows what the underlying programs governing the behavior of the universe are really like? I think there's accidental complexity and essential complexity, to use terms that Fred Brooks used in The Mythical Man-Month. And I think we have a great future in science of trying to discover the underlying processes. The grand unified theory may not be in the form of mathematical equations but may in fact be in the form of algorithms that resemble programs.
HASELTINE: The Internet has already become indispensable to our economy, in the same way as the telephone and electricity. So we are depending on something that we really don't understand, and it's getting more complex. Isn't that a concern?
LUCKY: Here's the thing I keep coming back to in my mind: I can't believe that we went to the moon in 1969 without computers. I think we probably couldn't do it anymore. You'd go up there, relying on a computer, and at a critical point it would crash. HAWLEY: Take that thought and map it onto the world of biotechnology. We're the first generation of human beings to fundamentally engineer new species, and also the first generation of people to fundamentally alter the climate of our planet in ways we haven't begun to understand. We'd better have more than a flimsy handle on this because the sorts of "gotchas" that can come bubbling out of that stew are really something.
HASELTINE: Mike, you point out that we're engineering new species. The definition of life, in most biologists' handbooks, is something that's self-replicating, something that incorporates energy from the outside world, something that creates macromolecules, and so forth. Those are some of the big signposts. Angela Belcher, who's now at MIT, recently made a virus that binds to certain inorganic compounds and grows liquid crystals. So, on paper, it satisfies all of the conditions for life. Do you think the computers of the future will be able to reproduce? Will they be alive?
BERNSTEIN: Well, do you think a computer virus is alive? I think that fits your definition too. They gain energy from the machines they take over. They replicate themselves.
HASELTINE: But computer viruses don't create their own macromolecules.
HARROW: I wonder if computers will eventually begin to replicate themselves, especially if we move into the area of biologic computers. There are already experiments with DNA being used to do computations and, within a year or so, we'll be approaching computers that have the same raw computational ability as a human brain.
HENG: I think the ultimate nightmare is software that can reconfigure itself and infiltrate the Internet and the Web and change all the software on Earth. And nothing you could do on the hardware side would make it any different.
BERNSTEIN: I think the problem with software reliability is the lack of liability, that there's no economic incentive for making reliable software.
HASELTINE: Two years ago, I heard Nathan Myhrvold, who was at the time the CTO of Microsoft, show a chart, and on the chart it said, "Here is Moore's law," showing that computer capability doubles every 18 months, and he said, "And here's the number of lines of code being generated by software engineers worldwide," and it tracked almost exactly the same curve. At the end of the talk, I said, "You know, Nathan, I didn't see where you were showing that the number of software engineers was going up exponentially. Who is going to be writing all this code?" Do tools like computers create a need for more engineers or fewer engineers?
DONOFRIO: More engineers.
HAWLEY: Well, everybody's got to be an engineer.
HARROW: More engineers.
HASELTINE: All right, if we have to be both engineers and technologists, why should the average person who is not a technologist care about what computers are going to do in the future?
HARROW: Very simply, because those computers are driving how we all work, live, and play.
HASELTINE: Give some really pithy examples. BERNSTEIN: Well, you won't be here. You'll be a hologram. You'll be able to have dinner at home tonight because you could come here as a hologram. We will extend that into all kinds of virtual realities.
AHO: We're definitely entering the world of virtuality. How we interact with people and organizations, whose rules we use, whose tax laws, and so on, is totally unexplored. We're going to need even more people who have technological understanding. We don't have to call them engineers, but they'll have to understand the tenets of the technology for us to have a productive society.
LUCKY: Why the average person should care is because this is what they're going to be doing for a living. As engineers have to move up the stack, it leaves this emptiness underneath where the average person has got to do that stuff using computers.
DONOFRIO: Actually, if you care that life science is the science of the 21st century, if you care about your life, you have to care about computers. You won't live to a ripe old age without them. x
This roundtable was held on June 24, 2002, as part ofIEEE's INFOCOM Conference. For a full transcript, see the IEEE CommunicationSociety's Web site at http://www.comsoc.org/headlines/EngineerRTFINAL.pdf.
