MIT MechE Symposium: Mechanical Engineering and the Information Age - Nam Pyo Suh, Papken der Torossian, Seth Lloyd
SETH LLOYD: So this conference, Mechanical Engineering in the Information Age, is designed to look at the role of information in mechanical engineering. Now my background is in physics. When I came here, I didn't know anything about mechanical engineering.
But I knew a bit about information. And over the past five years, I've been amazed at the extent at which problems in information and information processing permeate the whole fabric of mechanical engineering. You buy a car, it's got 30 microprocessors in it. Right? That's a lot of microprocessors. They all had to talk to each other in the right way, or else your car doesn't run properly.
Similarly I've had a great appreciation as-- my job, I try to build extremely small, powerful computers. And I've got a much greater appreciation for the extent to which mechanical engineering has contributed to the information revolution. Many of the primary breakthroughs that allow us to have the computer revolution at all were mechanical breakthroughs. Breakthroughs in materials, breakthroughs in processing, breakthroughs in manufacturing.
So this conference is here in recognition of this fact, and to discuss what's happened and what we're going to do about it. My name is Seth Lloyd. I am a member of a triumvirate that Professor Nam Suh set up. A three man group. My other organizers are Sunny Sue and Sanjay Sarma.
I should say that the triumvirate was an ancient Roman custom. And Sanjay Sarma is the Julius Caesar of our triumvirate.
He is the one who really did all the organization for this. Sunny and I just like the Pompeii and the Crassus.
Okay. So to keep us on schedule, I'd like to introduce our first speaker, Nam Suh, the head of the Department of Mechanical Engineering. He's the cross professor of mechanical engineering at MIT. Nam Suh.
NAM SUH: Thank you. Can you hear me if I talk from here? I guess so. Well certainly, I'd like to welcome all of you to MIT. And I'd like to thank our colleagues for organizing this. Seth Lloyd said he didn't know anything about mechanical engineering. At the time we interviewed him for a job, that's not what he said.
We hired him because he told us he knew a lot about the mechanic engineering.
Certainly, it has been a pleasure to work with my colleagues in the information area. Today you'll be hearing from very distinguished speakers, and especially the keynote speakers today, and the luncheon speaker, dinner speaker-- these three people, although they may not realize it, they had a profound influence the way we developed information technology.
In fact, they had a profound influence on me-- the way that I think about information. And indeed, many of the things we have done have been affected by their thinking, indirectly and directly. So I am looking forward to their talk today. And I hope you enjoy their presentation as well. May I show the next slide?
One of the reasons we decided that we should have this conference is that we should exchange information as to what all of our sister institutions are doing in the field of information technology. And also, we want to tell you about what we are doing.
And also, it's not a big secret as to what we have in mind. I hope that some of our sister institutions that do not have strong programs in information area will start programs and hire our PhDs. We are producing a lot of PhDs. And hopefully, you'll hire them. And that, in turn, will make room for our future students in our programs, and so forth.
And I guess one of the important issues is why IT in mechanical engineering. In fact, within our own department, we have had a great deal of debate on this issue. In fact, the first committee we formed in our department to look at this issue came back with a report that says we shouldn't do anything in this field. So we formed another committee.
Headed by Harry Asada. And that committee came back, said, yes, we should do something. So we said, gee, now we have unanimous support for the idea.
So this is how we justified getting into this business. And so since then, we did a number of different things. I want to review with you briefly what it is that we are trying to do.
It all has to do with our educational goal here in mechanical engineering. And I'm sure it's true of many other departments at MIT. Our job is to produce future leaders. So we ask ourselves what our future leaders need to know to lead industry, to be a leader in engineering, academia, and so on and so forth.
So our conclusion was that indeed, for future leaders, it's almost indispensable not to teach our students something about information technology. We have to give them the background information knowledge required so that they can become leaders.
Very few things are purely mechanical anymore. And indeed, the whole idea that the mechanical things consist of geometry I think is ancient, and we have to depart from that. And so we decided to go into this.
So what we have decided to do is, in addition to-- it's not in lieu of, or in place of-- but in addition to the traditional mechanical engineering subjects, we have to teach our students something about information technology. And also biology. So we decided that one of the major goals of the ME department is to transform.
Some of my colleagues think that this use of the word the word transform is too strong a word. But really, to transform, in my opinion, to transform mechanical engineering from a discipline that has been based primarily on physics into one that's based on physics, information, biology, while maintaining a strong foundation in design.
If you look at this, you can see that the words physics and biology are related to natural phenomena. But words like information is really, really related to synthesis, or artificial science, or how are we going to relate it. It turns out that that's what engineering is all about in some ways, because we do design, and design really is artificial science or synthesis.
So in 1995, we created d'Arbeloff Laboratory for Information Science and Technology under the leadership of Harry Asada as director and Ian Hunter as a co-director of the laboratory. So they are the ones who worked very hard to create the research base, create educational programs. So all the credit goes to those people, plus other colleagues who are in the d'Arbeloff lab.
So what we have done is we added very outstanding faculty members. And many of these colleagues have disciplinary backgrounds other than mechanical engineering. Seth Lloyd is a physicist. Sunny Sue is electrical engineer. And you'll here about what they are doing. And Sanjay Sarma is a mechanical engineer.
So you can see that the new intellectual frontiers these people are creating by working together is really what's making information technology possible within the Department of Mechanical Engineering here at MIT. So you'll hear about new subjects they are creating. And they have vigorous research programs, and so forth.
One of the things that I thought I should tell you a little bit about, our philosophy of our research-- and I keep talking about this with anyone that I can talk to-- is that if you plot in a horizontal axis researcher spectrum, and then along this axis, if you plot impacts, the research activity level. It's hard to see. This slide's not very good.
So this is the research spectrum. So we go from very basic research or fundamental research-- you can define anywhere you like. And this is technology innovation. And this is activity level along this axis.
And typical distribution is something like that. Most of the professors and a large number of researchers are in this middle area. They spend I think 90% of our colleagues in universities are in this area.
But if you look at the impact made by this kind of research, the curve is just the inverse. So I say at MIT we should concentrate at this end and that end of the research spectrum. And I say we shouldn't really care whether someone is at this end or that end, rather than argue about what is engineering science, and so on and so forth.
There are all sorts of basic research one can do, and there's all kinds of technology innovation one can do. Some people are able to do both. Some people want to stay at this end or the other end. And I say that that's fine. And then let the middle ground be left to someone else.
So I guess one of the things we need to talk about-- what is information. Information can be defined in many different ways. To prepare these slides, I looked at Webster's Dictionary, and there's many, many definitions. So for the purpose of what I want to say in a few minutes, I say information is the knowledge that must be added to achieve certain functional requirements. So information can be measured in terms of probability of success, in terms of achieving those functional requirements.
Then what is information technology? If that's information, what is information technology that we are concerned about here? It has to do with generation of information, storage of information, transmission of information. Codification. Codification should be down there. Somebody put up that word up there.
Algorithms. Comprehension of information-- deciphering information you transmit. Machines that produce information. Devices that deal with information. And equipment for doing all of the above. Okay?
So that's information technology, very broad. One of the issues some of our colleagues raised within our own department when we talked about this is, what is information? There is no such thing as information. But to me, all of these are included within the term information technology, or IT. May I show next?
I think there are two reasons why mechanical engineers must get involved with information technology. And this, again, is influenced greatly by my association with Papken der Torossian, Neil Pappalardo, and Alex d'Arbeloff. You'll be hearing from them later on.
Mechanical things are getting much, much smaller. I didn't realize how small they are getting until I joined Papken in the area of lithography. And it's really amazing. Literally in the cross section of our hair, we can put in 1,000 lines. And that's a lot of electric circuits on the cross-section of our hair.
And we can do this because of mechanical engineering technology, because of optics, and because of all of these things. And as a result of this, we started looking for optics people. And I think today, if I may say so, our department has the strongest optics perhaps group within MIT, if not in the country as a whole. May I show next slide?
The other thing that's happened is at the opposite end of the spectrum. Mechanical engineering is becoming very highly distributed. Factories are located all over. Design is done globally. You have to exchange information.
A large number of people in distant places got to work together to create the same final output. And that cannot be done without our understanding of networking. Without our understanding of how design should be done in a systematic fashion, and so forth.
It turns out that mechanical engineering has always dealt with information. We never called it information. We are always in the area of design and manufacturing. It's actually these two fields manipulate information to create artifacts. And that's the business we have been in, and we just never call them information.
So I'll show a very simple example of how we evolved. In my own research, we evolved asking some very basic questions. How one goes about designing simple things to large complex systems. And how it resulted in a completely new technology for creating software, as well as many other things.
So it turns out in axiomatic design, there are only two axioms. One is called independence axiom. The other one is called information axiom. And it requires that the functional requirements must be independent throughout the design process. And we have to define what we mean by functional requirements and so forth. But essentially, that's one of the axioms.
The other axiom says information content must be minimized during the entire design process. Very simple statement. From these, we have a large number of theorems, correlates, what have you, which can be used by students in coming up with good design.
So based on this kind of thinking, we in fact created a very complicated commercial software. And I'll just tell you a little bit about the process by which one goes about creating this kind of software. May I show next slide?
So in axiomatic design, we talk about going from mapping from domains to domain. First of all, we assume that there are four domains. Customer domain, functional domain, physical domain, process domain.
And Then as a next step, you have to map it between the domains. And when you map between the domains, you have to maintain independence of functional requirements. And that is, you have to maintain that from what it is you're trying to achieve to be independent from other things that you want to achieve.
And then after that you have to zig and zag and create a hierarchy. And you cannot decompose something by staying in one of the domains. You have to zig zag. So that the only way you can decompose is zig zag it. And then you can create hierarchical structure in each domain.
And then the independence axiom tells you which are the good designs, which are the bad designs. And then it tells you a procedure by which you have to follow your design to be able to satisfy these axioms.
So if applying this idea to a software, it looks like this. Applying this to software is sort of like a v model. We are going from functional domain to design domain, and that's at this part of the leg. Going from design domain to process domain is this part of the leg. in software design.
And essentially, we design along this leg first. And then this is, in typical, is very equivalent to object-oriented programming. But by doing this, we create object-oriented program, the structure of it, without any human intervention. So in fact, you can think of creating a software without much human intervention measured.
So as a result of this kind of thinking, as you design it, you have a matrix. Sorry that you cannot see this, but these are functional requirements, and this could be the design parameters. In this case, we are showing design parameters versus process variables. But you can plot the same thing where this is functional requirement and that's the design parameters.
These are what people might call object-oriented programming. This is what you call object. That is the data. And this is the method. So in terms of an object-oriented programming language, that's the way it's going to be.
So the result of this, without spending too much time, the result of this, you get something like this. This is like an electric circuit diagram. Each one of these little boxes are the modules at the lowest level. And the module we define is very specific. The module we define is a role of the design matrix.
The result of that is you get a diagram like this. We call it four diagram, just like an electric circuit diagram. Just like a capacitor and resistor, we have these modules.
It turns out that this is generated automatically once you do the axiomatic design. With a click of a button, you get this whole thing. Then the only thing the programmer does is stick in the modules. In fact, programmer who works on this model, it's all very simple modules, because this is a Role of design matrix.
This person can be located in India. This person can be in Boston, and so on and so forth. And all you have to do is just stick them in there, just like you pop in resistors or capacitors, and then you have a software.
I teach this to-- right now, I teach graduate students in my axiomary design class. And it turns out that one of the students is a PhD from MIT who worked for Xerox for about 15 years. He came back for a masters degree called systems design and management program. And
I have another student who has also equally long experience software design at Xerox. And they tell me that the way we did this took only 20% or less of the time it normally takes at Xerox. The software we developed. They know the software very well. They played with it. And then they work with the people who are doing this.
And in the case of hardware design, machine design, I was talking to the president of SVG the other day, Will Hightower. He says now he really thinks we can cut down the time it takes to develop new machines by a factor of 2. What used to take 30-some months, we can do it in 18 months or so. So systematic thinking, having a very rigorous criterion for good design, enabled us to do this.
The purpose of talking about this is the fact that not only I want to tell you a little bit about my own work, but the fact that we started out with a very basic research, asking the very simple question of what constitutes good design. And then having done all of that, now we can apply it to software design. This is technology innovation. Software design, hardware design, or what have you.
So I think the mechanical engineering departments must weave all of these information technology in teaching our students. Otherwise, I don't really think they can become leaders. Without having background information in technology area, I don't think our mechanical engineering students can ever become leaders. They can continue to do traditional mechanical engineers' work. But if our goal is to produce people who can become leaders, they have to have understanding in this very important area.
And having done some of this kind of work, and having seen what my colleagues-- I have amazing colleagues. That's really an understatement in many ways. Many of my colleagues, as you'll hear today and tomorrow, are really doing very, very amazing things. And I'm convinced that the kind of things that we can do in mechanical engineering will contribute to the future of the information era. And with that, I guess I'll stop. And that gives right on time. Yeah? Okay. Thank you very much.
SETH LLOYD: Actually, we have time for a question or two.
NAM SUH: Yes?
AUDIENCE: You mentioned, for example-- well, you were saying that MIT is taking the lead in IT--
NAM SUH: I didn't say that.
AUDIENCE: Excuse me, no. What I meant to say--
NAM SUH: I said we are doing a lot in it.
AUDIENCE: That was a wrong statement. What I meant to say is that this conference is in some sense going to influence other mechanical engineering departments throughout the country, perhaps. Or there might be an offshoot of that. And you mentioned, for example, optics as an area. And that's not necessarily typically something that all ME departments take as an leading area.
Do you see that is developing in the future? Do you think that that's typically going to happen in physics departments, or electrical engineering departments?
NAM SUH: Physics department more or less gave up optics, right? Because that's not how you get Nobel Prize. So you get out of that. And until we started looking into information technology area, we really didn't think optics was that important for us. We hired somebody on a part time basis to teach optics, but that a halfhearted attempt to offer something.
But then it's very clear, once you get into information technology area-- you cannot deal with small things, you cannot deal with information transmission, communication, measurement, all of those things, without optics technology. When you're trying to deal with atomic level tolerances, there's no other way.
And so I think it's very important topic. In fact, I'm very pleased that we are so strong in the optics area. It's amazing what they can do.
AUDIENCE: Donald [INAUDIBLE] from UCI. I have a question about, you mentioned about how many of the classes related to IT to the curriculum of mechanical engineering. I'm sure that MIT, as well as other departments, will have a curriculum.
How do you know what is worth adding to the curriculum? You were given a selection. So which ones are you sacrificing?
NAM SUH: Yeah. There's a long answer to that. And Harry Asada I guess may talk about this. There are a couple of things. One is that we created undergraduate level elective subject. Undergraduate and graduate level. Seth Lloyd is teaching in the area of information. What's the title?
SETH LLOYD: Information and probability.
NAM SUH: And then we are in the process of creating what Harry Asada calls gateway subjects. So our students can take a certain set of first level graduate subjects. And then they can move right into advanced graduate subjects in electrical engineering and computer science without having to take all the other prerequisites.
So we are trying to convince our colleagues in electrical engineering that our PhD students in mechanical engineering ought to be able to get a master's degree in electronic engineering as well, while they are working on PhD. But we are still in the discussion stage.
So there are lots of those things that are happening. And the fact that we are emphasizing information means that lots of other subjects we teach will bring some of these things. In design subjects and other subjects, we teach them about what do we mean by information, and so on and so forth. So both independent subjects, as well as incorporating some of these ideas into existing subjects.
SETH LLOYD: Actually, in the interest of time, I think perhaps we should get going. There will be--
--education, of which professors Suh will be a member. And during those panels, we're going to hope to have a very general discussion. So if you have questions, some of them might be probably saved for those panels.
Our next speaker is Papken Der Torossian. He is the CEO and chairman of the board of Silicon Valley Group. We're very lucky to have him here, because he flew out yesterday to come here. He's flying back to California this afternoon. And he's flying back here on Sunday.
And this morning, in talking with him about the ups and downs of the semiconductor, he's the only person in the last five years who's made the life of an MIT professor sound calm.
So-- and, Papken is, perhaps more than anybody here, responsible for bringing mechanical engineering techniques into making the information revolution happen. So Papken Der Torossian.
PAPKEN DER TEROSSIAN: I think the question that was asked Professor Suh about how do you get mechanical engineering, what do you teach here to really make it relevant? What courses you take out, and what is there?
I'm sure that a lot of you over here can spend a week discussing that. Because it's very, very important, and very relevant. Can you hear me? If I stand back here, can you hear me? I just want to know how sensitive it is. Can you hear me now? Yes? Thank you.
Okay. Let me just-- I'm a mechanical engineer by training. And to repeat what Professor Suh said, that the optics that I learned here was a freshman physics course, was that first two weeks of the freshman physics course. And the book was by Professor Sears, which was written in the '50s. So that gives you an idea how old I am.
And frankly, it's still relevant. So the message is, what the professors, what good universities-- Berkeley, MIT, Stanford, all the good universities in the country-- what you have to teach is some very, very basic, basic fundamentals. And then teach people how to learn and adapt themselves to the changing environment.
If you do two things-- just teach fundamentals and teach people how to learn, which I think MIT and Berkeley and others schools do very well-- then that's what you really have to do. Because the rest of the time, you don't know what's going to happen.
When I came to school here, when I was in high school, somebody, when I filled the application at MIT, they said, what do you want to learn? And I just said, mechanical engineering. Somebody said, why? Because I like cars. And that was the reason I chose mechanical engineering. I had to put something in there, because they said, what course you want to go?
And then I came here. And when I came here, believe it or not, they were not teaching anything about cars. The professor [? Heschel ?] Schwartz was a emeritus professor, but they weren't teaching courses in cars for 10 years. What they were teaching is jet engines, because that was there, I went there. So you have to always adapt.
Another thing I want to impart to you is that in physics, or mechanical engineering, or whatever you teach, what I always say that you really have to teach the basic things with freshman and sophomore. And then junior and senior, you amplify a little bit. And you spend seven, eight years getting a PhD. All you do is learn that f equals ma.
But when you get a PhD, you learn that f equals ma most of the time, until you reached the realm of quantum mechanics. And really, that's all you do. You don't really-- and that's what you apply in business.
When I first left here, I was hired by Hewlett Packard. Hewlett Packard at that time was primarily run by electrical engineers. Bill Hewlett and Dave Packard are electrical engineers. And by the way, Bill Hewlett is a graduate of MIT.
And I know that a lot of people associate them with Stanford because they basically have given more money to Stanford than Stanford. But Bill is a graduate of MIT. And at that time, I was only a mechanical engineer that was hired professionally.
And the electronics industry at that time was moving from big boxes with tubes into transistors. And I'm sure some of you realize when you go from a big box to transistor, lots of heat, and go transistors, small boxes, miniaturized, the biggest problem in those days-- the transistors were germanium and silicon.
They were very sensitive to temperature. And if you went over 150 degrees centigrade-- in fact, if you really much over 120 degrees centigrade-- a lot of these circuits did not work at all. So biggest problem was making the circuits work.
So they asked me about it. And I used some very simple principles of mechanical engineering. Frankly, a lot of estimates. And I just came up with some formulas which I said, you have to use.
And they took these little documents and made a little manual about if you have a box this big, you can put so much power into it. And some basic principles. How you can call power transistors, as you know. I did simple things like put grease under the transistor, which made the conduction better. And put a little mass at the bottom, or put a little heat dissipator. And they thought I was a magician.
The point is I showed them that I made a formula that temperature is voltage, and the heat transfer is current, and resistance is 1 over k. And everybody thought I was a magician. They understood electrical circuits. The point is the basic principles of what you learn is really what you really apply in life.
And I want to share with you some of the thoughts, and what we do here at Silicon Valley, and how mechanical engineering is really very, very relevant to the information age. Next slide, please.
Well, I want to spend some time here. And I also want to say very little about bioengineering. As you know, when I was going to school here, Professor Mann was here. He was working with the University of the Mass General working on prosthetics, and limbs, and things like that. At that time, we always think of--
Today, mechanical engineering is so essential in medicine, because we measure temperature, pressure, flow. And before, doctors always guessed at things. That's what they call practice medicine, because they really practiced on us.
But today, you can actually measure things. We can actually give them temperature, pressure. We can give them. And you find out a lot of this information is very, very relevant to what's going on in your body. Next slide, please.
This is-- by the way, you don't have to take any notes. All you have to do is give your email address to Professor Sunny or Lloyd, and then we'll email all this. This are all on a computer. We can email them to you. Or we can make copies of the overhead.
In the old days, as you know, we really had gears in mechanical engineering, and what we call the multiplier for the brawn. And mechanical engineers basically made leverage, gave leverage so that human beings can lift large weights. And nowadays, of course, mechanical engineering is empowering what we call the brain. Next one, please.
As Professor Lloyd said, mechanical engineering is at the core about information technology today. You cannot make semiconductors without mechanical engineering, frankly. I'm not going to be that selfish. There are some chemical engineering, and there's some materials science, and there are some other disciplines. But fundamentally, to get down to the small geometries and to really, really measure a lot of these things, mechanical engineering is at the heart of it.
And our company frankly, I will say this. You cannot make Pentiums. Which is, since 1994, every Pentium, or most of the Pentiums built, are built on our machine. And every generation of Pentiums that is advanced from 120 megahertz up to 1 gigahertz now.
The only way you can get that is to make the lines thinner, and put more transistors on it. And we at SVG are the enablers for that basic technology that was developed. And I'll show you how that works. So we are extremely important for the whole internet and information technology. Next slide, please.
For those of you who have not followed the semiconductor industry, basically, you know you take a big ingot that-- in fact, I have seen at MIT a big place where they used to grow ingots here. And they gave it to IBM.
You take an ingot. You saw it first. A mechanical engineering problem. Then you basically put photoresist. You spray photoresist on it. Which is not a mechanical engineering problem. How do you spray something that's kind of viscus smoothly? You've got to good accuracies of plus or minus 20 angstroms. That's what you have to really do on this. Only one micron thickness.
So what we used to do in the old days, we used to really charge a huge capacitor. And suddenly, put it on the motor. And the motor really spun it very fast, so that you can really get acceleration very fast, so that you can really get a good flow.
Then of course, after that, you have projected image, which is not in mechanical engineering, what we make. You have to project the real tiny transistors. In the old days, it used to be 5 microns. Now it's 0.18 microns in the technology today in production.
And then you develop it. And of course, then you etch it. You put some other stuff in it. That's another mechanical engineering. We make furnaces where we basically-- it's a heater, and you basically put some gases in it.
And you have to make sure that there are no stagnant flow. You have to really flow it properly. Another mechanical engineer phenomena. So really make sure that you don't have any particles, and you have a smooth flow. Another area of mechanical engineering. So I really think mechanical engineering goes into all of it. Next one, please.
Lithography of course, is basically the most difficult and the most important part of making semiconductors. That's the ability to put the finest image on the wafer. And if you can't do that, whether you etch it, whatever you etch, whatever you develop, whatever you do is nothing. That's the basic enabler that gets to thinner and thinner lines.
And believe it or not, optics-- fundamental, geometric optics that's taught in two weeks or three weeks-- that's all we use, year after year, for the last 30 years. So whatever you learn in two weeks, maybe you should amplify it a little bit some applications so that people can get better at it.
And like Professor Suh said, the physics department, thinks that's very elementary. But when you really start using it, really making machines with it, you find out that it's not that elementary. Things get very, very difficult. Next one, please.
This is a sketch of a typical machine. It is amazing sometimes that the damn thing works at all. Frankly, it's just so accurate and so precise. Just to give an idea, we actually do print 0.1 micron. And the accuracies are within 0.01 microns. So you have to make machines that measure 0.01 microns, because you cannot buy these things. And these machines, these carriages have to accelerate.
Just to give an idea. These things run about 100 wafers per hour. So it's about 1 and 1/2 wafers per minute. So in 60 seconds, we have to take a wafer, put it in, take it out, put it in, and then put about 50 or so, BASICALLY dyes on it. Take it out. All accurately.
And you have to do this. So these are very, very fast, and very accurate. Everybody can do things slowly. But when you start, you basically have to hit a machine, and you stop the vibration, and do it.
I used to make disk drives before at Hewlett Packard. But disk drives compared to this-- we always argue about how fast you can accelerate something and stop it. We find out that when you accelerate something, that's okay. You can get there. But then it vibrates. So how do you stop the vibration before you can start reading and printing?
And I think once people realized that in disk drives, the settling time sometimes is much longer than the access time. And I think in these cases, we have to bring it in, align it accurately, which is wafer to wafer, to layer to layer. You have to do 26 layers on top of each other, and align it within 0.01 microns, and do it. And you can understand what we're talking about.
The difference between physicists and mechanical engineers is that physicists talk about it, and mechanical engineers have to do it.
That's for Professor Lloyd. These things really have to go, you know? I was I was at a meeting the other day. And I pointed out to some people that when we went to the moon in 1969, the computers that we used to go to the moon were so elementary. I mean, compared to today's PCs in terms of power.
And those astronauts, literally, they had the [? ask ?] test. Quite different. They really, literally had to fly those things with their feel. You couldn't see very much, as you know. And it's really amazing that we were able to do that at the time we did it, when you think about it.
The amount of computer power we have today compared to what you had in the '60s is just-- in 1965 the IBM 360 was embedded. So '69, there's not much technology. And the PCs didn't come until about 1981, '82. And as you know, there were 8-bit machines, and they were very, very slow; elementary. Next one, please.
This is an actual picture of that schematic diagram. We used some robots. And frankly, these materials have to be very light, very stiff, and damped, very accurate. And so we do this, actually. Believe it or not, these machines do exist in production. Next one. And they are making wafers every day. And that's what enables the whole information technology.
That's another picture of what the wafer is. These are 8 inch wafers, and we're going to 12 inch wafers. And of course, then you can see making a 12 inch wafer, and making a smooth photoresist on it at 10 to 15 angstrom accuracy. You can see it gets much more difficult.
And we are working very closely, by the way, with MIT. Because the knowledge today is, we don't have all the knowledge. And we really try to tap on the knowledge of the university so that we can get the best technology available. In fact, I'll show you some slides later as to why it's so important for industry to work with the university today, whether it's MIT or Berkley.
Just because the information is transferring so fast today that the days when researchers made something, got some papers, and 10, 15, 20 years later, it went into industry by osmosis or diffusion, whatever it is, is no longer valid. You really have to get instant latest technology.
And as well as the professors. Professors need to work with industry to find out the real applications and. Nam tells me many times, when he comes to visit us, sees the machines, he suddenly thinks of 10 ideas. And he's students come in, and his professors.
Because what we do a lot of times, when we're trying to accomplish something, and we don't. We're in a foxhole. We don't see sometimes what's going on. And one professors and a student comes in. He sees all these things that we're doing wrong, and what we can do better. That's another, I think, interaction that's necessary between industry and university today. Another one, please.
Another picture of the robotics that we use. And another one. We use optical polishing to 1/300th of a wave. And we used to use old what I call German opticians, because we don't even make opticians in America anymore to really polish these things, and measure it, polished.
But now, believe it or not, in California, Richmond, there's a company called Tinsley, which we purchased. They developed basically a method, computer-controlled, where they are measuring while they're polishing with the feedback. And we can do spheric as well as aspheric lenses. And this was really done-- and that's probably the best place in the world where we can actually do aspheric lenses very accurately by computer-controlled optical polishing. Next one, please.
This is a picture about how when we rotate what I call the wafer, and we put photoresist on it, you can see all the flaws and the pressures that occur when you really want to make this thing very, very, very thin. The reason that you have to control within 10 to 15 angstroms is because every chip is about, maybe, nowadays, between 1 centimeter squared, or 2 centimeters.
And you have to make sure that every part of that wafer has a uniform flow. Otherwise you don't get the speed, and you don't get what I call CD control-- Critical Dimension control-- accurately if you don't have the thickness right. And therefore, you don't get good yield. And yield is the name of the game in this business. Next one, please.
Mechanical engineers, we also are very, very much interested in sensors, actuators. And of course, some of you are very [INAUDIBLE] controlling. As you know, controls about 15 years ago, control was very, very, very important. Now, of course, it's going more into biotechnology, as well as a lot of IT technology. But they're still important.
And all these technologies. We use planarization, which is the chemical mechanical polishing. It really is extremely important today. It's a very dirty process. It's really polishing, and then it's grinding. But except you have to do it very accurately, because if you don't do it right, you basically eat into the chips, and you have very low yield. That's another mechanical engineering process. Go ahead, please.
This picture probably gives you what we're talking about. As most of you know, 1 centimeter is 10,000 microns. And your hair is about 300 microns. So we do, just to get an idea, 1/1000ths of a hair. 0.18?
No, it's less than that, even. Maybe 1/3000th of a hair. So it's really very, very small dimensions we're dealing with. And we have to measure it, and fabricated it, and control it. And of course, keep the dust out. Next one, please.
This is a electromagnetic spectrum that gives you an idea. But basically, we are working today at-- basically, GNH lines are 436 nanometers. And the eyeline which was used until about five years ago, 365 nanometers.
We are using 248 nanometers now. And now we're going to 193. And of course, next is 157 nanometers. That's basically a source of light. Next one, please.
As I said, our company is the leading producer of lithography. We really do deep UV, shorter than 4,000 angstroms, or 400 nanometers, which is visible light. Then we are going with the 193. Next is 157. And then EUV.
And these are all optics. Until about 20006 to 2008, we're still going to be geometric optics. People want to know when the technology is going to stop. We're still using the optics. EUV can start considering whether we're going to be on optics.
And there's a lot of discussion now in universities whether we should go to the next. It's going to be on human cells, or on tissue instead of silicon. But right now I think for the next 10 years, we'll be using silicon and optical techniques. Next one, please.
And I think what I talked about is what mechanical engineers do to opticals in information technology. And I think I want to spend some time telling you what information technology has done to us in mechanical engineering. And it helps us in product design, in manufacturing, and they call mechatronics. I'm sure some of you here teach mechatronics in your school. Next one, please.
As you know, if da Vinci had computers and CAD systems, as you can see, the design that he-- his innovation was not any worse than we did, except he didn't have the tools. So the tools that we have today like the axiom-axiomatic design that we use really, really speed up, and really make mechanical engineering life very easy.
When I was designing things, we used to make parts that fit together. Sometimes we made mistakes and parts didn't fit. Very simple. The holes were at the wrong place. Today, you can actually do on CAD, you can actually simulate it and makes sure they're in the right place, and before you do it.
And equally, by the way, most of you work software. Software has the same incompatibilities as the hardware does. And most of the time, software guys get away with it-- get away with selling things that are not finished. Unfortunately in hardware, you can't. And your computer suddenly freezes. You put a new program in, a new application program. The damn thing doesn't work.
So what happens? Because there's a software incompatibility with it. But software guys won't admit that. They say, it's a user problem.
That's why they charge you 15% a year to just maintain it.
But those things can be done by axiomatic design. Axiomatic design is really nothing but the replacement of a good engineer. If you have a good engineer who really thinks clearly, you've probably used axiomatic design mentally. But axiomatic design makes a good engineer a great engineer by showing systematically what you have to do, and what avoids potential problems and conflicts. Next one, please.
Like I said, now, we're using PCs. We have computer networks now we're using. We're using the internet now for really sharing data and sharing knowledge. We used to use pictures. Picture, tell, and other things by the television.
Didn't work. Now we have used the internet. Actually, you can share data online instead of waiting for a long time. So actually, mechanical engineers now use that. Next one, please.
And I think this is a very interesting chart. And I think this is something that all of you in the academia really should really pay attention. This is important. It shows here that in the 1 megabyte, the red one is in the 1 megabyte. See how long it took? It took about almost 186, almost four, five years for it to mature to its maturity with the number of units.
On the left, on the ordinate is the number of millions of units. On the abscissa is the years. It took four or five years for it to mature. And then it took almost 10 years to decay. But the next generation, it took a faster ramp. And the next generation is faster than even yet.
And right now, we are really doing with 64-megabit megabytes going. You know, right now, it's dying. But look how fast that grew. And now we're going to 256 megabit, and it's really growing faster.
What the message really here is-- and I have another chart that we'll show you-- is that's why I think it's extremely important for universities and industry to work together. And then the next chart, please.
This is a chart that was recently published I think in Business Week. You can argue with the numbers. But fundamentally, it says that on the left side, on the ordinate, you see the percentage of US population penetrated. And the abscissa is the number of years since the product invented.
You can argue whether the telephone or the automobile or directive was invented. You can exchange IT by 10 years here and there. But the point is it took a almost let's say, 50 years to 60 years for the automotive to penetrate 25% of the population. And if you look at the farthest left chart, it took less than six years, or five years, for the internet to penetrate 25% of the US population.
And the message really is that things are moving a lot faster. And I think it's extremely, extremely important for us in industry and academia to work together and vise versa, because we don't have the knowledge. We a lot of us don't have. And you have a lot of new ideas and new knowledge.
And I think we give you a lot of application capability. We give something to sink your teeth into. And I really think that is extremely important for us to establish the collaborative effort that we have here. Next one, please.
In manufacturing, supply lines. And as you know, let me just give you an example that's not even in our industry. My brother-in-law is a vise president of a garment industry. So what's relevant about garment industry to our business?
It used to be Sears and Walmart and all these guys, they ordered something, and they basically got the clothes, and they put them in their warehouse. And they built huge warehouses. They distributed these things into their various stores. And then basically sold them.
Today what happens is this huge department store like Walmart, or Sears, Or any others, literally have thousands of distributors and suppliers. And they tell them to tag this model machine, whatever clothes they are making.
And in the clothing business, believe it or not, clothing is just as perishable as computers. If you don't sell something for the spring-- right now, they're working on the next fall fashion. And if the fashion doesn't sell, it's perishable. That's why you get them to all these outlet stores.
So they have to do they have to get this machine. They have to accept the design. They have to manufacturer it. They make it to various sizes. Put them into put the tags, and send it directly to the store.
And in the story, you have all these new computers now. They even tell you exactly where it came from. And within every day, they look at how much they are selling. Within two three days, if they sell, that much they take it, and they send it back. Or if they sell it a lot, they automatically reorder it.
What I'm saying is even as mundane an industry as the clothing industry, the whole information and the speed of operation is happening. So we in mechanical engineering, we in manufacturing, we have to do the thing. We ought to really improve our process lines process systems.
And the days when you designed a part, and you went to the machine shop. Three months later, you got it, and it was wrong. And you got it back again. It does not do that anymore. You really have to do things accurately and very, very fast. Next one, please.
Another area that mechanical engineering reliability. You cannot do things-- when I came to this industry 15 years ago, on machines probably worked about 150 in one week. One week is 160-some-odd hours. And the way they looked at it, a machine lasted a week without anybody touching it, they were okay.
But when you are going to get to a lights out, what I call automated factories, and you have 500 steps to make semiconductors, and you have a lot of machines, different machines, complex machines, different processes, and you want the lights out factory without, people which cause people contaminate the circuit .
You have to have a lot of machines that have to work. And uptime suddenly has to go into thousands, not hundreds of hours. These complex machines have to work reliably. And what better things to do in mechanical engineering than to make these things, all these dynamics things, really, really reliable. All this probabilistic theories. And you have to design things properly. It's extremely important. You cannot get automation without reliability.
As you know, the jet engine-- I mean, if you had an old airplanes, every few hundred hours, you have to really fix it. Jet engines last 12,000, 15,000 hours with [INAUDIBLE]. And very seldom you hear a jet engine fail in an airplane. You just take it for granted today, because of mechanical engineers. Those are the guys who were teaching at MIT when I came here. The next one, please.
Well, I think that all these diagnostic systems internally to really self-correction, right now, we're designing machines that instead of sending all these technicians all over the world, we can actually diagnose them remotely, and get the information, and then immediately feedback to the manufacturing floor.
So these are all systems that the new information age is helping mechanical engineers get systems fast. In the old days, you had to bring the part in, You had to spend on the job analyzing it. Next one, please. But today you have to do it faster.
And I think there is a professor, I'm sure, that teaches mechatronics, the whole system of working together of traditional systems, and all systems. Next one, please.
In any case, in the '70s, we had the servo technologies. Then we had the '80s, the IT technology. Now I think we have the entire confluence of all the other systems, and actually feedback systems, and using the technology to really get our job better. So that's what you have to really teach. Next one, please.
That's a complex chart, basically. I'm sure most of you can draw your own chart, or you can use this one when I send the email to you, about the general field of mechatronics, which is really a lot of the disciplines combine into one, what we call mechatronics. Next one.
Just to give you one slide on-- I don't want to slide the people who are doing bioengineering. I really feel you can spend a lot of time deciding about that's the next field of bioengineering. And I really think, I'm very pleased that MIT under professor Suh's leadership, and Neil Pappalardo's chairmanship, adopted the concept of creating your very strong bioengineering field at MIT.
Because I really feel that that's the next, basically, horizon for us to conquer, as to how we can get this complex body that we have. And I always say, God was a very good engineer. Because every time we look at the body, the way it works, it's amazing. So I think if we understand it, and measure it with a lot of mechanical engineering principles, and use some of that to really influence it and correct it, we can probably don't have to practice medicine. We can actually do medicine. Next one.
So basically, these are a lot of examples of-- we always think of sport engineering and other biotechnology as bioengineering. But basically, mechanical engineers do just pressure, temperature, flow, and all these things we have to use. That's it. Thank you. You can take that slide out.
So I just want to let you know that mechanical engineering is at the heart of things. And I think that the job here, again, is to teach some very, very basic principles. And teach people how to learn as they adapt themselves.
Like I said, I was the only mechanical engineer at HP that was accepted. But now, of course, things are a lot different. And they realize that mechanical engineering has a huge contribution to the IT industry. Okay. Thank you.
SETH LLOYD: Time for one question, and then we'll break. No questions? Thank you. Let's thank our speakers from the morning again.