ion etching to get rid of wet chemical etching. Steppers came along and gave us a 10- or 15-year respite in mask fabrication. Ion implantation came along and got us away from diffusion furnaces, which were terrible on the statistical issue that Chuck raised. My guess is that there is an engineer or a scientist out there somewhere who has ideas about how we are going to break through these barriers, and this will continue to grow. I do not think we need a big breakthrough for 15 or 20 years.
SEITZ: If I may add a couple of things. We thought that the scaling limit was somewhat below a quarter of a micron and would be due to tunneling, statistical fluctuations in threshold voltages at reduced voltages, and a few other effects. Now we know the limit is somewhere below a tenth of a micron. Over the next decade, what matters are not effects such as tunneling, but more mundane interconnection issues. The fabrication processes are improved in all of the areas where there is the most leverage, such as adding more and more layers of interconnect, which is really what has been limiting for some time.
Besides Bob, if, in your view, we can get a mere factor of 10 or 100 in the next 15 years, maybe the software people can take up some of the slack.
MISCHA SCHWARTZ: I would like to focus on a severe problem that Bill raised when he first started speaking, and that is the question of the downsizing of basic research in physics and devices by the major corporations in the last few years. Those of us in the academic world have been very concerned about this. The question is, Who is going to build the hardware platforms 10 years from now that the software is going to ride on? You have mentioned that maybe universities can pick up the slack. We keep hearing this. This is the reason why we are looking at increasing the basic research activities, things that are 10 years or more out. If I look at trying to rebuild Bell Labs, it is like reassembling Humpty-Dumpty to me. I do not think we can put it back together again. Bob Lucky could give us his view on it. My 15 or 20 years at Bell Labs convinces me it is gone and it cannot be replaced. National labs, in my view, require a larger cultural change than the universities do. I think industry must undergo a major change. Sending someone to work at a university is now considered a positive step for your career, rather than a detriment. I think this is the best place for us to turn.
I think innovations will come from all sorts of places. We have got to be in a position to capitalize on them when they do occur. That is the second major problem we have had in this country. We are not very good at this. We do the first start-up. Then, when we have a success, we find that East Asian countries, or even Europeans, now manufacture things in high volume better than we do. I do not think there is a simple answer to this, but somebody said we badly need a policy in this country that says how we are going to address this issue, not only in the silicon area, but in all areas of research. I think we no longer have national labs like AT&T and IBM, TI, Phillips, and Fairchild. The national labs are gone, and we have got to find a way to replace them.
SHUKRI WAKID: If you believe embedded computing is going to be real or very distributed, then intelligent sensors are one way to go. This means you need designs for chips that are a lot simpler, much simpler and more applied. By the same token, if federal digital signal processing is going to take off—and people sometimes say it is a barrier to computing because it is very difficult to do—then it is going to force a need for simpler design versus a more complex design. Do you want to say anything about this reverse trend of complexity?
SEITZ: Many of the processing media, sensor outputs and so on, of military embedded systems use digital signal processing chips. The typical digital signal processing chip dispenses with all the address translation hardware and other facilities required to run an operating system on processors used in desktops. There are at least 40 companies in the United States that make digital signal processors, ranging from boutique companies to the likes of TI and Motorola. As long as there is some money to be made in this area, I think it will continue to be healthy.