Podcast with Prof. Robert Wille - Design Automation Through the Ages

My guest today is Robert Wille, professor of computer science. Amongst other topics, Robert and I talk about design automation throughout the ages, and discuss the analogies between classical design and the quantum world.

Listen to additional episodes by selecting 'podcasts' on our Insights page

The full transcript is below

Yuval: Hello, Robert, and thanks for joining me today.

Robert: Hello, nice that you have me.

Yuval: So, who are you and what do you do?

Robert: I'm Robert Wille. I'm a Professor at the Johannes Kepler University in Austria. And for many years, I work in the domain of design automation. First starting in the design automation for classical circuits and systems. But now, also for almost 15 years, I'm also working towards developing corresponding methods for quantum computing. That's, I guess, why I'm here.

Yuval: And just for transparency, you are also part of our Technical Advisory Board, and we're very grateful for that. 

What does design automation mean?

Robert: It literally means automating the process of design. And in this particular case, of course, of the design of circuits and systems. But it could be many, many other things. The main motivation here is that we have to cope with substantial complexity when we're designing and developing today's systems. The next smartphone, the next AI solution, those are systems you cannot create and develop and design manually anymore. You need, at some point, design automation, because you cannot cope with the sheer complexity.

Today's systems are composed of millions or even billions of transistors or components, and you don't draw the netlist of such a system on a whiteboard anymore. You need automatic methods for synthesis, for compilation, for checking that everything works correctly, and that's what we develop. And the main challenge, indeed, is how to cope with this enormous complexity.

Yuval: So, is that like VHDL or Verilog or these kinds of languages that have been developed for a while? Could you give me a little bit more detail about your specific work, what your group is interested in?

Robert: Programming languages or hardware description languages like VHDL are some of the means we're using. And I would say 50% of our work and research, we are still using to develop methods for classical computing, but the other 50% also for emerging technologies, including quantum computing.

And since you asked for a precise example, a typical scenario is: you have designed something, for example, in VHDL such as a particular control system, something which controls a train or an aircraft or something like that. The traffic preemption is a simple example we use in our lectures. And you have developed that in VHDL, and now you might ask yourself, "Okay, can you deploy that so that you would feel comfortable that this system is now used in practice? Or do you, rather, would like to have something like a check, which makes sure that nothing bad is happening?" Nothing unintended is happening with the system. So what do you usually do?

You have your design, for example, in VHDL. And then you have certain properties or safety guarantees you want to provide. And then, we develop methods so that you can simply put your design and certain properties, for example, to be checked into a black box and, in a push-button fashion, this black box determines whether or not your property indeed holds or whether or not there might be a situation where something bad is happening; the properties failing, something like that.

This is one example, usually the verification example. Another example: synthesis. You write your VHDL code, you'll hit the button, and then the VHDL code will be synthesized. Or compiled when we talk about software system for a particular architecture, for a particular technology and so on and so forth. And all those tools, where you simply throw your design into it, and then it automatically generates a netlist or generates a proof condition; those are the tools we are developing; this is basically what we're doing.

Yuval: So, the process of doing this for electronic design netlist and synthesis and verification has been around for many years, how is quantum similar and how is quantum different than these traditional methods?

Robert: There are lots of similarities. Indeed, when we look at the history, these tools you just mentioned, they are not just available for a couple of years, but we, as a community, are developing them for decades. And it's interesting to see the parallel development of classical computing and quantum computing, because originally, in classical computing, electrical engineers built the first computers, and then over time, they evolved so that they suddenly, as I said, recognize, "Okay, we need automation."

Indeed, they founded an entire domain, computer science or informatics, around that. So, at some point, computers were developed by electrical engineers, and then at some point, the community recognized, "Okay, we need computer scientists. We need people who develop software, who develop automatic methods for that." And it's very interesting that, in quantum computing, we kind of see a very similar development. Because the first ideas of quantum computing, of course, came from people from theory, then physicists have built the first quantum computers.

And now, we more and more recognize that, in a similar fashion as we have seen that many decades back for conventional classical computing, we now also need people who can develop software for quantum computing. Who can realize certain applications for quantum computing. So, this parallel is really interesting to see. On the same side, we can also see the challenges. We have something in the classical computing world, what we call a design gap or verification gap, which means that our ability to properly design something and to exploit the technology we have is limited.

So, we have, in the classical world till today, a situation where we can develop or realize more powerful machines, than we actually can design useful applications and software for. And in quantum computing, what's interesting in the past years when I started on quantum computing, we didn't have these powerful quantum computers so far. We had very prototypical systems and they are still limited. But now, we see this development in the past years.

And my feeling is that we are approaching in the quantum world something that we have seen in the classical world: some sort of a design gap that we may very soon reach a situation where we might have more powerful quantum computers than methods and means to use them properly or exploit them properly. And this is sort of what we are working on, to avoid this design gap, or at least to keep this gap as small as possible.

Yuval: When people design abstraction languages, whether it's hardware description languages or the parallel in quantum, sometimes critics come and say, "Well, when you design an abstraction layer, you lose something. You lose the ability to take advantage of the unique properties of the machine." And today in quantum, various manufacturers have different properties, different architectures. So, is it too early to design an abstraction layer for quantum or too late? 

Robert: Very good question. And I have sort of two different answers to that: On the one hand side, right now, indeed, we are in a situation where we have to have a close interaction with the physicists who are building the quantum computers. I consider myself a computer scientist and, in this regard, I like abstractions. Because although, I have a rough understanding how the physics of quantum computing work, I'm certainly not an expert in quantum physics. And in this regard, I rely on abstractions because that's the only way how I can exploit my background in computer science in order to tackle the complexity. And right now, you're absolutely right that, nevertheless, we need to be close to the physicist to understand the concerns. Because today's machines are still limited and we really have to try to respect and to satisfy the physical constraints as best as possible.

However, in the long run, if the physics community is making the quantum computers better and better and more scalable, more reliable, less error prone and stuff like that, I'm pretty sure we need more abstractions. And again, the nice example is when you just check out how development goes: we have this exponential growth in the classical computing power. And right now, if you check out the roadmaps by IBM, Honeywell and so on and so forth, you see a very similar development. Sooner than later, I assume, we're reaching a situation where complexity is increasing rapidly. And then, one of the main ways of coping with this complexity is abstractions.

In a similar way, today, we're not building, working with transistors anymore in order to realize an application. We will have, in quantum computing, also a situation where we will use high-level languages and automatic tools and compilers, which might not be that machine-specific anymore at certain abstraction levels.

And I think it's totally okay and worthwhile to do research on those abstractions already now, so that we are prepared for a situation in which complexity grows so significantly. And for a situation, we might be able to handle the physical requirements on those lower levels and we can abstract them on the higher level. So, I really believe both are necessary. We need to cope with these physical details today, but we should also prepare ourselves that we can abstract more from it in the future.

Yuval: On classical computers, all these abstraction layers already exist, right? If I design a webpage, I don't need to know how CMOS transistors work. And there are many, many software layers between the web page and what's actually runs on the hardware. So, that's good. And let's assume that that was also the situation with quantum computers, that you had an abstraction there.

But today, because of coherence issues and errors, many algorithms are written in hybrid classical-quantum way. Is there a special case of this abstraction once you get into hybrid algorithms or is it just two separate camps with a little bit of interface between them?

Robert: To be honest, I don't really know. I'm not sure where this whole development will lead us eventually. Right now, my understanding, my feeling is: if you want to realize an application for quantum computing, you need an understanding about quantum computing. The way how we program a quantum computer, how are we going to use it, right now, still requires the corresponding designer to understand how quantum computing works in principle. Not physically, but at least methodologically. You need to have an understanding about superposition, entanglement, these kind of things.

We may be able to abstract away from that in the future, but at the moment, I don't see how this could really work out today. And this is, I guess, also one of the large challenges when we talk about in the community about what do we have to accomplish in order to establish quantum computing.

Of course, we need physicists who build us the computers. We need the software and so on and so forth. But we also have this huge discussion, you're very familiar with that, on the quantum workforce, right? So, we need people trained for this particular community. And right now, my strong feeling is today, we still need these highly trained experts. We need to educate people. They have to have a background in quantum computing at some level of abstractions. Maybe in the future, we will reach a point where we can abstract away from that, but I do see this more in the midterm or long-term future.

Right now, I don't see anyone who is capable of implementing or realizing something, an application of a quantum computer, relying solely on conventional classical HDLs, for example. You still need some sort of quantum description means and an understanding of that.

Yuval: You are a professor at a Austrian University, and I see that in the US, certain universities are starting to offer Masters in Quantum Information Science or a PhD program. Could you tell me what the situation is in Europe with regards to educating the workforce on quantum?

Robert: It's becoming an issue. So, people start bringing that into their lectures and stuff like that. I said I work in quantum computing for 10, 15 years now. And particularly in the Master programs, I already had lectures, which I called emerging technologies, where I covered topics like quantum computing. And this because I always felt like,  the young generation who is pursuing their Masters or Bachelor degrees, maybe their PhD degrees now, they, most certainly, will be exposed to quantum computing because they have their entire life in front of them.

I kept saying to my students, "I am a professor. I've got tenure." I probably could say, "Well, I don't care anymore because I'm done more or less with my career. I'm just doing it because I'm still have a passion about it. But you guys, so my students, you still have lots of decades in front of you as computer scientists. You should be aware of what's going on." And I'm doing this only for 10, 15 years.

And now also, in the universities, programs are built up. I know also in our university, we are putting a focus on that. Literally, every European country, probably every country in the world, has identified quantum computing is becoming a thing. How established it's going to be? Nobody really knows! But we know that people should be educated about it. I'm not saying that we are completely changing our curriculums these days, so we still teach our students these conventional classical programming languages. Which, of course, makes sense because we're not talking about quantum computing is going to replace conventional computing. We're going to have just more variety of computing paradigms to be used and quantum computing is one of them. And we are teaching our students how to deal with that.

At least, we are beginning with that. So in my lecture, I cover things like, "Okay, how do you use certain tools, like Qiskit, like Cirq and so on and so forth." And realizing these toy applications and examples they have. But I also have to admit that we're not realizing fully fledged applications yet. But I'm pretty sure in the next years also my own lectures will change in the sense that we also have not just toy example to be playing around with, but real world practical, relevant examples, our students will have to implement on a quantum system. So, this is certainly changing. And in universities, this is a thing these days, definitely.

Yuval: And keeping on the European angle, when you look at the big industrial companies in Europe, do they rely on the universities to provide the manpower that they need for the quantum computing exploration projects? Or do they do something else today?

Robert: Good question. I would say a little bit both of that. Right now, you can really talk to our company partners and when I check out what's happening in particularly in Europe, I see lots of companies who say somehow, "Okay. I heard about this quantum computing thing. Can you tell me what this is about?" So, companies who are not really want to make their hands dirty yet, but who feel like, "Okay, I probably should be informed. Am I missing something?" So, we have, definitely, companies like that, which is fine. I'm not saying everybody should jump on the quantum computing train right now.

But when I started 10, 15 years ago and I did this quantum computing thing, I was the crazy guy with this future technology which is so far in the future. So, I'm already happy that right now, we got feedback, particularly from companies, who are saying, "Well, this seems to be a thing. Can you teach me? Can you tutor me? Am I missing on something?" This is one way. And then there are other companies who say, "Okay, we want to be involved right now."

And so, you asked the question whether or not this is driven by universities or academia or by industry. I would say as how I see it, that companies are now also starting and taking, using money to fund, for example, universities or research projects or to fund their own research laboratories within their companies.

So, I have a little bit feeling that both government and industrial funding on quantum computing significantly have increased. I guess, in an effort to prepare themselves for a time where quantum computing gets established and to be ready for that. So, I wouldn't say it's only academia. I also see a lot of companies investing in academia to develop this field. But frankly spoken, in Europe, I'm not aware of any kind of success stories in the sense like, "Okay, we have a European company who is actually making lots of money with quantum computing." But they're investing and I'm pretty sure they are investing because they expect maybe in a couple of years, some sort of return of investment.

So, I would say both government and industry are significantly investing in the technology and trying to establish Europe as an important player in the world on this technology.

Yuval: As we get close to the end of our conversation, let's assume I gave you a magic wand, and now you can control the work plan of the large software and hardware companies and the small ones as well. What do you want them to work on for the next two or three years in quantum?

Robert: It's a good question because, literally, I'm not sure if I should say that publicly, but I have a little feeling that what the society, the community, industry, governments are currently doing, is to be honest, perfectly in line for what I could wish for. Because, again, 10, 15 years ago, I started working on quantum computing and I was this crazy guy working on this future technology. And I had fun and I really enjoyed that academia is making it possible to work on these emerging technologies.

And at that time, I really would love to have this magic wand to sort of convince people that what I'm doing actually has a relevance, practical relevance. Right now, I don't see that I need this wand so much because everyone around, whenever I talk to industry, to governments, or the funding agencies and so on, I have the feeling that I don't have to convince anyone anymore.

And now, suddenly, topics I worked on in the past 10, 15 years, suddenly gain relevance. So, maybe if you allow, I would reserve this wand maybe for future days, when I really need it. Right now, I'm just happy that this entire community is developing in a direction where I'm really happy with. Of course, we have to be careful that we are not overreaching it. And there are still a lot of challenges. And I mean, people talking about maybe a quantum window that with the expectations are too high.

But right now, I'm really happy with the community is evolving and what's going on. And it's really fun to really see what's popping up. So, as said, I'm not sure if I should say that publicly, but right now, I would say I'm really happy with the development as they are so far. So, maybe let's keep it this way.

Yuval: I'm grateful that crazy guys like you have been working on this for 10, 15 years, so that we are where we are today in really, what I think, is the start of an amazing revolution. Robert, it's been great speaking with you today. Where can people get in touch with you to learn more about your work?

Robert: The easiest way is probably just Google my name. Most likely, you will find my webpage. It's www.rwille.de. Or you can also follow me if you like on Twitter, which is my Twitter handle is @rbrtwll. And looking forward to hearing from anybody who's interested in these kinds of topics.

Yuval: That's perfect. Thanks so much for joining me today.

Robert: Thank you very much for having me. It was fun.