The Power of Data Podcast
Episode 84: Discovering The Power Of Quantum Computing
Guest: Dr. David Snelling, Program Director of Artificial Intelligence at Fujitsu
Interviewer: Anthony Scriffignano, SVP and Chief Data Scientist at Dun and Bradstreet
Hello and welcome to the power of data Podcast. I'm Anthony Scriffignano SVP and Chief Data Scientist at Dun and Bradstreet. Today, I'm thrilled to be joined on the podcast by Dr. David Snelling, Program Director of artificial intelligence at Fujitsu. Welcome, David, how are you today?
Oh, not too bad. It's been a good week.
It's really good to have you here. It's unusual that we're all getting very comfortable with virtual conversations, and we meet people that we wouldn't have an opportunity necessarily to meet in real life, and certainly this is one of those cases for me - so, I've been really looking forward to talking to you today. By way of introduction to our audience, could you just give us a little bit of an understanding of your amazing background and what it is that you do and your role at Fujitsu today.
So, as you probably can detect from the accent, I'm not native to the European environment, but came to Europe in about 1980s, was then what we would call like a .com era, but for supercomputing and the .com supercomputing company I went for did what .com supercomputing companies do and went bust, stranded me in Europe; spent a little while consulting in high performance computing at the time, and then went to teach in academia at the University of Leicester in Manchester for a while, before joining Fujitsu some 25 years ago. In Fujitsu, I spent most of those 25 years working in the laboratories, started again, of course, with supercomputing, did some grid computing, which is a precursor to cloud computing, did web services in some green data centers artificial intelligence, obviously, from my title, and now I'm involved with the quantum computing strategy within Fujitsu as well. So, basically a nice large mix of very different things to do, and now I've been with the business side of the function for about the last five years.
That's a fascinating journey, and I'm smiling while you're talking about it, because it's all the cool stuff.
You know, you start talking about high performance computing and supercomputer; you know, what's happening, at least from my perspective, is that we have to keep inventing new terminology to describe things. What we call high performance computing, at the time that we coined that word meant something very different than what we're going to be talking about today. So now, what do we call it right? Higher performance computing, it's a different thing. And we struggle with the fact that this terminology that we use, sort of has general and then very specific meaning in particular context, and the context of quantum computing, which we want to talk about today, we have all kinds of new language, we talk about superposition and entanglement, we talk about all kinds of real world problems that are very hard to understand, if one doesn't have a background in physics, the sorts of things that are at the heart of these quantum computers are very different than the zeros and ones and bits and bytes that we're used to in classical computing, and so invariably, when we start to talk about quantum computing, we have to sneak in terminology like I just did in order to lay the groundwork to even be able to talk about quantum computing. So, having said all that, what are some of the real-world challenges that you face other than language in talking about quantum computing, evangelizing it, and really addressing the day-to-day challenges that are in front of us.
One of the things that I like to think about in terms of quantum computing is that we would think that it was a scientist’s dream, but it is no longer a scientist’s dream. It's a reality, but it is still a nightmare for the engineer.
And so, one of the biggest challenges as a consumer of quantum technology, if you will, that we face, is actually the state of the science and the engineering at the moment. So, we are in a kind of a tipping point where the technology is possible; we know it works, but it's not yet to the place where we can use it on a day-to-day basis. So, it's back in the early days of the transistor, where you had to be able to use a soldering iron in order to make use of a transistor. That's where we are now with quantum computing.
Except now we have to start by creating an environment that's colder than outer space.
Well, yeah it is - I mean, if we go into the technical details, but quantum state is very delicate, and in fact, you mentioned and outer space, actually, outer space is way too hot, right, because the quantum state only exists, and then only for a short period of time, at a few micro Kelvins above absolute zero, just ever so cold; and that means that you have a difficult and complex environment that you have to maintain, in order for the computer to work. It has to be shielded from any kind of magnetic radiation or vibration, or just don't sneeze, okay? I mean, coughing is one thing, but definitely don't sneeze in the vicinity, or you'll ruin your quantum state. And so, with all that comes a whole new stack of technologies and capabilities that you need for care and treating. So those are some of the challenges that come with it.
And then there's the inside of that thing that you have to create, and then you have to somehow observe the answer on the outside, so, there's the very real problem of communicating with what I'll call the inside and the outside, which is all different set of problems that have to be solved. So, this is one of those onion type problems that every time you un-peal you get to an problem that seems almost impossible, but almost impossible is a synonym for possible, if you have the right mindset.
And a lot of determination, and we're definitely seeing the determination, now people pushing the boundaries to try and get from where they are today to actually making use of these devices.
So, you mentioned sort of passingly there, that it's also an issue that one has to interact with these states very quickly, because they are not persistent, they are not in any way, something that you can just go away and come back and things will be the way you left them. This is not an environment that has a good analog. Ironically, when we had classical computing, we could say, well think about on and off, and zero and one and plus and minus, and we can conceptualize that even though the real world doesn't behave like that at all. Quantum computers are very much like the real world, which is inherently very hard to describe. So, the types of problems that we address with quantum computing are different in an unusual way. Can you talk a little bit about how you have to think differently in order to get into, let's assume they work, and assume we have these things called quantum computing machines, environments, whatever you want to say? What types of problems do we work on? We're not gonna have a quantum word processor tomorrow, right?
No, and we probably wouldn't want one, I think one of the things we have learned already about quantum computing is that it will guarantee the long existence of classical computing. Okay, yeah, we're going to be using our silicon chips and our laptops for very, very long time. So, I'd like to do a little bit of a hierarchy of the way in which we can deal with quantum computers. Okay. So, first of all, you've got the whole category of what are called commentarial optimization problems. These are those problems that individual parts of a problem are very easy to solve, but there are just so many different options that if you were to try one option, every nanosecond, the universe would still free solid, before you finished even halfway getting to the answer, huge amounts of activity. Now, quantum computers are designed to solve those kinds of problems, they're designed to look for them. And in fact, some of the more, shall we say, larger quantum devices, called Quantum annealers actually solve that particular kind of problem, they only solve commentarial optimization problems. And so those will be one category of problems that are already describable already programmable for the kinds of devices we're expecting to see over the coming years. That's the subclass of quantum annealer. Now, the next class is what we would call the quantum gate machines, and here are the problems that come into two different interesting categories. One of the problems is because the quantum gate machine is actually a quantum system, if the problem you're trying to solve is, in fact, a quantum system, like say, you're trying to design a drug, or you're trying to design a new material property, you can actually build the computer to emulate that quantum system that you want to solve, and then turn it on, and it will give you the answer. It's like the days of analog computers where we actually built circuits that behaved like the control system of an aircraft, so that we could simulate the aircraft that will be one of the early wind situations for quantum computing. And then the final analysis is when we can actually build quantum gate machines. And then we'll be able to solve problems and start to move out of quantum science and you know, material science and so forth, and into things like finance, logistics, possibly machine learning.
So, I want to try to pull on a few threads of what you just said. I can imagine some folks that might be listening to this might be hearing a lot of language that feels very abstract. What are some of the things other than you know, you talked about drug discovery, you talked about decay type problems, quantum annealing, if we are just thinking in terms of if I were a business person, not a computer scientist, and someone was coming to me, and they said, we have a fully available gate machine that you're talking about, where we have one of these very large scale quantum annealers. Great, what kinds of problems in the real world that I face as a business owner, as a business leader, what are some of the things that I find either impossible or improbable to solve today, that now I can do?
Well, I think the answer to that question is in the way you asked it, because you asked as the business person, what problems can quantum computers solve that I can't solve today? Well, that's exactly the ones that they can, the problems that you can't solve today. Now, the problem with that is that it's a little bit of this unknown unknowns, because some of the problems that quantum computers will be able to address you don't realize you have yet because they're just beyond the scope of anything you would ever consider asking about.
Let me see if I can suggest one of those to you just to be a little bit provocative. So, I don't want to get into I'd love to, but I don't want to get into all the detail around why but quantum computers with respect to factoring present a particular problem for encrypted data. A particular challenge I should say that things may be won't be as encrypted as we think they are in the future. So, the world is thinking now about quantum resilient encryption, what would that look like? That's a problem we don't have today. We don't need that today, because we don't have anyone going and using this technology to unpack what we're building today; but tomorrow, we're gonna have that problem. What do you think quantum resilient encryption looks like in the future.
So, the two things about the quantum encryption space that that are worth highlighting. One is that when we can crack our current encryption algorithms, and so forth, it won't be the current information that is at risk, it will be the information you're generating today, that will be at risk. So, all the secret conversations that you're having now are the ones that somebody is storing away, because you think they're safe, because they're all encrypted. And then somebody will come along when they do have a quantum computer and have a look at that conversation. Now, if that conversation happens to contain your credit card, you're probably all right, because that changes every two or three years. And so it's really pretty much irrelevant issue to worry about. If, however, it's the name of some secret spy hiding somewhere, then the longevity of the information becomes the issue that you need to worry about. Now, there are a number of algorithms being developed by the National Institute of Standards in the US long running competition, to develop quantum resistant algorithms for encryption, that's one avenue. The fact that there were some 14 submissions and they're down to three, doesn't actually build a lot of confidence that the three that are left are going to be that good going forward, either, because nobody has had a chance to sit around and try and crack them using the tools of quantum computing. So, we don't know. There are other ways to do this, you start with quantum safe communication using quantum communication technology. That's another topic, but that's not so much quantum computing that's using other quantum properties to solve the problem.
That was a perfect segue to my next question, which is, let's assume that that problem of communicating safely in an encrypted way is a persistent problem, no pun intended. What about quantum communication? What will that look like? What does entanglement bring to the table that doesn't exist now? And how might that play out in the future? And maybe we should explain entanglement?
Okay, so it's first a little bit on quantum properties. There are sort of two major properties that quantum computing exploits, one of them is superposition. This is the classical Schrodinger’s cat is both dead and alive at the same time. But the important thing about it in terms of quantum computing is if I have a memory of say, just four bits, because the bits can be in a superposition of zero and one, any computation I do on those four bits I'm actually doing for all 16 possible values. And now you can see what happens when you start to get 60 to 100 qubits, then you've got lots of different parallel computations all happening at the same time. And the other property is entanglement, which is what Einstein described as spooky action at a distance. And what it is, is if you have two quantum particles, if you will, that are entangled with each other, to separate them by a galaxy or two from each other, when you look at one of them, you will know what happened to the other one. And that capability basically gives you the ability to build a computer, that doesn't have to worry about the time it takes a signal to go across the machine, you've got a whole new kind of challenge with that, if we come back to the quantum encryption and communication, using these properties, particularly the entanglement property allows us to do something called quantum key distribution, okay. And it's the ability to use a quantum channel to pass a key that you're going to use for one conversation only, and then you're gonna throw it away. So, it's a rich random key, this is safe, because no quantum computer in the world can crack a key, that's as long as your conversation because it could be anything, it could find all the possible conversations, and you wouldn't know whether it was this podcast or my bank account that you have go by. And so, the thing that you get with the quantum communication is the ability to verify that the key I've given you that we're going to be using for the next 20 minutes or something has not been seen by anybody else, the problem is that that's all it guarantees; and that means that if somebody listens in listening turns out to be a successful denial of service attack just by listening, because until I know that I've sent you a safe key, we can't start our conversation. So, there are lots of stacked problems, from encryption to the challenges of quantum communication, to authentication, to non-repudiation that just keeps stacking up as we go.
There's no easy button here.
There's no easy button here. And if you're looking for a career in data security, you're onto it. Go for it.
Yeah, this field is wide open for innovation. If anyone's listening to this, that is either young and entering the field or has young entering field type people in their life that they can influence. This is for sure, a career and it's an exciting career. It's rare as human beings that we have the opportunity to see the birth of something truly revolutionary and thought evolutionary Yes. revolutionary, not so often. This one's revolutionary, we need new language to talk about it, we need new ways of thinking about the problems. It's a really, really, really rich field, and one which is changing way faster than any smart people can completely understand, which is why it's so important to have these types of conversations. I want to ask a slightly different question. We alluded to it at the beginning, quantum computing does not push away classical computing, there's an almost guaranteed marriage here, where there are systems that are hybrid systems where quantum computers do what they do best, and classical computers do the other stuff, and they work together. What do those sorts of problems look like in the future.
The thing that we've discovered with the hybrid kind of solution is that we always talk about our problems. And more and more today, we're talking about our problems in terms of large scale data, huge amounts of data, whether that be satellite images, or tweets on the internet, but huge amounts of data that need to be processed. The quantum computer, whether it be a quantum annealer, or quantum gate computer, is not the data crunching part of the problem, right? classical computing will be to take information that's out there, that's sort of the information capture side, whether you use Internet of Things, or whether you use satellites, or whatever, to capture the data, then there's a phase called data analysis. And that's where classical computing lives, I think what happens next is you want to start trying to use that knowledge that you've gained the understanding the machine learning the artificial intelligence, and then you need to make a decision on what to do; and it's when you come to that decision making, that you've got so many options that you cannot address that problem by simply asking. And so, the quantum computer sticks on the end of that. And I can tell you which one of those billions of possible answers, combinations is the right one.
Yeah. And you know, since this is the power of data that we're talking about, I think it's a very important point that data processing the discovery of curation of data, the cleansing of data, the manipulation of data, not so much in the quantum world. However, when we think about the problems that are brought about by big data, all of those V's that we don't talk about anymore, volume, velocity, veracity, variety, value, all of those create overwhelming computationally complex problems of analysis. In the AI space, we can build hybrid systems that literally try 1000s or 10s of 1000s of options and compete with each other to get the best answer. But that space is expanding at a rate that's arguably unmeasurable right now, we're not going to be able to use these tricks much longer because the data will overwhelm the problem, and then confirmation bias and we'll start to things start to happen. And then when we get into a situation like we're in now, where the world is disrupted, all of that data from the past doesn't look like the system we're trying to understand. Now, what do I do? So, I think that quantum really gives us a tool that we don't have today, which is to deal with things that are computationally overwhelming in the time that we have to work on.
And the key is in the computationally overwhelming, because it'll be a long time before quantum computers can give us any real advantage on the scale of the data problem that we're facing. I mean, you alluded to it earlier that getting information into and out of a quantum computer is a bigger task and a huge engineering task today, and it's likely to stay that way for some time to come.
So, we did talk in passing about drug discovery, and some of the other areas where there's some really exciting promise, where are the areas that give you hope? That maybe by bringing this technology forward, some real human problems, some real-world thing will turn into something more helpful?
So you have one of them. Obviously, in the past couple of years, we've learned a lot about the importance of understanding our natural environment, in particular, the little tiny microscopic virus environments that we live in. So, the biomedical area for research is definitely one that will benefit from the quantum transformation that we're facing in the near future. Other societal problems that really are a high focal point, or anything that comes around logistics and management, the number of combination challenges that we're facing today, where do we have the data, we have the information, but the choices are just so innumerable, that we don't know how to make the right trade offs, to do the right thing for the right communities. Those kinds of large scale application domains and logistics, etc, are one of the things that we're definitely going to be seen as an advantage in the future with quantum computing.
A lot of the analysis that's done right now on things like supply chain disruption and resiliency, it's all first order, maybe second order, because then people's brains start to get foggy. So, you can understand if you look at your supply chain, you can understand your vendors and your customers - but when you start getting to your vendors, customers and your customers vendors and take that out another step and another step, too many choices, too many if thens, and really the error overwhelms the analysis, so you stop.
It's an interesting one to highlight because it is because those problems are in particularly the computational optimization ones that I'm most familiar with, because the work I've been doing lately. Those problems are explosive, with what's called an exponential curve, you get to some point where instantly gets too big to manage. And it gets to there so fast, but there's a gap between what is trivial, and what we can grasp as a human being, because we are actually very clever creatures. And so, we put a human being on it and a human being can deal with the first three levels of supply chain optimization, and just does it because he's smart, or she's smart, she's done it before. And you add one more layer to it, and it's impossible for human beings to do. And so the thing that makes it possible today with human intelligence is what makes us not look at those problems in terms of trying to solve them today. And there's a huge potential of going one step further. And that's when quantum will start to make the big difference.
So, some of the people that tend to go one step further, faster are the bad guys. Is there any quantum malfeasance that you envision in the near-term future?
So, the obvious one is encryption.
Let's just leave that off.
Okay, so you could find other places where you have an information advantage, of course, I'm a good guy. So, I haven't even thought about the bad things I could do with quantum computing.
The good guys have to catch the bad guys, right
So, the thing that we would be looking for are places where a bad actor needs to be able to make a more effective solution than what you can online gambling, for example, the first online gambler that has a quantum computer might be able to win the game every time.
Quantum card counting.
Quantum card counting lovely phrase. I'll borrow that in future, if I may.
Sure. The thing that comes to mind to me is the malefactor who figures out that they can use information like this to optimize money laundering, for example, the good news, I think, is that there are some algorithms within the quantum space. Right now Grover's algorithm, for example, that can help us find things that we don't completely understand. So, I could use quantum technology and quantum algorithms to observe the supposition that that type of behavior that was happening, as well as someone else could engage in it, I don't think there's any first mover advantage to the bad guy there. So, I think we're still okay for a while.
Yeah. So, the money laundering, anti-money laundering is an interesting example of one of the things we've alluded to and that is the different thinking necessary to do quantum computation. So, if I was trying to say, detect money laundering, okay, there are a whole batch of things I could do to detect money laundering, but what you might do in the quantum space is to think about the problem differently. And I want to find the most likely way for money laundering to take place within this network of systems. So, rather than trying to detect it, you try and determine what it's going to be before it happens, and therefore, then you can much more precisely place your bets. And so, this is a lovely example of how you might think differently in order to address a challenge in whatever domain you're in.
In my group, I call that proto quantum thinking, how do you need to change the way you think, to get yourself ready for a world where you have these capabilities available to address the problems that will then exist?
Absolutely. So, one of the things we're doing at Fujitsu, if I could do a quick advertisement here?
Is we've developed what we call a quantum value assessment. And the sort of the cheerful way of talking about this is we're going to train our customer base to be intelligent buyers of quantum computing, because the quantum computing today is definitely writing a series of hype curves. And knowing when your organization can make use of quantum computing, what you have to do in advance to prepare for that, be able to intelligently purchase your quantum solutions down the road, is something that we're not ready for yet. And we're working on strategies to help our client base become ready for that. We might not win the business at the end, but will know that whatever they buy, they will get what they thought they were getting, because they become an intelligent buyer. So, in addition to sort of the career in quantum computing that's available to a lot of young, very bright people, a lot of business people need to start thinking about quantum computing, not in terms of understanding how it works, but understanding what it means to their business, and how that would be impactful, and we're happy to help people do that.
Amen. And kudos to you for thinking that way. Einstein, since you brought him up, He said that, you know, things should be made as simple as possible, but no simpler, which I love. And he also said that we can't solve a problem with the same mindset that we used to create it. So, you know, this is a great example of changing our thinking which is an important critical function and change leadership. So, as we're drawing to the end, I could talk to you all day, I really appreciate you being so intellectually generous with your time and your passion here. What advice would you give to anyone that's still listening to us about how they should think going forward if they want to enter this field? Other than get really good at math? What can we tell anyone who wants to enter this field? Maybe two things get really good at math and don't panic? What better advice would you give them?
What additional advice, because I think those are two very good advice’s. The thing that is really important is to get comfortable with the idea of thinking differently, because you set up an environment in the quantum computer to solve a problem, you don't set up the environment to solve the problem you're looking at. You set up an environment that says when this environment is correct, that light will turn blue, or red, and depending on whether it turns blue or red, I know what my answer was, it's thinking upside down about the structure of your question. And then you run your quantum computer and 9999 times out of 10,000, the lights red, but once it's green, that's okay. If the lights red, that's the answer to my problem. And it's this other way of thinking that tolerates Well, probabilistic answers for starters; it might or might not be the right answer every time. So, you need to at least try a couple of times, and so that this other way of thinking is probably right up there with the math, not everybody has to do the math - but I think anybody trying to do quantum universe thinking needs to be thinking differently than the normal way that we think.
For sure. The one other thing I would suggest and be interested in your reflection on it is get really good at collaborating with other people. I think you just alluded to it, the space is too big, as too complex, you cannot be alone a pioneer, you have to work with other people, you have to get really good at saying what you think out loud, you have to get really good at talking to each other like we're doing today; and that sort of yes and, you know, embracing it's the scientific process. It's inviting dissent and testing what you think.
Absolutely. The nature of collaboration, testing problems, let me explain this to you – and, halfway through explaining it to you, you get the answer. You know, it's the classic software engineering, rubber duck, right? But you need to have those people around you just so that you understand it. You need to be able to explain it to people and trying to explain a quantum algorithm to somebody will double your understanding of your own algorithm.
Well, this has been an absolutely fascinating conversation. I can't thank you enough for taking the time to engage with me. It's been definitely the highlight of my day and for sure my week. I hope that you've taken something from it and I guarantee our listeners as well, thank you very, very much.
It's been my pleasure.