Hartmut Neven, Engineering Director at Google, founder and manager of the Quantum Artificial Intelligence lab, has expounded on the development of quantum technologies saying that there are various ways to apply quantum computers in the future, however, it might take 10 years before we could witness applications that would matter to ordinary people.
Sputnik: You said in an earlier interview that quantum supremacy is a turning point. It will do what was previously impossible. However many researchers doubt that they will be able to outdo classical supercomputers any time soon in terms of productivity. So what kinds of quantum computer tasks are you developing that will be solved if quantum supremacy is proven by year’s end?
Hartmut Neven: So, quantum supremacy, indeed, deals with a benchmark problem which is not related to applications; it’s only there to set. On certain tasks, a quantum computer is much faster than a classical machine; that is the goal we would like to achieve. Once you are fast on some tasks, you can go to the next phase and ask the question what would you put this new computation capability to use for? And there we have various ideas. At this point, we only know of one application for sure we can do – and this application is called 35 random numbers. This is a production of random numbers that you can receive from the server, and you don’t have to trust the provider of this server to make sure you get fresh random numbers. That has various uses in information technology.
Sputnik: If quantum excellence was to be achieved, how can it change ordinary people’s lives?
Hartmut Neven: Besides the application certifiable random numbers, there are a few other areas where we hope we can apply quantum computers to in the near term. I mean not in 10 years, when we hope to have an error correcting machine. These applications are in quantum simulation, in machine learning and optimisation. I should explain this a little bit. In quantum simulation, you are trying to, as the name suggests, simulate systems where quantum effects play a role. You asked why ordinary persons should care…
Let’s say, batteries are for electric cars, today those batteries are very heavy and it takes a lot of time to charge them. So, car companies would like to have much lighter, faster chargeable batteries. But because the processes involved in a battery, like quasi-electrons flow from cathode to anode, there’re quantum processes involved; and therefore, if an engineer has a new idea, a better way to construct a battery, they would have to physically build that battery, take it to the lab and measure it. That’s a very slow process; therefore, only a few hundred catalyst materials, also catalyst electrolyte materials, get tested every year.
If you had a quantum computer, you could check with your computer millions of materials and take only the most promising ones to the lab; so, you would dramatically accelerate chemistry materials research, the development of pharmaceuticals. And that is, of course, something that ordinary people will feel; if you have more efficient solar cells, more efficient batteries, catalysts in industries that make plastics that are more biodegradable; many applications are on those lines.
Sputnik: When may Google announce that quantum excellence has been achieved? Is it even possible to put a deadline on such a task?
Hartmut Neven: That’s a kind of a million-dollar question. We hope that we might be able to do something application-related over the next few years. But the more conservative estimate would say that we have to first invent error correction, and that is about 10 years away. So, if worse comes to worst, it might still take 10 years before you see useful applications that would matter to ordinary people.
Sputnik: What are the main obstacles to reaching quantum supremacy?
Hartmut Neven: To reach quantum supremacy you need to have a number of q-bits; that number is not so large, it’s about 50 – and various labs and companies have processes or systems with more than 50 q-bits – but you also have to be able to control them very precisely. That is the challenge that [we have]; so far, people haven’t managed to do that.
Sputnik: How can the development of quantum computing improve AI algorithms?
Hartmut Neven: One area is in machine learning, which is the main area these days of AI; which is essentially teaching a computer to do a certain task. Let’s say, you wanted to have the computer detect whether or not there’s a face in an image. You train it a little bit like you would train a child; you give examples with pictures, like “here’s the face”, and then you have images where there’s no face. And essentially what such a machine learning programme does is that it has many knobs that you can turn. Initially, you don’t know how to turn these knobs well for the machine to do the task right; so, you present the first example and, let’s say, by chance the learner gets it right and you say: “that’s great, let’s keep this setting". Then you bring the next example and, assume, now it’s wrong; you say “let me tune it a little bit differently so that next time I present this image you will do better". Then you keep repeating this process, you show images, you see whether it does it right or wrong and accordingly you tweak a little bit those knobs until it gets it right. It’s a very slow process, you need to present many examples; mathematicians call this process an optimisation process. Quantum computers have been proven to be better at this process; so if we have a working quantum processor, we can accelerate optimisation, which means we can accelerate machine learning or training, we can accelerate that part of machine learning.
Sputnik: What are you working on now in your lab?
Hartmut Neven: We have called out quantum supremacy as an important point, not an endpoint, we’ve singled it as a milestone along the way, it’s a way post. Sometimes people say that quantum supremacy is just a marketing gimmick that the Google team came up with, but that is not correct. It’s actually a valid engineering milestone, which means you are much faster on one task than any classical computer. And once you have achieved this goal, you can ask a more difficult question, which is whether you can also be much faster than classical computers on a useful task. First, you will need to show it on some task, and that is something we still would like to achieve.
Sputnik: Are there any negative effects of any of your creations that you regret and that you never anticipated?
Hartmut Neven: In my career, I’ve done many different things from autonomous robots and self-driving cars to neural networks, to face recognition and quantum computing. And, of course, that’s kind of interesting to see how technologies develop; for example, it was one of my companies who did the first face filters – when you take the picture of the face and then you can change it – and these days all kids in the world have fun doing this. I would have never imagined this; I thought it was a little gimmick. And the very first of these applications sold was done by a company in Japan, and I wouldn’t have anticipated that.
Sputnik: Do you feel like your inventions have changed the world? That the world will never be the same...
Hartmut Neven: I wouldn’t think so. I think that any of things that happen, happens always through a collection of many humans; and every single human only does a slight element in the evolution of things. We are all co-creators of the world around us; each of us is just a tiny bit.
