In 1935, the famous physicist Erwin Schrödinger was debating with his friend Albert Einstein about the nature of a fundamental concept in quantum mechanics – a field that was, at the time, still very new. To illustrate his point, Schrödinger proposed a thought experiment wherein a (rather unfortunate) cat sealed in a box is both alive and dead simultaneously – up until the moment someone opens the box. Decades later, that abstract paradox is still very much alive, and enabling the development of a new generation of computers.
These quantum computers use bits (called qubits) that, unlike the binary bits in today’s electronics, can simultaneously exist in many states between on and off. And although the word gets overused in science, this emerging technology really is revolutionary. A fully developed quantum computer is predicted to be able to perform calculations that would be impossible for a traditional supercomputer, even with thousands of years of processing time.
In this episode, our experts chat about the current state of quantum computers and explain why the mind-bending theories of quantum make coming to work a lot of fun.
Irfan Siddiqi is a professor at UC Berkeley, where he leads the Quantum Nanoelectronics Laboratory, a collaborative group dedicated to developing new and improved superconducting qubits. He is also a faculty scientist at Berkeley Lab, where he leads the Advanced Quantum Testbed and the Quantum Systems Accelerator – a DOE National Quantum Information Science Research Center.
Zahra Pedramrazi is a project scientist at the Advanced Quantum Testbed. During her physics undergraduate, she took a quantum class with Irfan, and became hooked on the field. She is currently focused on the fabrication of superconducting qubits, working to refine their design in order to overcome the limitations of current qubits.
"Thus, the task is, not so much to see what no one has yet seen; but to think what nobody has yet thought, about that which everybody sees." ― Erwin Schrödinger
“How wonderful that we have met with a paradox. Now we have some hope of making progress.” ― Niels Bohr
You're listening to A Day in the Half Life. I'm Aliyah Kovner at Berkeley lab, and in this episode, we're going to hear from two quantum computing researchers. Right now, quantum computing systems are a super hot topic because of the immense potential they have to usher in a whole new world of computing. Truly a quantum leap forward in technology, if you'll excuse the 90s pop culture pun. This conversation is going to focus more on the present and future of quantum computing, rather than its history, because the field is so young. Quantum computers started out as a purely conceptual entity back in the eighties when physicist Paul Benioff first proposed the development of such a device, which could theoretically perform calculations that are impossible on a conventional computer. But it wasn't until 1998 that a prototype -- a tiny prototype -- was built and not until 2017 that a commercially usable quantum computer was completed. Because quantum computers will have a lot of applications in cybersecurity.
The current rush to develop and improve quantum computing systems is reminiscent of the space race of the 50s and 60s, with institutions and governments around the world working hard to have the best technology. But these systems will also have a lot of other uses and will be used very collaboratively, like for vastly accelerating drug discovery or improving cosmology simulations of things like black holes and the big bang. The core of a quantum computer, and the reason why they are exponentially more powerful than a traditional computer is the quantum bit. The computerized devices in your home, including whatever you're listening to this podcast on, are made of binary bits, little electrical switches that encode information by being either on or off. But quantum bits can exist in a huge variety of other states in between on and off, allowing them to store way, way more information.
And these computers are processing that information in ways that more closely approximate reality; after all, the subatomic particles that all visible matter is made of also exist in these in between the states. However, physicists themselves still don't fully understand quantum physics. So building computers that operate off quantum principles is basically one part engineering, one part philosophy. Special disclaimer, if this sounds confusing, that's because it is! Unless you're a quantum physicist like our guests, it's hard to wrap your head around how these systems work. In all our episodes so far, we've talked about some mind-bending science, but this topic might take the top prize. Luckily, our guests do a great job explaining.
So let's jump in. Our established guest is Irfan Siddiqi a researcher at Berkeley Lab, UC Berkeley professor, and director for the nationally funded Advanced Quantum Testbed. Our early career guest is Zahra Pedramrazi, a former student of Irfan's who caught his love for quantum physics during her undergrad. And now, following her Ph.D., works with him at the test bed.
Okay. So to start out, I know that this is a very complicated field, that isn't very easy to boil down into basic concepts, but how would you define a quantum computer? What is it in the simplest terms you can describe?
So thank you for the question Aliyah. What is, in fact, a quantum computer? A quantum computer is very similar if you like to your classical computer in that it's a device that is set up for advanced mathematics or for computations. However, the architecture at the level of the individual constituent parts is fundamentally different in a classical computer. We have transistors. These are basically switches that go between zero and one and these switches, or bits, that are there are all independent. So one goes ahead and tunes each one of them, one at a time to zero in one. In a quantum computer, the quantum bits, if you like are such that they periodically need to talk to each other and you can actually talk to all the bits at one time, making a very complex array that in fact can execute certain computations in a very efficient manner. And that's fundamentally the difference in the coupling between the bits in a quantum computer versus a classical one.
And can you tell me a little bit more about the differences between a classical computer made with semiconductors and a quantum computer? I understand that a classical computer can only exist in one of two states, but a quantum computer can exist in many more states. Is that a sort of an approximation of it?
Absolutely. You're, you're, you're hitting on the heart of, of quantum computation. And in fact, quantum information processing is the science of putting knobs on entanglement. So the idea is that each one of these bits is a quantum mechanical object, just like Schrodinger's cat, which can be both asleep and awake at the same time. Each one of these bits can be in combination of zero and one at the same time.
And one way that we can think about it is to imagine a sphere where one pole at one pole, we have a state zero and at the other, we have the state one. So remember what defines this state in quantum mechanics is probability. So our qubits state has some probability to exist in a linear superposition of these two states of zero and one. So going back to the sphere image, it means that our qubit state, instead of being at the poles only can also be anywhere on the surface of this sphere. So instead of having either zero or one, now we're in this probabilistic regime.
And this also has an effect on the information propagation that we have in classical computers and quantum computers. So when we, when you have an array of bits in a classical computer, each bit can be zero or one, and they don't really interact with one another or the environment. Now, if we have an array of qubits for a quantum computer, it's a different story. You have these spheres sitting next to one another, and the qubit states between them can now interact with one another as well as with the environment. So this is called quantum entanglement. And so the information propagation in quantum computers can can now be quite leaky since these interactions can be unwanted and they can cause information loss. So the challenge in quantum mechanics is how to control this information flow because sometimes we do want the qubits to interact with one another and that's how we can do different computations. But sometimes we don't.
Yeah. And to riff on what Zahra said, once we put these knobs onto these quantum systems, they have a tendency to become classical. And they don't want to exist in this superposition state. Very rapidly, they want to go to either zero or one. And so that raises an interesting question. What are the right materials for actually putting together and building a quantum computer after all, everything in the universe, obeys quantum mechanics. So you should be able to make a quantum computer with anything, with any physical system would work. And in reality, we have manifestations of quantum computing systems that involved ions that are trapped in an electromagnetic field, neutral atoms that can also be trapped circuits that will have electrons serving as the bits, or even my field: superconducting circuits. And these [superconducting qubits] are much larger, but they collectively behave as, as a single bit. What's different in all of these systems is that they're coupling to the environment is fundamentally different.
Objects, which are small and isolated, for example can live a very long time where this quantum superposition or entanglement property is unperturbed for a long time. But by the fact that they don't communicate with their environment also makes it difficult for us to actually do processing with them. Not only can the environment not read them, we cannot read them. Right. And it's difficult to actually go ahead to get two of those bits to talk to each other. On the other end of the spectrum, electrical circuits are big, but they have short lifetimes. And in which case one can then think about what's the best combination for a quantum computer. Do you want these bits to be easy to program, but short lifetime or difficult to access, but long coherence? And we don't have the answer on that just yet.
So when you say lifetime, do you mean the total span of time that this device would function or in terms of like processing time?
You're actually hitting on a very important point. In fact, it's both of those things. So there is a hierarchy of lifetimes. One can start to think about if I have a zero state and a one state in my quantum bit, and I want it to be alive and dead at the same time. So zero plus one. When I start off, there is a lifetime associated with that zero plus one state going to let's say the pure zero state, right. And that's sort of the, the ultimate limit, right? At that point, your, your bit is no longer usable. Then there are times associated with each operation. The goal would be to do a large number of quantum operation flips of those bits, or logical operations, before this quantum of information decays.
I see. But well, I sort of see, so as it's performing the computation, you can imagine spheres with different, kind of being at different points along the sphere, but if you measure, but once you measure, they'll all go back to the pole.
That's right. So once you measure it, you have to bring back a classical reality.
So it, but there's like a split, split moment when you can record that information of each bit's state of being
That's right. And in fact, it's even more interesting that we don't even record it. We have faith that it's there. So...
[Laughing] So how do you know what they say?
That's a great question. And that leads into quantum verification validation. How do you know that your quantum system is the system you think it is, you know, that it's performed the computation that you wanted to, to perform, but to step back from that, just a, just a little bit, the idea is that physical systems can exist in a much more complex space than you can measure them. And that is really the critical idea of quantum mechanics saying that you know, the cat can be in all sorts of different combinations, you know, asleep, awake, et cetera. So these are all the possible states a physical system can occupy. You just cannot find it in all of them, at the same time.
You only find them in a certain set of states, which are stable, right? And those stable states are the ones that are on the poles let's say of the spheres, the cat being asleep or awake, right? So that complexity, which is underneath the surface of quantum systems, is what you're tapping in this computation. You want to run the computation in that space where you can access this larger manifold of states, but you cannot look. Because once you look, you, you go ahead and collapse it back to the classical world. So the way a quantum calculation is set up is you initialize all these bits. You let them run, you let them talk to each other. You let them access all these crazy states that are there. And then at the end, they give you an answer. Now, how do you know that actually the quantum system that you had is the one that you programmed in and what have you? So we run all sorts of diagnostic circuits saying that, you know, let's put in a set of circuits or inputs where the output will be different based on quantum physics or classical physics, right? And those, those give us the diagnostic tests that tell you that you actually had this degree of correlation between your bits and it, it manifests itself as a quantum mentor, which is different than a classical computer.
Yeah. And to add to that, the power of quantum computers comes from the fact that we can take, we can use these types of entanglement and superpositions and like different interactions that the qubits have from one another to have different pathways for the solution can take different pathways that was not possible for a classical computer. And so that is, that is the reason that, for example, for a large data set, that a quantum computer can can have a computation that is exponentially faster or predicted to be exponentially faster than a classical computer.
I have so many questions that are just like bordering on philosophy and too much into, um -
Oh I love philosophical questions, because I love those Aliyah, because quantum mechanics, philosophy technology, these are all three sides of a three-sided coin.
[Laughing] Okay. So I guess one thing that pops into my head while you were both talking, there is, so if you are trying to use a quantum computing system to solve a complex problem, and you, you input that, and then you look too early, so it's reverting to the classical on-off kind of bit, do you still get an answer? I mean, does a malfunctioning quantum computer still act like a computer or is it just not functional if you look too early?
Yeah, that's a great question. So for example, if, if the user looks too early or the environment causes decoherence what'll happen is the signature that one sees is, when you look at the output state, so you go ahead and measure all the bits in your register. You will see things that are called mixed states and the mixed states are how we describe classical outputs. So you will not see these fancy signatures of entanglement or, or coherence that is put in there. So that is telling you that the information that you had is going into channels that we, you were not expecting it to go into. So the short answer is no, you, you don't get a meaningful answer. It basically says, or you get an answer with much less certainty.
So I guess that the challenge that we have in quantum computing at the moment is the noise mitigation and error correction. So how can we control the noise and like decrease the noise and how can we correct for the errors that can collapse these states?
So in this context, what does the term noise mean?
Of course. So an important property of a qubit is a coherent is its coherence time. And which you can think of it as a time through which qubits is viable and different processes and measurement can be carried out on it. Is so longer coherence times are desirable. And what is detrimental, eh, to the qubit and the coherence time are the different noise sources. And basically what we mean of a noise in a quantum computer is any type of signal loss or additional signal that we don't want. Classically, you don't have much of a signal loss because there's not much interaction that you have with the environment, but in quantum computing, you can actually lose this information, which is what you want to measure, if your system interacts with something that is not desired. What I'm interested in is realizing a novel qubit design, uh, that can lead to longer coherence time and the way I'm going about doing so is through designing and changing different elements in the circuit of the superconducting qubits. And that makes them protected against some types of noise sources and meaning that those types of noise source can no longer cause decoherence. So a lot of my time right now is in the fabrication, where we use the fabrication techniques that are mature in semiconductor industry, such as wet chemistry and different lithography and deposition techniques to fabricate these chip scale qubit devices. So there's a lot of exciting work happening, yeah.
So that leads well kind of into my next question, that I was wondering: what is a quantum computer made of? I've seen a couple of pictures of them and they look like a sci-fi chandelier, basically is sort of the best way I can describe them. But what are these different parts made of? And these are, you're talking about current qubits and improving them, what materials are going into these sorts of devices?
Yeah, that's a great question. So, as we mentioned, we're working on superconducting qubits, and so we design and fabricate these qubits and their circuitry in house. And our qubits are 2D. They're made out of tiny structures in orders of less than a hundred nanometers -- so a few microns. And they're made out of aluminum and niobium materials on a silicon chip,
Just to give you a sense, a human hair is 80 microns. So that's 80,000 nanometers, so it's really tiny.
[Laugh] Ah, right.
And so these qubits operate at very low temperatures of around 10 millikelvin and we can achieve such temperatures using custom-made dilution fridges that can cool down our samples. And so we have our qubits, uh we package them, we cool them down in our fridges, but now the measuring and controlling aspect of it is a big challenge, since as you, as you know, we live in room temperature. So we have, we have lines that are carefully engineered so that they can go from, from room temperature to 10 millikelvin and we can apply and read microwave signals in an out of our qubits. And we also have magnets so that we can apply magnetic fields, which gives us another degree of control over our qubits. So we use a microwave pulses as well as magnetic fields to measure our qubits. And so it's a, quite a complicated fridge. [Laugh] And I think you might've seen the chandelier like picture of such fridges and it's a pendulum design. So it goes vertically down in temperature. So the top part is room temperature, and then it goes down to 10 millikelvin. So depending on the dilution fridge, we can either operate one chip or multiple chips, where each chip can be as simple as only having a single qubit on or have a multi qubit processor
And to add on to Zara's point, and your point Aliyah about the fancy chandelier, which I love this imagery. So the reason for this kind of complexity is really telling you about the density of information you can store it in the quantum versus classical world, that little quantum chip, if it had 300 bits, you would need more numbers than the particles in the known universe to describe that state.
That's how fast that complexity goes up cause it's like an exponential increase with the entanglement. And this is really an outstanding challenge of how to convert and interconvert between the quantum information, which is very dense at the level of your chip to classical signals that come out of the fridge.
Right. And so what is kind of the current state of quantum computers? There are, there are prototypes that do work or close to working? Kind of give me a snapshot of the field as of this moment.
So we have quantum computers that do work. We can do measurements, we can do computation. And recently there was a paper that demonstrated the first time that like quantum computers can perform a task that a classical computer cannot. But but the, the problem here is the fact that they're very noisy. And so there a lot of errors. And a lot of the effort is how can we reduce this noise? How can we have longer coherence times so that we can do more processes and we can do more complicated type of computations. The field is very interdisciplinary right now. So there's a lot of efforts coming from computer science, engineering, and physics, where we're attacking this this field at different angles. And as Irfan also mentioned, there's a lot of effort on the, on the hardware side, especially because right now we're at the stage that we have multiple qubits in one chip. And so as we increase the number of the qubits, the number of the lines that goes into it that are needed for measurement and control is going to increase as well. And so how can we have a fridge that can accommodate all of those wires and still have like a low noise, and how can we reduce the noise? So these are just a few of the challenges that we're, we're working on.
Yeah, Aliyah, I think the, the key part is what does work mean? [Laughing] So quantum, in terms of working, we can do something, right? To use another fuzzy phrase, right? We can do something and with these computers and in particular, what we can do is demonstrations of functionality. So we can test, you know, is the idea of a certain quantum algorithm at the level of two bits, say, working. Or three bits or 10 bits, right. Is the, is the principle sound? Are there physics reasons why it should not work? So what's very exciting, is that what we've seen so far is that there are no indications that it's not working, right. And then the question is, well, what's the next step? When we are able to run these algorithms today, we run them in a fashion that they are not exceeding any classical computers in terms of capacity.
So you can't go to a sort of nascent quantum computer and say, well, I'm going to replace my Pentium with this, right, it's simply, it's not close at all in terms of computational power. There is of course, the beautiful demonstration by the Google group of this quantum advantage. However, these types of experiments are done in, in problems that are really cooked up, if you like, or generated to demonstrate this property, right. We have not demonstrated for algorithms that we are using today. Right, for computing, let's say the energy surface of molecules or batteries, or any physical problems that that are being carried out at the lab. Right. So that's sort of the next step. How does one go ahead and actually then configure one of these computers to solve a problem of scientific import or a scientific problem? The challenge is that we really need to have a radical change of architecture.
And what I mean by the following, what I mean by that is the following. If you take a noisy computer, as, as I was mentioning, that has a short lifetime, and you just keep adding bits, that doesn't mean the computer is getting more powerful, right. Because in fact, it takes time to actually go ahead and put all of these bits in a complex quantum state. So let's imagine you have a chain and you start entangling them on one side of the chain. Okay. By the time you get to the other side, you already lost the coherence in the beginning of the chain, because it was not, there was not enough coherence to make it live long enough, you know, for that process. So simply adding qubits, which are noisy, or which have low coherence, doesn't increase the capacity of your computer.
So although this is perhaps a debated point, I think a strong preponderance of folks believe that you really need to mitigate errors and correct errors, right, to go ahead and make a computer that's more powerful. So it's not simply about going ahead and making materials more perfect. I mean, of course we do want to do that. That makes error correction easier, the better your qubits are. But to go from this physical layer to use the quantum verbiage, so what we have now are physical qubits. So each piece of metal, each transistor like, or each qubit that we make is one quantum bit. When you go into error correction, you use many of these physical qubits to make one logical bit. So my real zero or one is now distributed over, let's say 10 bits, right. Or a hundred physical bits. And that's a much more robust way to encode this information. So the challenge in going to this logical qubit formulation, where error correction is being used is given the quality that we have, the lifetimes that we have, and the error correcting codes that we know, we're envisioning a hundred thousand to a million physical qubits for each logical bit. So that's a very far cry from the 50 qubits that we have today. So there has to be some radical progress in how we actually decode errors and correct for errors in an efficient way.
So the, so you're saying that there's a difference between what is made as like one physical bit versus what is like a functioning quantum bit?
Absolutely. Right. So for example, if you wanted to share a secret or store some information with your friends, if you only tell one friend that friend can forget, so it's better to tell three of them, right. Or five of them. So you're doing the same with quantum information. You're going to de-localize it in a way that in fact, it's stored over the collective state of many individual bits.
Okay. But when someone talks about how many bits their, their quantum computer that they're building or experimenting on has, are they generally referring to the physical number?
There are only physical qubits. Okay. So any numbers that you typically see, or if not exclusively see in the press or physical qubits.
And so you mentioned 50, is ,that's how many are on the current quantum computer that you are both working on? And can you tell me about that computer? Like how long have, has it been operational?
Sure. So we've been playing with quantum bits for a decade and a half now. So each one of these go into a dilution refrigerator that use a helium gas that's compressed into a liquid and then pumped on at low temperatures to go to 10 millikelvin. So each one of them are running, if you like, some flavor of quantum computer. Some with fewer bits, some with more bits, so on and so forth. But typically the largest one that we're running are running about eight qubits at one time, and they're running multiple chips of eight qubits. And we have designs that go beyond that, but it goes back into the quality of your bit, right? We increase the number of qubits as the quality goes up, right? So as my [test]beds are able to sustain a bigger and bigger superposition, I can think about a 16 qubit or a 32 Qubit chip.
So these dilution refrigerators are currently on campus, but we are also eagerly awaiting the, the completion of building 73. So this is a structure that LBNL has put in, an investment, and this is right next to the botanical garden, a nice wooden building, so it doesn't have ferrous metals to create all sorts of noise that would wreak havoc on your qubits. And when that's up and running in about a year and a half, then we hope to to house four dilution refrigerators there, and really have that as being the home for the test bed that we've put together here at LBNL, the advanced quantum test bed and other programs that are running across the enterprise.
So one thing I was really looking forward to talking to you both about is, I know a lot of people who are computer programmers, just being from the bay area. So I was thinking about this episode and I was thinking about how programming would change if quantum computers took over. Is it, is coding for quantum computer absolutely different than coding for a traditional computer, and will there have to be a huge shift?
Yeah, I think that, that's a great question. And part of it has to do with the fact that the stage of development is very nascent on quantum computers. So for classical machines, we are typically not thinking all the time about the specific instructions and the specific motions and voltages that go down to the transistors. We are programming at a much level, higher level of abstraction, whether it's the mathematics of the formula that you put in, or the syntax of a programming language or HTML, or VHDL where you have pictures, and you know, more graphical things that are there. So in the evolution of quantum technologies, we will also get there. But we're not there yet. Where we were about five years ago was completely bespoke. Everything had to be programmed by hand. Eventually we will get there, right where you won't know what's inside the, under the hood, it's quantum under the hood, you'll just see it moving faster. But at the moment, we're not there yet. But also to answer the question that you asked, will programmers have to do things very differently? I think we very much welcome, and it will be critical to have those programmers put their input in and learn all those inner parts in, in this intermediate era where it's quasi bespoke. Right? So in fact, it's going to be a co-design process.
So I think this is a very exciting time that we are, and right now, because we're at the stage that one does not have to have full understanding of what the gate is in order to operate and, and apply different algorithms and actually do computation. So what a gate is, is it's, it's the core of the quantum mechanics of the qubits. So you apply some type of operation on the state of your, of your qubit. So basically you're, you're telling your qubit, for example, go for, go from states... If you imagine the sphere again, you'll have, you're at state zero, you're telling it: go to state one, for example. And so you're, you're doing some type of operation to your, to your qubit. And so we're at the stage that those gates are semi-automated so we can use them. And now we need like higher levels, which are the algorithms, to run a code so that we can do things, optimize things, and we can do them faster and like a more efficient way to actually do these types of processes. But we're slowly getting to the point that we would want software engineers help in understanding on how to optimize it, because if we want to operate at the classical level of computers, then then we need to have that knowledge.
Yeah. So in fact, to follow on that, if you were to think about, let's say a classical process, that's just a comparitor. So something that compares, you know, what's in your right hand and what's your left hand. So that's a classical gate, right. And there are "and gates", do you want to see, are they the same; or, "or gates," or this or that; or "X or gates." So in a classical computer, you're not worried about the information running away on you. So when one is optimizing a compiler or a code in classical, a classical computing is to get it to run faster, right? Or easier with less resources, so on and so forth. So of course in quantum, you want to also do that, but it's complicated by the fact that you can't run them over a very long time, because as you asked earlier, it runs out of gas, right? The coherence goes away. [Laugh] So this compilation right now is very complicated in that not only we're trying to make the execution simple, but to actually make it run because you may not have the coherence that you need to actually run the gate.
It's interesting though, because I mean, classical computing is you know, very linear. It's like, if this, then that followed by this, execute that, but do quantum computers need to follow that kind of linear logic because they're at their core, not linear?
Yeah, that's very interesting.
So I guess like one, one way to think about it, the reason classical computers are, are linear is like the essence of your bit or transistor. It's either zero or one it's either on or off. But for quantum computers, as as we mentioned, it's, it's not, that's not the case. You have entanglement, you have superposition -- but what we mean by like like one gate before the other is because each one does something to our system. And when it does that, like, it, it changes like the, it changes your state. And so after that, you want to apply something else and sometimes it's more optimized to do one then do the other, because you save time. So that, because as, as Irfan mentioned before, superconducting qubits, coherence time is limited. We don't have an infinite amount of time. So we have like some limited time and we want to apply, we want to be able to do as much processes as we can on, on the qubit. And so it's important how to put one gate, then the second, then the third so that we can we can do all of them. Yeah.
And exactly as Zahra said, there, there are certain things in classical computing that follow a timeline, as you were asking, right? So you run box A, then the result of a feeds into B right. Then feeds into C. So that is also true, right? At some level in quantum things.
Irfan, I got a sense of what Zahra is working on these days, but what is your personal focus currently?
So, what I find tremendously interesting and exciting about this field is that there's really two sides. This is now a two-sided coin where you have this quantum information technologies on one side, you know, computers and sensors and all the wonderful things that we could be doing with modern equipment. But there's also really fundamentals of quantum mechanics on the other side, we really need to learn more about the theory to move the technology forward. And as the technology gets better, we learn more about the theory. So interesting boundaries that are coming up at the moment, you know, what really distinguishes a classical computation and a quantum one? Where quantum better, whereas classical better? And this is an interesting and tricky subject. So for example, there are some algorithms which are called hybrid algorithms, Aliyah. So in these hybrid algorithms, you only do the hardest part of the problem. Quantum mechanically, the rest of it, you use the classical computers. So on the surface, this seems really good.
So, no reason to make all of these other things, quantum. But it's not known if the computational power of such hybrid systems is better than a classical system. So I really would like to answer that question because it tells us do we have to go all out, all quantum all the way, or are there intermediate milestones where quantum can have advantage? So that's a technical problem. Related to that problem, the other side of that coin is how complicated a superposition can I actually build in a quantum system? We have never tested that question, right? We started with superpositions of one atom, two atoms, few atoms, tens and tens of atoms. Does quantum mechanics work for 1 million atoms in an ensemble? I don't know. Right. And the reason why I'm not a cynic, or I'm not a conspiracy theorist or anything like this, but I do want to point to the fact that we change theories of physics every a hundred years, or so, and quantum mechanics has been around for awhile. So theories of physics change when you have new tools, right? Quantum mechanics was invented when we had better spectroscopy other tools to study the atom. And we realized that our previous notion wasn't right. So now that we actually have better tools to build more complex quantum systems, is quantum mechanics in its present form the final theory? From a philosophical, cynical physicist point of view, I would say no, but I'm excited to figure out what the next step is.
And so by, by building these quantum computing systems, you, you kind of are going to get at that question inadvertently or, or is there, was that even just one main reason to build them?
It's definitely a main reason. It's, it's it's not a side product. I would say it is an equal partner.
So I have read, though, that there are some things that classical computers or current computers really can't do, or they, there just isn't enough processing power that could ever be built that could do, like modeling the states of molecules, modeling states of being in nature, because nature itself right, is, is quantum. So for things like that, those, those are systems that only a quantum computer could solve. Is that right?
That's absolutely right. So there are classes of problems that are classically intractable. And in fact then, then one looks at the overlap of problems that are classically hard, but quantum-ly doable.
So I think this is still an open research question, and there's a lot of research and effort putting into this, and one main avenue to showcase the power of quantum computer and is when we're dealing with quantum systems. So for example, in biology, sometimes it's very important to understand the energy levels of a molecule. So, and understanding these energy levels can be quite complicated because a molecule is a, is a combination of a lot of different atoms that are bonded differently to one another. And that creates a wave function that is kind of like a fingerprint of a molecule, but this wave function can also change. So for example, like if your molecule is slightly energized, or if it's slightly excited, it can have like a different wave function. And you might think like, who cares about like the quantum wave function of like some system, but it's actually very important. So understanding this wave function and understanding how we can isolate it and how we can, for example, go to the ground state, which is like the most stable state or the lowest state of your molecule, is important. And if, if we want to manipulate these molecules, we have to be able to understand them. These types of calculations can be quite hard for classical computers. And so that's basically where quantum computers can be quite powerful.
And so understanding, understanding these molecules and their energy levels is very important because if we want to manipulate them on, if we want to use them, which is something that we truly want for example, in a pharmaceutical solutions or drug delivery, if we want to use a complicated molecule to a deliver drug and cure cancer, we need to understand that the wave function of this molecule,
What are the problems that are being fed into these small qubit systems, experimental systems that you're working with now?
Right. So in terms of algorithms, if we were to break them down into two classes, we can think, talk about things that are pure, pure-bred quantum algorithms. So these do quantum things start to finish, and there are a few exemplars that are sort of well known in the literature, and they kind of fuel all the other algorithms that are there. So one of them is called Grover's search. So if you were going through a phone book, right, and I wanted to find the name Kovner in the phone book, I would have to sort of go page-by-page. And on average, I have to go through half of the pages in the phone book to find this name. So that's because you're doing a linear search, right? And the probability that you will find this name by chance: one half. If I actually make a quantum superposition of all the names in the phone book, what I can do is I can query all the names at once.
And in fact, I will get an answer which is square root of time faster. So I will reduce the number of queries by square root of n [number of names]. So that's a, that's a known speed up. So already very good, if your n is large, right? If you have a very large database or phone book or search that you need to do, you can do this. The other algorithm is Shor's algorithm, right? Where we go ahead and take a number and factor it into its prime factors. And this has an exponential speed up on the best-known classical algorithm. So notice the words I used, this is an exponential speed up on the best-known classical algorithm. There may be someone that knows a better classical algorithm. That will be a very rich person, no doubt once they find that. But that's another example of a, an algorithm.
These are pure quantum ones. They require lots of qubits. They require long lifetimes, right? So these are, if we can achieve these, then tremendous results will happen, right. Because we can prove that there is advantage there. However, it's not so easy to get to this error corrected stage [where these algorithms could run]. So then there are algorithms which are called hybrid algorithms, as I mentioned. And so one of the ideas there is the following. If you want to calculate, you gave the example of a molecule, right? And you want to figure out the energy of a molecule or a catalyst that's used in batteries or fertilizer production or, or pharmaceutical design. Well, as you put all these atoms together, it gets really complicated, right? We can write the equation in the quantum mechanics that we know, but our classical computer cannot solve it in any real time for [large] numbers of atoms.
So if we wanted to solve, let's say the the, for the full dynamics, all the full bells and whistles of an arrangement of atoms, maybe we can solve it for 50 particles or a hundred. That's not very many, right. A complicated molecule will have thousands if not more. And so the point is, what you do in that case is you take your quantum system as a hardware solver. It artificially kind of arranges these atoms and automatically calculates their energy. And all you need to do is just flip through all the different configurations. So I take my quantum computer [and set it to solve] configuration one, the hard part is calculating the energy. The quantum computer does that. It finds the energy, and then I flip it classically to the next orientation. It finds the energy, right? It's almost as if you're looking at a locker, right, that's on your door and you forgot the combination. You would keep changing the numbers one at a time, right. And pulling on it. So the quantum computer does the hard work, but you have to shuffle through all these possibilities classically.
And so that's an algorithm that people are running. And the question is, the resources that you need to do the classical shoveling, do they outweigh the benefits of having that quantum computation? This is not known.
Right, well, I mean the, the benefit, I mean, the benefit is you couldn't do it at all otherwise, right? So that's a good benefit!
You are an experimentalists at heart, right! [Laughing] Because that's the, that's the argument I make. [Laughing] I'm saying that even if it's not better, at least it's a solution, right? And I think that's part of what's driving the, the, the excitement at the moment -- can we actually scale in any reasonable fashion, right, these hybrid algorithms that we have now, and what would it require to do that scaling?
Are quantum computers improving conventional computers as well?
Yes, I definitely think so. Because in two ways. Number one, we learn more about what classical computers can actually do. Right? You learn about the boundary between quantum and classical, because for a long time in classical computing, we had a certain architecture or a certain methodology we've been following. And quantum is forcing us to think, well, is that really where the boundary is? Or have we simply not tried, right, with classical computers to do different things? The other idea is that quantum algorithms can also inspire different ways to look at classical algorithms, right? You may ask, well, what if I use the same workflow, but don't have the entanglement. Is there some advantage? There may be, in many cases. And the famous example of this that has come up recently is something that's called the Netflix problem. Do you have Netflix, Aliyah. Do you use this kind of streaming service?
Yes. Too much.
Yes, it tells you the the predictions, right, for what you should be watching based on what you have watched. Right. And the way that some of these matrices are calculated, it's called a preference matrix. So you start to tabulate a list of users and their habits right. Of what they actually watch. Right. And then there will be certain holes. For example, if you have watched three particular shows, the real question is will people that watch that sort of shows, are they likely to watch another show, right? And so in actually dealing with these matrices or this mathematics of this group of mathematics of large numbers, one tries to go ahead and solve them or come up with preferences based on that. And there was an algorithm that was proposed that was quantum mechanical by Prakash and Kerenidis some years ago that was supposedly going to actually solve this Netflix problem in a much more bonafide efficient way than a classical algorithm. And a famous story is of course, that a student of, one of our one of the colleagues, famous people in the person in the field, Scott Aaronson put a young Ewin Tang on this problem to prove that this quantum algorithm is actually the best in class. And what he did was find a better classical algorithm for doing this. And if I've gotten the story, the details of the story, right. That's how it proceeded in the press. So indeed quantum algorithms, the search for quantum algorithms may lead to better classical ones.
That's. Yeah. That's very cool. So one thing that I definitely wanted to ask you both is if you for you, what's fun about this field. I mean, if you're physicists at heart, how did you end up as experimental quantum computer builders, as opposed to a different field, or as opposed to something different entirely, how did you get to where you are?
Good question for Zahra.
[Laughing] So I think for the for me, it's like, I still consider myself as a physicist, but for me what's fun about it is the fact that we can play with something that is very in comprehensible, like something that quantum mechanics and actually do computations. And so I remember in my third year of undergrad, I took my first quantum mechanics class. Actually, I took it with Irfan, and the, the whole subject was very fascinating to me, like the particle wave duality or the Schrodinger's cat. And now we can actually use these phenomenas, we, we can control and manipulate and measure the entanglement of the quantum states and do computation. So for me, it's, it's amazing. And I think what attracted me to the field is the fact that there's so much yet to be learned and understood. The field is at like as, as we've discussed, it's at a very exciting stage. There's a lot of effort and, oh, we're at the stage that we don't know, we don't have the perfect answer yet. So we have some decent solutions that are functional, but they're not optimal yet. And so there is, so everything is up for investigation. Nothing is set on stone. So it, and I think that creates a perfect environment for research and innovation, and that's the reason I, I joined. Yeah.
Yeah. And I share that enthusiasm, right, with Zahra saying that there's so much that's not known. We're starting to actually just look at very, very basic ideas of how quantum systems operate. And the fact that we now have tools to actually settle some of these debates that were theory debates or philosophical debates, you know, decades ago is, is pretty cool. Right. You can actually answer a question that someone raised decades ago with some hard experiment, with respect to how I got into the field. I think it's completely circumstantial at best. [Laughing] I started off as an undergraduate and I had a tiny part of a research project at Harvard that looked at neutron trapping. So I was at atomic physicist, if you like, or playing part in an atomic physics group had nothing to do with quantum computing. And so that's very far away from superconducting circuits, but I grew up in New York City and there was, during the summer, an opportunity for an internship at a company in New York City called Hypres that worked on superconducting circuits. And it paid a few thousand dollars, which was, you know, for me a big number, you know ...
Yeah, as a college intern, that's big bucks! [Laughing].
[Laughing] Yeah, exactly! Absolutely. So I took that. I applied to that. I got that internship and I worked at Hypres right, looking at superconducting electronics, classical electronics that were there. And the the CEO of that company at that time was a student of, of who I also became a graduate student of Dan Prober at Yale university. So they introduced me to Dan and then I went to Yale and we went and then started working on detectors for astronomy. So again, nothing to do with quantum computing, because qubits were not invented yet for the superconducting domain. [Laughing] The first superconducting qubit is measured in Japan in 1999. Right. So I'm finishing up my Ph.D. right around that time, so this seems to be a very interesting topic. And then right around that time, other great person in my life would be, Dan Prober, of course, being the graduate advisor and a fantastic advisor, but my postdoc advisor and another great person, Michel Devoret, came on a sabbatical to to Yale, right. And he started working on qubits. And then the rest is history. I think after that.
That's a, that story is a nice way of reminding people that a lot of these nascent fields kind of start out by a couple people spreading ideas around and drawing in others with their own curiosity, and how that can be the foundation of these, these movements. So it's nice that your, your history was sort of through just meeting people and sharing interests with them.
Yeah. And very much so Aliyah, I resonate with that concept. And I tell students as well, when they're starting, it's good to plan out your career, of course, in terms of where you want to go and study you know, in science, which fields, but don't completely write out or underwrite out the joy of discovery science, of serendipity, or following where your heart goes, right. Because as long as you follow the scientific method, put in the hard work, it'll lead to something interesting no doubt.
Did you ever think the quantum mechanics you learned in class would actually be useful for something?
So it's very interesting because I think the essence, like the quantum computers is where we're really using the essence of quantum, um quantum mechanics, and we're actually seeing it in play. And it's, it's very powerful to actually see something like very abstract. Like you read about this, Schroginder's cat in a box that is both that dead and alive. And you don't know if it's dead or alive until you open it. And we're seeing the same concepts in, in quantum computing that you'll have a state and you don't know if it's zero or one and the moment you measure it, you collapse it. So we're playing with within the boundary. You don't want to collapse it. You want to be able to do different processes on it and and do some type of computation. So, yeah.
Yeah. And what Zahra is referring to here, it was something I call the first and second quantum revolution, right? So quantum mechanics is revolutionary in sort of the intellectual sense. So the first revolution was that the world is grainy. Everything can be grainy, right? It's quantized. The Greeks knew about the atom, right. You know, it was lost of course with Aristotle for sometime, but the Greeks knew about the atom. They knew, had thought about things that were grainy so on and so forth. But the second quantum revolution really says that you can put these grains in a superposition, and this is just absolutely bizarre, right? Because it contradicts, you know, physical intuition, physical reality that we've, we are used to in the daily world. The idea that things, these grains can exist in a superposition, this totally bizarre. Right. And it seems that all the experiments that we do say that's how the world is. We just haven't seen it that way.
Yeah. This is one of the things that I think about too much, but just what would happen if we could really zoom in on an atom then an electron, you know, what would we see about how these particles exist and you know, what they look like? What about after an electron? You know, what do its parts look like and how will we ever know what's there?
Better tools, haha.
Better science through better tools. That's the LBNL way.
So true. Okay. So one final question. Zahra: Was Irfan a good teacher?
Oh no, I should log off now. [Laughing]
Thanks for listening to A Day in the Half Life. Oh, and P.S. Zahra does think Irfan is a good teacher. This episode is dedicated to the memory of Glenn Roberts, Jr., a science writer who could make the most complex physics stories fun and easy to read.