SPEAKER_03 0:40

Hey Ciprian, how are you doing?

SPEAKER_01 0:42

Hey, Patrick. I'm doing great. Looking forward for another episode of Entangle Things.

SPEAKER_03 0:47

Well, actually, this is a gonna be a great one, I know for sure, because it's we're being joined by Google. I can't believe it's been five years and it's the first time we've had anybody from Google. Um, and also they're from Boston. So, Emma, could you mind introducing yourself to our audience?

SPEAKER_02 1:00

Yes, hello. I'm the Googler from Boston. Um, my name uh my name's Emma Rosenfeld. I did indeed grow up in the Boston area, uh, Medfield, just near Natick, so not too far from you in Quincy.

SPEAKER_03 1:12

Exactly.

SPEAKER_02 1:13

Um but yeah, I've been fascinated by physics and um building new technologies with physics for as long as I can remember. Um, I got interested at the very start in in high school when I kind of you know took my first classes and realized that you could make a puzzle that you solve with math to impact technologies and build new technologies. Um yeah, so I grew up in the Boston area, went to school there, um, did my PhD with Misha Lucan at Harvard, focusing on how can we actually make like sound waves and mechanical structures go quantum, colder and smaller, and try to explore what kind of new technologies we could build with that. Um so again, trying to push things to the very small or the very cold and see what comes out for new technologies, what kind of capabilities do they have? Um and you know, I'm more broadly interested in uh, you know, communication, sensing, and computation and the quantum advantage that comes from that. I spent a summer at Google X where I worked on trying to uh actually detect magnetic fields coming from the heart to diagnose new heart conditions. Started a project there at the intersection of medical technology and quantum mechanics, uh, and then went to quantum computing and industry um at Amazon for a couple of years, and now I'm at Google.

SPEAKER_03 2:25

And and you're doing research there, right? You're diving into scientists. So you haven't changed your mode of thinking for quite a long time, it sounds like.

SPEAKER_02 2:33

Well, I mean, I like to think that I've thought about different types of problems a bit over the years, but um yeah, I mean, I have just loved the scientific process, trying to find new questions and new directions to explore that will have some material impact on the world before, at least in my lifetime.

SPEAKER_03 2:52

Oh, I think you're at the right place at the right time.

SPEAKER_02 2:55

Yeah, thanks. Thank you. Yeah, and Google's been fantastic.

SPEAKER_03 2:58

Cool. So can you talk about what you're thinking about working on now or or what what do we want to talk about today?

SPEAKER_02 3:07

Well, yeah, uh, I mean, at Google, we have an effort in quantum computing. We're trying to build the world's first you know, fault-tolerant quantum computer and solve problems that normal computers classical able to get us to. Right. Um and our approach there is to use superconducting qubits. So we use metal printed on a semiconductor. And this kind of piggybacks on all the development that's happened in the uh digital electronics community, the development that's happening there in nanofabrication, for example. And the idea is to build something that is truly lithographically scalable so that we can harness these tools that have been developed to scale some quantum hardware.

SPEAKER_03 3:47

And we've got a lot of experience with some semiconductors now.

SPEAKER_02 3:50

And yeah, exactly. We have a lot of experience with semiconductors and the fabrication of uh chips, you know. And the idea is to try to build something using that technological development, and it has been for the past, I would say, um, decade or so. Um, but at the same time, you know, we're also at an inflection point in the field where we're really seeing that we are in this new regime where we're pretty confident that some hardware platform is probably going to work. We don't know quite which one yet, but we think we're gonna get there.

SPEAKER_03 4:22

But it might be multiples. Maybe we'll have different modalities.

SPEAKER_02 4:25

It could be different modalities. And it could be that depending on the application you want to run, you're better off using uh neutral atoms or trapped ions or superconducting qubits, and depending on um, you know, the trade-off of performance and speed that comes along with each of them. Um, so we we feel like you know, it's it's been fantastic to see this shift in the last two years where it's like we've gone from during during my PhD, it was like nobody knew about what quantum computing was, and it was kind of like, oh, this is this interesting kind of scientific question, and yeah, maybe we can get there, but there are all these open questions in the hardware, and now people are like, okay, let's prepare cryptography, for example.

SPEAKER_03 5:02

Yeah. Cyprian can relate because AI was not on the uh most people's bingo card five years ago, and now it's everywhere.

SPEAKER_02 5:11

Yeah, and that I would say that pivot comes from both hardware developments and also demonstrations of something called quantum error correction, which is where you know we're actually able to uh solve one of the most important problems in quantum computing, how fragile quantum computers are by using redundancy in hardware to store information in a more reliable way.

SPEAKER_03 5:32

Okay, so so that's the I mean, that's the story we've heard for a long time. And it seems like we're making bounds all the time on that, but it's it's very incremental. Is there a is there a holy grail that we're looking for that will unlock it all, or is it going to be incremental all the way down, you think?

SPEAKER_02 5:51

Well, I think that there are going to be challenges at the hardware level that are different depending on the hardware approach at different scales. So for us, for example, you know, we have all of these controls that are a chandelier in our fridge. We have a line per qubit coming down. And that's great for research purposes where you want to probe and you want to understand what every single qubit is doing. And uh, you know, it's great for scientists like me because I get to really run the experiment and get in there and see what's happening. Um, but it doesn't scale, of course. You need it now for prototyping and for understanding physics, but um, but it doesn't scale. And, you know, we have a variety of solutions that we've been thinking about and uh that we're pursuing. Um, but it it feels like we've we've just started to kind of scratch the surface on that from an engineering kind of starting to treat it from an engineering and scaling point of view. Um so I'm actually more optimistic than ever in terms of superconducting qubit hardware because I think that there are you know still still a lot of low-lying fruits to be had there on the on the control problem, which is probably the biggest um, you know, the biggest Achilles, Achilles heel of our hardware as it is today. And we we knew this would never, you know, the controls would never scale. So so we're working on on hardware solutions for that. Um and then maybe we can get to after we solved that problem, maybe we can get to like, I don't know, the kind of scale.

SPEAKER_03 7:20

I mean that that would definitely be useful.

SPEAKER_02 7:23

Yeah, yeah. But we need to solve that problem first.

SPEAKER_03 7:26

Okay.

SPEAKER_02 7:27

Um and then I mean, I'm not I'm not I I haven't worked directly in ions or or neutral atoms, but um my impression is that you know they can kind of scale to the 10,000 sort of scale with the with increasing laser power and field of view of the um objectives.

SPEAKER_03 7:46

And that's physical cubans, not logical.

SPEAKER_02 7:48

That's that's physical. Um, maybe 100,000. Um and there are some improvements to be made on speed, uh, but then they'll need to if to scale up further than that, there might be some kind of you know scientific development that's needed. So I would say to answer, yeah, in short, Patrick, um, you know, I think that there are still some holy grails to be found at different scales, but I think that there are going to be interesting developments, scientific results that have material value along the way.

SPEAKER_03 8:15

I hadn't thought about the fact that every modality has its own barriers, other than superconducting versus like Majorana Fermion, which is a completely different animal. Exactly. Every one of them has started from a different point with different engineering obstacles along the way. And actually, superconducting seems like it's it's taken its on first. It's hit its its limit.

SPEAKER_02 8:39

Yeah, yeah, yeah.

SPEAKER_03 8:39

And it's working through it.

SPEAKER_02 8:41

It exactly. We're we're working through it. And we uh, you know, this isn't it's it's not a surprise for us. We've been thinking about these things the whole time. Um, but it's important to be able to prototype, demonstrate key concepts like quantum error correction in a way where you have a test bed that you can really you know study exactly what's going on at every qubit at every point in time. Um another challenge is uh you know materials and materials quality. And um, you know, I think similarly there we have um we've known that we're going to need to uh you know treat this problem with a more dedicated engineering team. Um, but for now we we we're prototyping with kind of the thing that that we know works. Um and I see this as really a a reason for optimism because I I don't think we've hit up against any technical barrier along these key Achilles heels.

SPEAKER_03 9:42

We don't need cold fusion along the way.

SPEAKER_02 9:44

We don't need cold fusion along the way, exactly.

SPEAKER_01 9:46

It's 10 years away, by the way. So cold fusion?

SPEAKER_02 9:49

Oh, great, good to know.

SPEAKER_01 9:50

Always I thought that was just fusion. Things continue like that for the past 50. So um one, yeah, uh a lot of talk has been uh uh made in the past maybe 12 to uh 18 months about the advances in fault tolerance, right, and and quantum error correction. So I'd really love to get your point on uh A uh do you agree that we're moving forward at a uh serious pace there, or B, do you think it's just uh kind of siloed uh smaller breakthroughs? Are we at the point where quantum error correction works, or we're still haven't reached the point where it has like a net positive uh kind of impact on the systems?

SPEAKER_02 10:40

Well, it definitely has a net positive impact. We've measured a net positive impact across multiple hearts, you know, both in the superconducting qubit community and neutral atoms. So we are storing information in a way that that information stays at a higher quality than if you did it kind of the naive way of plugging information into your qubits. So just a brief recap of what quantum error correction is. It's this idea that you have each individual qubit in your quantum computer is fragile. We're trying to program energy into, we're trying to program information into energy scales that are very small. And so any noise from the environment, you know, a radio tower nearby, um, depending on your hardware modality, the vibration of an elevator on the ground floor, building you're in, whatever, can actually throw the information off. And the the solution, and this was known as a problem. This was why when quantum computing was first proposed as a field 30, 40 years ago, it was like, oh, okay, you know, we this would be an interesting, but it's still just an academic kind of pipe dream because uh the information is so fragile.

SPEAKER_03 11:41

It's like keeping a candle on in a hurricane.

SPEAKER_02 11:44

Exactly, exactly. And uh really quantum error correction was seen as the saving grace of the whole idea of building a quantum computer because the physical error rates, if you program information in a physical qubit and you try to do a logic operation on that physical qubit, it's gonna fail one time in a hundred or one time in a thousand, or best case one time in ten thousand. And you might say, oh, that's actually pretty low. That sounds quite good. But in reality, the applications that we want to do have uh circuit depths. We need to do many, many, many of these gates on the order of a million to like trillions of them. And so that means that you need these gates to perform exquisitely well, down at 10 to the minus six, 10 to the minus 12, 10 to the minus 15, even. And that's a yawning chasm between what we have and where we need to go. And so people looked at the field and they said, oh, okay, you know, it's it's it's never gonna work for the applications we want to run. But quantum error correction has this key aspect, which is that when you program information redundantly across multiple physical qubits and you store the information in this kind of distributed way over those physical qubits, then you're able to do local checks on patches of these physical qubits without affecting the underlying encoded information. And that allows you to detect if an error has happened and try to deduce where it's happened and fix it without breaking any of the entanglement and superposition that you've worked so carefully to engineer at the distributed or logical qubit level. And the more resource that you put into that, the larger the system size that you're distributing over, you get an exponential advantage. And it's really this exponential scaling that has made people think, okay, maybe quantum computing is possible. Because if you start with a patch of just, you know, uh three by three qubits or five by five, that doesn't get you to the performance you need. But with an exponential scaling, you you actually, you know, that's a very powerful uh distribution. That means that once you start adding resources, you get uh you get down to the error rates that you need fairly quickly. Um so we're at this point where we are starting to see in quantum error correction across multiple hardware platforms, we're starting to see that we can program information in this redundant way. And that information is a higher reliability than the encoded information in each individual physical qubit. And that's amazing. That's a very exciting time to be to be a scientist in this field because it's like, okay, this key idea that we needed for building a quantum computer is actually, you know, in principle, it works. And that's a breakthrough. Um, there have also been some breakthroughs on um, you know, being able to perform logic gates on that information, again, in multiple hardware platforms, which is a great cross-check of these ideas. Um, and then simultaneously, uh, there's been a lot of research into saying how can we make these codes more efficient? There's something called LTPC codes, low density parity check codes, which are actually used in 5G. And uh, you know, this is a direction that's really been um pioneered recently in the field, uh, and primarily uh, you know, um take an interest in the in the neutral atom community because they have arbitrary connectivity. You need better connectivity, more qubits talking to more qubits to get them to work. Um, but generally, this is this is a direction of interest for uh for the entire quantum computing community because it means that the resources needed, the reliability, um you can get the same reliability with fewer resources in a quantum error-corrected way with these with these new codes. So there's the research of the codes, there's the nascent experimental demonstrations, and yes, we're seeing this reliable improvement at small scales. And now the challenge is to see as we really scale up to tens of thousands of qubits, does this exponential scaling still hold? And if so, we should be able to get to noise rates fairly quickly, if it does hold, uh, to be able to actually have application-relevant performance.

SPEAKER_01 15:30

Yeah, I think that's very like uh exciting times because I distinctly remember when we were um starting the podcast, and even in the first couple of years, uh there was no definitive proof that, to your point, quantum error correction is gonna actually work, right? Um and switching gears just a little bit, one of the things that we were uh discussing a lot in the early days, right, on the superconducting uh kind of modality was the challenge of building uh two-qubit gates, right? Because of the topology and and and and things like that. Uh and it proved to be like generally in the industry, building a one-qubit gate, right, proves to be relatively easy, right? Uh when you have to build the the two-qubit gates, almost all modalities prove to have certain challenges. So, like generally speaking, do you think we've overcome those challenges? Or that is still something that that we haven't like properly solved uh uh currently.

SPEAKER_02 16:36

I I think we're pretty much good to go on the two-qubit gate front. Um, do you mean physical qubit or logical qubit or both?

SPEAKER_01 16:43

Uh mostly physical, right? Because with logical, that's a different story. Error correction comes into play in other things. Yeah.

SPEAKER_02 16:50

Yeah, no, we're we're uh we two-cubit gates have been performing well for um a few years now, you know, well below the percent level error. Um you're right that in the beginning, when people are spawning new hardware modalities, um, that is a definitely a linchpin to solve. Um you know, my own PhD was in trying to understand for NV centers how do we get two cubic gates between them in a way that is that's scalable. NV centers are defects in the solid state, the motivation being you have a solid state chip that works at room or elevated temperatures, and you can uh entangle distant color centers through various proposals in order to scale a solid state technology. But um that was a linchpin when you're really like at the beginning of trying to invent a new quantum technology and say, okay, you know, there's this degree of freedom that I have that I can control and I can read it out and I can initialize it into a known state. Um two QBKs can be a challenge there. But we have we are well beyond that in the uh superconducting qubit community. Um we are entangling um transmon qubits at Google, um, but there are other qubit modalities that you can that you can choose in terms of the the shape of the metal on on the semiconductor. Um and generally the process is to have you know the two loops of current essentially that are your qubits, um have them printed close together, and then have something that causes an electric field to be shared between those loops of current. And that means, yeah.

SPEAKER_03 18:20

Is that a T-gate?

SPEAKER_02 18:21

No, no, sorry. This I'm I'm talking about the interaction at the kind of bare bones level, the the building block of our quantum computer, the physical qubit level. We can talk about a T-gate in a second.

SPEAKER_03 18:32

Okay.

SPEAKER_02 18:32

Um But that's pretty much that's pretty much solved, and we reliably perform uh two qubit gates.

SPEAKER_03 18:40

Okay, so uh what's a T gate then? Let me ask that question.

SPEAKER_02 18:44

Yes, so uh a T gate is a particular type of logic gate that we'll need to perform at the logical qubit level. So this is a quantum error corrected operation.

SPEAKER_03 18:55

Which is kind of where we are now because people have been talking about physical for a long time. We're starting to enter the age where logical and it it really drives me nuts when people write an article and they talk about qubits and they don't differentiate because there's a big difference between there there is a big difference, yes.

SPEAKER_02 19:10

One of them will get us to where we need to go.

SPEAKER_03 19:12

And and T gates is on the good side of that ledger.

SPEAKER_02 19:15

And well, yeah, a T gate is is a general it's it's an operation, it's a logical operation you can apply to a qubit. So specifically, I don't know if you're familiar with the Hadamard gate that rotates from the Z-axis on the block sphere to the X-axis. So so you you have energy configurations that in your qubit that store information, and changing those energy configurations amounts to a logic gate. A T gate in particular rotates your qubit along the equator of the block sphere. So it actually changes the phase of the superposition. You have a combination of zero and one, and it changes what is that linear combination, that coefficient in that linear combination. And the reason why people care about it so much is because it's really hard to error correct. For most of the error correction techniques that we are pursuing, there's this gate that is non-clifford, a T gate. It's called a non-Clifford gate as opposed to a Clifford gate because it has certain mathematical properties that mean that it changes the states in a particular way. And um, if you're familiar with um, you know, the block sphere where you take states from one axis all the way zero to an equal superposition, zero plus one on the x-axis, um a non-Clifford gate will actually rotate you off the axes of a block sphere. But um, the reason why it's hard to error correct is because it's it's not native to most of the error correction techniques that we are using. So we're encoding information redundantly and then performing logic gates on this encoded information. And it's really hard to do a T gate with most of the encodings that we're pursuing, at least the encodings that are very efficient. And so this actually sets a bottleneck for running quantum algorithms because then you can say, well, what do you do? You need this non-Clifford gate. Without the non-Clifford gate, you only have a classical computer. All the other Clifford operations will only give you the power of a classical computer.

SPEAKER_03 21:06

You're losing a dimension without it.

SPEAKER_02 21:08

You're you exactly, exactly. And so um the approach in the community has been to say, well, let's create these kind of patches on the sides of physical qubits that together themselves make up logical qubits. And we're gonna use these as a resource to supply T states, which is a resource for running a T gate, to our bigger quantum computer. So it's this special use case that's applied on the side. And then the challenge becomes okay, how do you create that resource at very high fidelity? Because you can think about it as like a little bit of food that you're supplying to your quantum computer that gives it the power that it needs to run. But the performance of it, once you supply this food, it has to be very high fidelity. You can't error correct it once it's in there. And if if you went up to someone and you said, you know, how long is it going to take to um, you know, factor a large number or uh You know, run some scientific simulation on quantum hardware, they would say probably a long time because it takes so many resources to create these T states at high fidelity. This it takes a ton of qubits and a ton of time to create these T states at high fidelity. And the incumbent way that people were thinking to create this resource is something called magic state distillation, which takes a ton of qubits and a ton of time. And so people were actually designing quantum algorithms to reduce the T gate count because it's so expensive. So it was really that's how just to give you a picture, that's how much of a bottleneck it was. But then, you know, there were there's this idea that was developed over the course of a few years in the community and um kind of refined and um and finished by my colleague um Craig Gidney at Google, uh, called Magic State Cultivation, which is where um they were able to use some techniques from quantum error correction, um, in particular post-selecting on errors, just throwing out if you see flags in your in your circuit that you don't like, throwing out the shot and restarting it, that means that you do you could create a T-state with much fewer resources than Magic State distillation. And this brought the bottleneck down for running quantum algorithms.

SPEAKER_03 23:13

Very cool. So that seems to be a very emerging uh place to reinforce and to get the success that we need if you're solving this big problem.

SPEAKER_02 23:25

That's that's right. It it was with magic state distillation, the incumbent technique, that's very costly. You want to really, you know, make sure the T count is is as low as possible that you need for running a quantum algorithm. Uh, but with Magic State cultivation, this actually brings the resource cost down quite a bit. And we have some work um that recently came out that actually shows on hardware that Magic State cultivation works.

SPEAKER_03 23:47

Excellent.

SPEAKER_02 23:47

So we tested these techniques on our Willow processor and um we demonstrated that these ideas that were theorized in late 2024 of bringing the record resource cost down for T-States actually works in practice.

SPEAKER_03 24:01

Very cool.

SPEAKER_01 24:02

You you mentioned that once you create them, you can't error correct them. It would be really interesting to kind of get a little bit of more detail in that. Like what are the things that are limiting that?

SPEAKER_02 24:16

Um, it has to do with the way that you are encoding the logical information in your physical qubits. Um, you are using particular checks when you do so, local checks of the information that have a certain um symmetry and uh certain axis. And if you try to do a T-gate in most of the ways that we are thinking about quantum error correcting, and you did your checks, you would destroy the logical qubit information.

SPEAKER_03 24:44

There's no way to do a T-gate and have these checks at the same time that provide the robustness to I mean the checks themselves seem like a cheat because you can't read a quantum state without destroying it.

SPEAKER_02 24:56

Exactly. And that's one of the key innovations of quantum error correction. We were able to develop a way, the community was able to develop a way where um you have these local checks, and the way that they work is they kind of check a dimension of your state space that you don't really care about. You care about it in the sense that it's providing you checks, but you're not actually storing your information there. And then at the same time, you encode your information in a different regime of your state space that would be destroyed if any of the checks fail, but isn't measured by the checks.

SPEAKER_03 25:25

So this is like if I had a bread I'm cooking, a bad analogy, but uh if I want to know how big the bread is, I but if I measure the bread size, I destroy it. So instead I measure the temperature.

SPEAKER_02 25:38

Yeah, it's it's it's kind of indirect in that way. And that's part of the challenge as well of quantum American correction because it's it is indirect, and because of that, we have to make sense of our checks as best as we can, and that's an entire area of research. It's called decoding, where you get all these flags back and it's telling you locally what might have happened, and you have to figure out okay, globally, what do we think happened?

SPEAKER_03 25:59

This whole space is just puzzles. It's just it's just puzzles.

SPEAKER_02 26:03

That's why it's so much fun. Who wouldn't want to work in this space?

SPEAKER_03 26:08

Very cool.

SPEAKER_02 26:09

Yeah. But those puzzles, but this encoding, again, it doesn't work for every single gate that we need. And this is something I think you know, we should be talking more about when we share, you know, logic results in quantum error correction because it's it's kind of like a little secret of okay, we can do you know all of these gates, but there are still these non-clifford operations that need to be treated very carefully and do set a bottleneck. But I want to be clear that with with magic state cultivation, you know, there's a path forward towards actually significantly reducing this bottleneck first from a theory inception to now it's experimental uh validation. And again, I think this is an area where we're just scratching the surface, and there are a ton more experiments and generations of this project to come.

SPEAKER_03 26:49

Very cool.

SPEAKER_01 26:50

You mentioned that that folks were uh literally trying to minimize the number of T gates in in algorithms. So I would like to switch gears a little bit at the higher level. Where do you think we are in terms of developing algorithms? Because we've heard it like, oh, uh we still only have like a very limited number of algorithms, and then there are other people saying, look, until we have working fault-tolerant quantum computers, don't expect kind of a significant increase in the number of algorithms. What's your view on the state? Where are we? Are we making progress? Are we waiting for the actual fault-tolerant machines?

SPEAKER_02 27:30

Well, first of all, I think it's a bit of an under-resource direction um broadly in the field. There's a lot of focus on creating the hardware uh in this way that is sort of protected in your own approach and um and has its own challenges and opportunities that we're kind of learning across each other in the community. Um, but algorithms actually, if if you look at like the number of people exploring quantum algorithms and really doing rigorous analyses of quantum advantage and verifiability, it's it's fairly small. So again, that's a reason for optimism because we need more mathematicians?

SPEAKER_03 28:03

We need more competitions.

SPEAKER_02 28:04

Well, we need more everyone in this field. We need we we need all kinds of skill sets, and uh, mathematicians is one of them. Um but okay, I think that's one thing. Um, the other thing is that you know, we we know that there are a lot of scientific applications of quantum computing. It's it's a it's a scientific tool that will change the approach for a lot of our molecular and material technologies. So instead of having a top-down kind of approach where you need to guess and check in the development of uh pharmaceuticals or fertilizers, and and you have all of these different combinations and you're guessing and checking and you're studying them, um, you can have this sort of bottom-up um posture where you're able to simulate, okay, I have this particular molecule of interest, and there's this area of the ligand that you know is gonna be really important to bind to this particular aspect of the cell in this environment. And you have an idea of what the structure of the problem looks like, but you don't have a computer uh powerful enough to simulate. So it's gonna be these very like surgical applications that are in basic science. And I mean, I'm a scientist, but I don't know, I think it's pretty clear that if we have a different tool to pursue questions fundamentally in basic science, that that'll speed up our development across multiple sectors.

SPEAKER_03 29:23

And this is what Feynman was talking about when he suggested that we needed a quantum computer is the physics. Cyprian and I, full disclosure, Cyprin and I are both programmers, so we're looking for the algorithms. Okay, sure. He he's like Grover's, I like Shore's. It's not lost on me that that Shore is a mathematician, he's not a physicist.

SPEAKER_02 29:40

Right.

SPEAKER_03 29:41

So it's just that different side of the uh the thing. But you're a physicist, so that you see the physics and the chemistry, and we definitely see that that's where a lot of the benefits will will accrue to. For example, there's no algorithm that I know of that affects the sensing world, and that's a huge part of what we're gonna get out of this.

SPEAKER_02 29:58

Well, yeah, I mean, and quantum sensing, quantum sensing is its own technology, you know, outside of quantum computing, pretty much. I mean, you could imagine transducing sensing signals into a quantum computer, but that's extremely challenging. We don't know how to do that, and and that's you know, that's that's down the road. So that's that's kind of a different area. But yeah, broadly, I would lump, as you said, Patrick, I would lump these applications of quantum computers into two buckets. One is the scientific and the kind of simulation and probing of basic science questions, and the other one is more computational things like optimization, uh factoring and such. Search. Um, but but yeah, I think, I mean, I think to answer Cyprian's question, um I think it's kind of both things that you said, that we are um still, you know, waiting for quantum hardware to be able to figure out all of the applications. And yet we know that there's this different approach and uh, you know, broad class of problems waiting for us once we do. But it's hard to answer what exactly the value is out of any of them because, for example, like for example, lithium-ion cathode, I'm sorry, lithium-ion batteries, figuring out the cathode is a hard problem. And you can use a quantum computer in order to try to get an answer to maybe what the material structure should look like. Um, but that doesn't answer yet, well, how do you synthesize the material at scale in an economically feasible way? Um, you know, what is the environmental cost of creating that material? Maybe if you can use your quantum computer to try to create a material that is lower environmental cost to study, but it's really like an answer to basic science questions that doesn't answer all of the development technical questions down the road. And I think that's where people get confused on what is the value of a quantum computer and where are the applications really, because it's it really depends on you know how much you you value answering these base these questions in basic science, and there's still a lot of unknowns, even if you do.

SPEAKER_01 31:51

It sounds like I think that's that's that's great that you mentioned that because we've seen a lot of that confusion. Uh, people tend to think that quantum computers are gonna be this magic thing that's gonna invent overnight for us some amazing materials that would uh solve all the problems of the world, right? Which is a little bit, I think, uh uh kind of parallel to the truth, because as you mentioned, right? Well, in 50 years, maybe quantum computing is gonna answer some of the very hard questions exactly not gonna be able to kind of do the whole thing for us.

SPEAKER_02 32:29

Exactly. It should be a surgical tool that is an incisive piece of uh an important piece, a limiting piece of a bigger puzzle. And um I'm really excited about that because it kind of means that there's there's a lot of there's a lot of reason to to think that there could be a massive impact. There's just a lot of uncertainty around it. Yeah.

SPEAKER_03 32:50

So we're we're past 30 minutes and it's gone by instantly almost. Um I and we don't have to end right now, but we have to start wrapping up. Is there anything else that you wanted to talk about or bring up or that we should inform the audience of?

SPEAKER_02 33:04

Well, I think you know, we just talked about a lot of uh potential applications of quantum computing, and I just want to reiterate that um hype can be very damaging to the field if you if you overpromise, like like Cyprian was saying, that uh quantum is gonna solve all of our problems or it's gonna you know revolutionize the world in solving a particular problem end-to-end without a broader context for um the other tools that are needed, then um, this can degrade public trust. And at Google, we try to fight against this by doing you know peer-reviewed publications and sharing our results and um all of our you know, our data when we publish is just uh publicly available so anyone can download it and play with it, raw data. Very cool. Um and it's it's a it's a tight balance to walk because you want to share all of these you know reasons for optimism in a different scientific approach to problems. Um, but at the same time, you can't um necessarily promise a particular number on a lot of these, a lot of these issues that you're solving.

SPEAKER_03 34:06

And there's a lot of room for misunderstanding. If you watch if you watch social media, there's lots of people saying, oh my God, they just did this and it's something they did seven years ago.

SPEAKER_02 34:15

That's that's right. That's right.

SPEAKER_03 34:16

So so there's misunderstanding out there.

SPEAKER_02 34:18

That's that's right, exactly. And um, yeah, and I think as as research continues in quantum error correction and as we s we start to scale hardware uh more, there's still going to be, you know, various engineering challenges along the way. But um because we've shown that the basic concepts work and we have an idea of how to get to the next scale or two, I think that there's that's why you're see seeing all of this um, you know, the posture shift in the community of saying, okay, this is this is something that isn't gonna work for the next 20 years to like, okay, we we need to we need to prepare for this.

SPEAKER_03 34:53

Yeah. Yeah. It's been a great conversation. I hope you uh you had as good a time as we did talking, and uh we hope you join us again.

SPEAKER_02 35:01

Okay, yeah, thank you. Thank you so much for having us. And and thank you for the work you're doing here. Um, you know, in our community, we I think we don't do quite enough of sharing what we're doing with people who haven't spent a decade banging their head against the wall on these problems. So um I really appreciate the discussion and the questions that you're asking, what you're doing to share uh to share our work with the the broader public.

SPEAKER_03 35:20

Thank you. Well, let's synthesize from one more thing, Patrick.

SPEAKER_01 35:23

I I I would like to mention here because I believe it is very important, right? Uh especially as Emma was talking about the hype and the danger of hype, right? I I would like to say that the the fundamental work that these very, very large companies like Google are doing in this space, I think it's paramount for for driving not only the scientific advancement, but also the trust. I think, as you mentioned, you do have a massive responsibility in keeping the trust. And uh I I think uh uh in in your case, you are just doing it like textbook and perfect way, right? It's it's very important. It's very important like for us, for the community, right, to um not be lured by some results that six months from now prove to be, I don't know, vaporware or or or whatever, right? The reader is is extremely important. So uh I speak for Patrick as well. We also appreciate what you folks are doing because you are one of the one of the driving forces in this world.

SPEAKER_03 36:29

And far better than letting just governments be the ones because then the stuff would be deep, dark, deep, dark secrets. And and so the sharing is is is wonderful. So thank you for that.

SPEAKER_02 36:38

Oh, yeah, of course. Yeah, it it takes a global village, and um, we hope that the broader community, you know, also keeps maintaining the transparency and trust that that they've shown us across different hardware modalities and institutions.

unknown 36:52

Cool.

SPEAKER_03 36:53

Well, I think that's it for now. Thanks again, and we'll see you every next time.

SPEAKER_02 36:57

Okay, yeah, thanks so much for having us.

SPEAKER_00 36:59

Bye.

SPEAKER_01 36:59

Bye, everyone.

SPEAKER_00 37:02

Cybercrime is one of the biggest threats to businesses of all sizes and industries. With almost half a million open cyber positions, the problem is compounded by the lack of available talent in the marketplace. At Pulsar Security, our elite team of highly credentialed experts collaborate with you to assess your current defenses and develop solutions tailored to your specific needs. With services ranging from cybersecurity education to advanced penetration testing and red teaming, you can start reducing your risks today. Visit pulsarsecurity.com and let's secure your digital future together.