How Do Spin Qubits Work?
Brandon Severin says that 20 years ago people did not believe you could rotate the spin of an individual electron. During my PhD, I saw teenagers do it in their lunch break as part of their labs.
In this episode · 24 chapters
- The Google announcement, and why it matters
- Reading a molecule with NMR spectroscopy
- The bearish case: not enough algorithms
- More qubits, or better qubits
- Trapped ions, and Newton's cradle
- Fidelity, unpacked
- How many spares does one good qubit need?
- What a spin qubit actually is
- The transistor already in your phone
- Turning an existing chip quantum
- Does nature already hold the algorithm?
- The quantum circus
- Why spin qubits could scale
- Who observes the observer?
- Copenhagen, or many worlds
- Using AI to wrangle the cats
- Who the software is for
- Empire of AI, and the GPU comparison
- The two camps in quantum
- The vibe in quantum, and the vibe in AI
- Kevin Kelly's question
- Why is it taking so long?
- What one algorithm could change
- Why quantum feels different
And the beautiful thing is with a quantum computer, we’ll actually be able to understand what’s going on at that fundamental level for the first time ever.
Brandon Severin founded Conductor Quantum, which builds on the transistors the chip industry already prints by the trillion every day. “The big thing for spin qubits is that the manufacturing chains are already in place.” He explains how a spin qubit isolates one electron.
Severin breaks down Google's new algorithm for reading molecular structure, and how AI now does the voltage tuning that took him weeks by hand during his PhD. He also explains why he prefers many worlds to Copenhagen, and why progress is slow.
Brandon Severin, founder of Conductor Quantum, argues on the Thinking On Paper podcast that spin qubits built from ordinary silicon transistors offer a route to the millions of qubits quantum computing needs. He calls calibration “a huge bottleneck to scaling”, which Conductor Quantum tackles with AI, and he rates Google's latest quantum result as a real advance.
How does a spin qubit store information?
A spin qubit stores information in whether a single electron's spin points up or down, and Conductor Quantum's approach isolates those electrons by adding extra gates to a standard transistor. Severin says the whole process runs on ordinary voltages, with “no lasers, no vacuum”. EeroQ is also building towards electron spin qubits, but with electrons floating on superfluid helium.
How does Brandon Severin rate Google's latest quantum result?
Severin rates Google's latest quantum result as a real advance: an algorithm that extends NMR (nuclear magnetic resonance, the standard way to work out a molecule's structure) to show how atoms interact over longer distances. He calls Google's claim that no classical computer can match its speed reasonable for now, though he expects someone may eventually find a classical algorithm that can.
What is the bearish case for quantum computing, according to Severin?
Severin names the shortage of algorithms as the bearish case for quantum computing. Hand someone a million qubits today, he says, and “no one really knows” how they would use them.
How does Conductor Quantum use AI to calibrate qubits?
Conductor Quantum uses AI to set the voltages that form each qubit and to recalibrate it when it drifts. Severin says he spent weeks trying to do this by hand at the start of his PhD at Oxford without getting close, and he describes the software as the precursor of an operating system for quantum computers.
Why does Severin think quantum progress is taking so long?
Severin splits the quantum industry into those building one qubit at a time, often with a university spin-out mindset, and those designing for a million qubits from the start. Asked why progress is taking so long, he says there are not enough people “building rocket ships”.
Where Jeremy and Severin disagree
Jeremy asks whether someone could make a single qubit so good that it debunks the case for scale. Severin answers with a DARPA study which, he says, shows you would still need a million qubits to run RSA even at very high fidelity, because quantum computers can only tolerate around one mistake in a thousand.
Links from this episode
Questions
How do spin qubits work?
A spin qubit stores information in whether a single electron's spin points up or down, and Conductor Quantum's approach isolates those electrons by adding extra gates to a standard transistor. Severin says the whole process runs on ordinary voltages, with “no lasers, no vacuum”.
What is Conductor Quantum?
Brandon Severin founded Conductor Quantum, which builds on the transistors the chip industry already prints by the trillion every day.
Why does Brandon Severin back spin qubits for scale?
Brandon Severin, founder of Conductor Quantum, argues on the Thinking On Paper podcast that spin qubits built from ordinary silicon transistors offer a route to the millions of qubits quantum computing needs. “The big thing for spin qubits is that the manufacturing chains are already in place.”
Which interpretation of quantum mechanics does Severin prefer?
He also explains why he prefers many worlds to Copenhagen, and why progress is slow.
Transcript
Brandon Severin the world as we know it is built on quantum mechanics, quantum interactions, the natural world, the table, the microphone, we have this podcast, the silicon chips that power the computers, molecules, drugs. At the fundamental level, all of that is all quantum.
Brandon Severin And the beautiful thing is with a quantum computer, we'll actually be able to understand what's going on at that fundamental level for the first time ever. At the moment, we're just kind of making approximations and being clever, but it's, know, okay, how can we know if this drug will work in clinical trials before it gets there?
Brandon Severin How can we make better fertilizers so we can feed the world that's just all down to understanding nature at the end of the day.
Jeremy Gilbertson Disruptors and curious minds. Welcome to another episode of thinking on paper. My name is Jeremy. This is Mark. We unpack the future with the people building it today.
Mark Fielding Our guest is Brandon Severin. He's the founder of
Mark Fielding Conductor Quantum, Welcome to the show, Brandon. Thank you for thinking on paper with us.
Brandon Severin Amazing. Thanks for having me on guys. Pleasure to be on here.
The Google announcement, and why it matters
Jeremy Gilbertson Tell us about this Google announcement that just came out, why it's important and why it has this interesting tie to the nature piece that we're talking about.
Brandon Severin Yeah, the Google announcement is very timely. It's important because one, look, the Google quantum team is great and they basically be able to show that they have a chip with over a hundred quantum bits, qubits, which
Brandon Severin They're able to control them, operate them, but most importantly, they've developed a new algorithm that allows them to start looking at how atomic nuclei interact with each other. how we do this normally is using NMR, so nuclear magnetic resonance.
Brandon Severin and it's how you figure out the structure of molecules. We've been doing this for decades now and Google's basically figured out a way to enhance NMR analysis on a quantum computer. So it's that direct correlation of, I have this atom here and it's, let's say, an oxygen and it's bonded to other hydrogens and next to it I have a carbon. What's like...
Brandon Severin what is the structure of that interaction? What does that shape of that molecule look like? And we've basically been doing that classically using NMR, but with now this latest algorithm from Google, we can get more of an idea of that atomic structure at longer ranges. And the way they've done it is actually quite clever, to be honest with you.
Reading a molecule with NMR spectroscopy
Mark Fielding How have they done it?
Brandon Severin Yeah, so it's this... The problem with NMR spectroscopy is that it's very short-ranged. I.e. if I have, let's say, an atom here and an atom here, you know, I can sort of figure out their interactions. But if I go further along the molecule and I have an atom over here, I don't really know, like, how this atom all the way over here interacts with this first one. ⁓
Brandon Severin And what Google have done is basically put together an algorithm that can actually figure out those longer range correlations to this nearby atom. And that allows you to basically have more of an idea of what the structure is. You don't need to kind of second guess what's going on. And they've done it on relatively small molecules, know, like 18 atoms, 28 atoms or so. ⁓ But it's like very promising. And the algorithm itself, like how have they done it is basically simulating the nucleus of this atom as an individual qubit on their chip. ⁓ And they basically run a plethora of operations to get the chip into a certain state. Then they perturb, basically give one of those qubits a little kick and then run that first set of operations in reverse across the rest of the chip and see how, what they get back.
Brandon Severin you know, is how it's different to what they sent, basically. And that gives you an idea of how that first qubit is coupled to all the other qubits. And that you can sort of think about that then as the nuclei. So the nucleus, how that nucleus of your physical atom is coupled to all the rest of the atoms in your molecule. That's effectively what they've done. And look, I mean, it's a first. They argue that no classical computer can do what they've done as fast as they've done. And I think, you know, first glance at the paper and what I can tell so far, I think it's actually a reasonable claim. Of course, I think someone will probably be able to like work really hard and come up with an algorithm that can run on a classical computer. But for now, mean, Google has a great result, right? It works.
Mark Fielding Can you give us like a real life example of what the algorithm or what future iterations of the algorithm could be used for in a useful way?
Brandon Severin any molecule where you want to know, you you synthesize the molecule in a lab, having an idea of what that molecule looks like, its structure is very important as to how effective that protein, that drug will be. at the moment, this was kind of the big idea with Alpha Fold and kind of predicting protein structures from Google DeepMind, but bear in mind that's predicting protein structures.
Brandon Severin not actually calculating them physically with hard evidence. And that's what this NMR spectroscopy brings insight to is, okay, what is the shape of this organic molecule? And therefore, what's the structure? How well will it bind to, you know, a harmful disease, ⁓ your cells, a ⁓ viral cell, for example. That's where this goes further down the road.
Brandon Severin And of course it's small molecules now, but it's the fact that they've been able to do this is, I think, really, really impressive.
The bearish case: not enough algorithms
Brandon Severin the bearish case for quantum, to be honest, is actually the lack of algorithms. You know, it's, okay, I give you a million qubits today, how will you use them? You know, like no one really knows. But we're already seeing that. Okay, as we get more qubits, we can develop better algorithms and test them and see how they work.
Jeremy Gilbertson What's the bottleneck in creating algorithms, quantum algorithms?
Mark Fielding Nature, Maybe you don't create an algorithm. You simply find, you discover where it exists in nature. ⁓ Is that a stupid comment?
Jeremy Gilbertson Ooh, fair enough.
Brandon Severin No, not really, because you're effectively trying. It's this kind of going back to this, the intro to the podcast, which is like you have this weird limitation of you want to simulate nature, but you also only have so many qubits. So then you kind of want to develop your algorithm for the computer you have access to, because then you can also verify that it works. The Google team put out this whole thing about like, this is the first quantum verifiable algorithm, it's kind of true. They just get a number back, which means that if you had a quantum computer, you could run their algorithm, check the Google result. actually, nature itself could run the Google quantum algorithm and check the result. That's the beautiful thing about it. When you think about it, that's the level we're playing at with quantum computing.
More qubits, or better qubits
Jeremy Gilbertson first thing we talked about, more qubits, or in the intro I talked about the more qubit camp versus the better qubit camp. And then there's also the next piece of the puzzle is, which type of quantum, which type of qubit is the best and which one are you using? So let's pick a couple. One, we're gonna stay with spin because that's what you're involved with. Let's talk about spin and let's say, let's pick like trapped ion So let's look at spin and trapped ion Can you break down?
Jeremy Gilbertson both of those two and tell me about the difference between one and the other.
Brandon Severin everybody will say their qubit is better than the other person's qubit. Exactly. Literally, literally. Yeah, literally. So let's start with trapped ions. they're probably one of the few qubits that actually just in when you hear them, they describe what they actually are where you have
Jeremy Gilbertson Absolutely, but Ford's better than Chevy, Toyota's better than, yeah, right.
Mark Fielding all been on the podcast and they've all said that their Qubit is the best. We've had IonQ on, we've had IBM, they've all said the same.
Trapped ions, and Newton's cradle
Brandon Severin Literal ions, so calcium ions, cesium ions, whatever you want. Well, actually that's more of the neutral ion, but calcium ions. ⁓ And you trap them in space. You have a column of ions trapped. How do you do that? Physical lasers, you literally, you know, use this laser cooling effect where you have ions floating and you shine a very well, you know, focused laser on them and you trap them within that.
Brandon Severin or beam, ⁓ kind of like something out of Star Wars, like you the Force. And you effectively start to build, know, I kind of describe it as like a Newton's cradle, but with physical ions. And the beautiful thing about ions is they are ions, you know, once you've trapped them they have their kind of this well that they sit in and you can operate on them.
Brandon Severin The beautiful thing there is well controlled, great environment, what's known as ⁓ the key metric in terms of better gives us very high fidelities, meaning that.
Brandon Severin when you tell your qubit or your ion to be in state one, it is in state one, you 99,000 times out of a hundred thousand. ⁓ Whereas other qubits don't necessarily have such high fidelities, but they'll have maybe other benefits like long coherence times, meaning your qubit stays in this quantum state for a very long time.
Fidelity, unpacked
Jeremy Gilbertson if I could just unpack fidelity a little bit, is that kind of a ratio or the combination of gate speed and coherence time together or no? Okay.
Brandon Severin It's almost, you're literally measuring the reliability of your qubit.
Jeremy Gilbertson of it being in this state or this state. Got it.
Brandon Severin do an op, and then there's almost like more layers of fidelity. So I perform an operation on my qubit and I expect this answer.
Brandon Severin is it correct? How many times out of 100 is it correct? Basically, that could be, let's say, your single qubit gate fidelity, and then you pair two qubits and you do a two qubit gate and you get a two qubit gate fidelity. Getting down the rabbit hole here, but this...
Jeremy Gilbertson Got it, got it.
Brandon Severin Good question. From what we're seeing.
Brandon Severin So single qubit gate fidelities are typically always higher than two qubit gate fidelities.
Brandon Severin I think it's just by nature of the system, right? Probabilistically, it's harder to do a two-qubit gate more reliably than a single-qubit gate. And by that, we mean like you're pulsing on one qubit, you know, you're, let's say, rotating it from zero to one, and then you're doing the same thing on the other qubit, and you're measuring out the combined state of those two. And there's just more things that can go wrong, right, as you add on that extra qubit.
Brandon Severin But you know, good quality single and two qubit gates, you can build a quantum computer. So once you've met a good threshold, then you can argue, with X number of qubits and this level of fidelity, I'm good to go. You know, we've ticked the boxes. ⁓ But the more qubits you have, it gets difficult to manage. So then you need extra qubits to correct for these errors, basically, if you're...
Brandon Severin that you want to use for logical operations.
How many spares does one good qubit need?
Jeremy Gilbertson the more qubit versus better qubit. Having one to back up the one that doesn't behave properly, right?
Brandon Severin Exactly. And it's not just one you need to back up, it's like hundreds if not thousands of qubits, you know, that you need for one logical qubit. You need like a thousand error correcting qubits.
Jeremy Gilbertson Thanks for diving into that stuff with us. Back to the trapped ions. You've got this thing suspended in a, I guess a vacuum is kind of, yeah, suspended in this little vacuum trap. It's kind of like Ghostbusters when they suck in the ghosts. It's an ion trap. It's probably similar to that tech, right? Ghostbusters were quantum.
Brandon Severin Look, if that works 100 % and you have it in this trap like a Newton's cradle, then you fire different electromagnetic waves at it. So whether it's laser light or whatever your ion of your choosing depends on basically the frequency or wavelength of light you shine on. So red, blue, green, whatever. And that excites your ion, that excites basically the electrons in that.
Brandon Severin different levels. Einstein, going, you know, back to the OG of quantum mechanics again, when that ion loses its energy, it emits light, and then you can measure that light. The hard part is measuring the light, because it's that light's going to come out as an individual frequency, it could get absorbed by anything in your huge vacuum, you need to collect it, you have all these different like efficiencies of detection. need to collect the light, need to transmit down a fiber and then you need to be able to actually detect it with a photodetector. That being said, the ion trap teams that are working on this, Oxford Ionics, now acquired by IonQ Continuum, they're great at doing this stuff. You know, like really, really good. They have some of the highest qubit fidelities in the world.
What a spin qubit actually is
Jeremy Gilbertson let's start talking about spin. there's like a silicon layer and we're literally poking holes in the silicon layer and trying to convince electrons to jump in those holes and stay there. Like, let's talk me through that, man.
Mark Fielding is a spin qubit? think is the question.
Brandon Severin two ways to approach this. There's the very bottom, bottom up way, where you start with the physical spin qubit, or there's a top down way, which is you start with a silicon chip. let's start with the bottom up way, because that's the fundamental question. What is a spin qubit? So a spin qubit is...
Brandon Severin orientated in a magnetic field. And you can view that electron similar to this NMR nuclear magnetic resonance. That electron behaves as like a tiny bar magnet similar to the magnets or a compass needle, you know, on planet earth, it wants to align with the magnetic field, right? Your compass needle wants to point north. It's kind of similar feel with your electron. It either wants to be spin pointing up or spin pointing down and up, let's say aligned with the magnetic field or down against it. And those are your two quantum states. Those are your two bits. Spin up, spin down. Zero, one.
Brandon Severin Now the hard part is, well, okay, great. I think like, arguably you could say this is probably like one of nature's most fundamental qubit. You can't get smaller than an electron spin qubit. Then the hard part is, okay, how do I isolate individual electron? Good luck. And that's where the silicon comes in. That's where we come in as a company, which is...
The transistor already in your phone
Brandon Severin You look at the transistors on your desk, the transistors in your phone. Those are basically tiny, tiny switches. You've got like 50 billion transistors in your pocket right now that allow you to control billions of electrons flowing across them per second. And that's on or off. The current's flowing, the electrons are flowing, transistors on, current's not flowing, electrons stop flowing, off.
Brandon Severin And it's like, okay, can I turn that into a quantum bit? I got billions of electrons flowing, but I only need one electron. And the genius part of spin qubits is...
Brandon Severin If I just slightly modify my transistor to give me a bit more control over the flow of electrons and still keep the same semiconductor process that prints trillions of these transistors every day but just say look let's just kind of tune this transistor up kind of like fast and furious you know you tune up a car add a bit of NOS or whatever yeah but very small very small can I start isolating individual electrons
Brandon Severin The answer is literally, yeah, you can. And the trick is basically a transistor works on having what's known as a gate electrode. It gates the flows of electrons going from one side of the transistor to the other. If you add a few more gates, you can basically start to isolate electrons between these gates. Well, it's not between them.
Brandon Severin all of these gates is all controlled by standard voltages, classical electronics. So no lasers, no vacuum. Same way your electronics work now. And if you finely tune the voltages you apply to those gates, you can isolate an electron. You can make it jump to the left, jump to the right.
Brandon Severin You can have it sit there for, you know, what seems like an eternity in the quantum world. In reality, it's not that long. I mean, we're talking like milliseconds. But because going back to, you know, your point on operation speed, you can get that electron to flip between states zero and one on the order of nanoseconds. Your gate speed is on the nanoseconds. So if you have milliseconds of your electron living in this quantum world that you've built for it.
Brandon Severin You can do thousands and thousands of operations within that time. And that's your spinky bit.
Turning an existing chip quantum
Mark Fielding You're taking an existing chip and using what's happening on the transistors and going down to the quantum level and controlling the electrons which are already passing in their billions through these gates.
Brandon Severin Exactly that. So it was like, think the late 2000s, you know, it's like, we've reached peak semiconductor that, the node size can't get any smaller. The transistors are so small. You used to say like electrons could tunnel from one side to the other. We're like, that's not a bug. That's a feature. Great.
Mark Fielding Well, that's the next step of evolution,
Does nature already hold the algorithm?
Mark Fielding technological evolution. Yes, incredible. So the Google algorithm, said if nature could run the algorithm, any quantum computer can run the algorithm. So if you have a trapped ion quantum computer, a superconducting or one of your spin qubits, it can run that algorithm. there's algorithms are qubit agnostic. They can run on any
Brandon Severin Ideally, yes. mean, you know, there are certain algorithms that will pertain themselves better to certain systems. It's just more easy to visualize and physically understand how that interacts with the underlying hardware. But that's the whole point of these algorithms is that, regardless of your quantum sort of basis or modality, you know, they work because algorithms are effectively gates.
Brandon Severin You know, the same way that the algorithm, let's say, we run on a laptop should also work on a classical computer made out of hole punch cards or buckets of water. You know, it's just operations on zeros and ones effectively.
The quantum circus
Jeremy Gilbertson whenever we talk to quantum guests, I always have this have this picture in my mind of a ringleader, taming a lion, under the big top, the lions, the cubit, and we're trying to like tame the lion in the corner. And then like the split second, the lions tamed Mark's got to come over and whisper a calculation in its head or a data set for it to rip on like is that does that a quantum circus, Jeremy, essentially you're describing a quantum circus.
Jeremy Gilbertson That's, ⁓ there it is. Trademark, you heard it first on Thinking on Paper, the quantum circus.
Mark Fielding Okay, in the quantum circus where you have all of the different systems entertaining the crowd, what can you do with a spin qubit that you can't do with other qubits? And
Why spin qubits could scale
Brandon Severin The big thing for spin qubits is that the manufacturing chains are already in place.
Brandon Severin it's not necessarily what we can do with them, right? That's not, you know, that's not answering the question, it gives you an idea of the sense of scale that spin qubits can approach without having to like kind of reinvent, you know, a whole new manufacturing paradigm. Because we are really trying to like leverage a trillion dollar industry, a multi-trillion dollar industry that's been like perfected by humankind for the past 50 years. When you think about
Brandon Severin And I think about the most impressive technology ever. It's definitely just what we can do with an EUV machine or whatever you use basically, but to pattern these silicon wafers into transistors and push humanity forward in ways we haven't seen before. spin qubits is very unique in that. Photonics come close and you can also make those an industrial fab.
Brandon Severin kind of need to really change the recipe. Physically what makes a spin qubit unique is you can move them. there's a lot of work in the spin qubit community now where people are literally figuring out how to, what's known as shuttle, move these qubits across. So we talked about isolating an electron, in your transistor. But let's say you want to let qubit here to talk to qubit here on the other side of the chip.
Brandon Severin How do you do do you use some quantum entanglement at a distance sort of effects? It's too far for that to happen. You actually can move, shuttle these qubits across the chip, get them to chat, move them back. The line tamer thing is real. And the amazing thing is it's such a, you pick the perfect animal, which is it's basically part of the cat family. And you can go all the way back to Schrodinger's cat.
Jeremy Gilbertson Get them to chat. So the lion tamer thing is real.
Brandon Severin Right? You know, you're born natural. And that's effectively what a quantum computer is. You've got Schrödinger's cat, which is this old thought experiment of what a qubit is and measurement and collapse and whatever. you've got a cat in a box, is it alive or dead? Well, quantum mechanics basically says that once you open the box, the cat used to be in superposition, i.e. being both alive?
Brandon Severin and dead or neither nor at the same time. But once you open the box, you measure it and it collapses into one of those states in being either alive or dead. With a quantum computer, you're basically trying to keep the box closed while you manipulate the cat into the likelihood of being alive or dead when you open it without killing the cat or keeping it alive. So literally exactly your lion tamer circus is very accurate.
Jeremy Gilbertson Very, very interesting.
Who observes the observer?
Jeremy Gilbertson so how do you, the biggest, the biggest hump for me, hurdle for me to overcome in understanding all of this when I first got into it was how literal I took the observation picture someone standing there in a lab coat with a clipboard, observing and nothing quantum can happen until this observation takes place and then.
Jeremy Gilbertson Who's observing the observer, observing the observer? Like, how do you talk someone out of that theoretical place into, let's land the plane and get tangible with it?
Brandon Severin tough. mean, because it, yeah, because the deeper you think about it, the, you, it's very philosophical. And I mean, there's, still two camps of interpretation of this, right? There's known as what's the many theory.
Jeremy Gilbertson It's hard, right? Yeah.
Copenhagen, or many worlds
Brandon Severin the Copenhagen interpretation and even that's not necessarily just the idea of the observer but it's related and you know there's this joke a great way of understanding the difference between the two is this the stroke is just like if I'm in Copenhagen and you see me there that means I'm there right whereas the many worlds theory says well I could be in Copenhagen I could be somewhere else in a different universe you're just not part of that universe
Brandon Severin What you believe is up to you. I prefer the many worlds. I think it's a lot more, a much more interesting way to live your life. Knowing that in some other universe, your lion could be still alive or dead, depending on what world you want to live in. Look, the thought experiments go on and on and on and some have actually never been solved. know, Wigner's friend and so actors or observers? Acting or observing, If we're to continue the quantum circus analogy and wrangling these cats...
Using AI to wrangle the cats
Brandon Severin How are you wrangling the cats? So you're doing something different, aren't you? Because you're using artificial intelligence to wrangle the cats, to wrangle the qubits, to control the qubits. Is that correct? Can you bring artificial intelligence into this conversation and how you're using it
Brandon Severin We're just trying to get the cat into the cage to start with. when we think about our billions of electrons, it's like, okay, well, I need to isolate one. So basically, I need to kind of form my qubit. I need to get my cat into the circus. And that's where we come in right from the beginning.
Brandon Severin And so the AI comes in where it's, okay, I'm applying these voltages to isolate my electron or puddles of electrons, you know, let's say 10 or so, and then I can, slowly deplete that puddle so one's left. And that's a really hard thing to do. But the joke is, is what you wouldn't believe is...
Brandon Severin You learn how to do this as part of your PhD if you work in spin qubits. You learn how to manipulate those voltages, let's say five, ten different voltages or so, by hand so you can isolate one or two electrons next to each other. I did it myself. At Oxford the first two weeks, my PhD, it took me weeks. I didn't even get close. I got to maybe 50, 100 electrons or so. I said, like, no way. If we want a billion qubits.
Brandon Severin You can't have a billion brandons, right? ⁓ And that's where the AI comes in. Yeah, literally, yeah. One in each of the worlds, you're fine. And that's where the AI comes in, which is basically, this is like loading the cat into the circus. You can think about it as like the bootloader for your operating system on a classical instrument. You you press that power button,
Jeremy Gilbertson Well, one in each of the worlds, right?
Brandon Severin files of a whole sequence of instructions as to what voltage is to set each transistor and to load what software from where. That's what we're building from the ground up. So it's controlling those voltages such that we can first form that qubit. And once we've got it, it's okay. Great. Let's do operations on that qubit. Uh-oh, you know, the qubit's gone out of whack. The electrons disappeared. Let's get it back. Let's recalibrate it, things like that. And to go all the way back to the Google result, think that's...
Brandon Severin You know, it's, that's why it's so hard. These qubits don't come for granted. Even when you have them printed on a chip, you really need to calibrate them and fine tune them to start doing interesting things.
Who the software is for
Mark Fielding who's this for? who should be listening to this going, okay, I want to learn about these different AI models that you have?
Brandon Severin first of all, it's for ourselves. We are our own users, as well as other people in the spin qubit and quantum community. The models are effectively analyzing the signals you get from current flowing through that transistor, saying, OK, based on, let's say, certain changes in current, what's the next best voltage I should apply to get me closer to
Jeremy Gilbertson Ooh, just like the word predictor root of what LLMs are doing anyway, right? We're just voltage predicting. if Mark is making spin qubits and wants to make more of them, he could probably come to you and say, help me automate this process. I want to make more spin qubits.
Jeremy Gilbertson I've got a pile of wrangled cats back there. I've got a lot of spin qubits, but they aren't behaving properly. Can you help me understand how to make those work better together? Is there an efficiency angle to what you're doing as well? Optimization?
Brandon Severin 100 % ⁓ software can work at night, it can work at all times of the day. PhD students and researchers get tired. And I think the important thing to understand is, this is basically the precursor of an operating system.
Brandon Severin That's where Mark one day will interact with this software without even realizing it. It's okay, I want to run my quantum algorithm, I want to design a new drug.
Brandon Severin you're doing that through conductor, know, conductor OS or whatever it's called at that point in time. And that goes and compiles all the way down to basically these voltages pulsing on individual qubits, individual.
Empire of AI, and the GPU comparison
Jeremy Gilbertson we have a book club and we just finished the empire of AI. Karen Hao's look at, the start of open AI and a lot of the behind the scenes things that a lot of people really fascinating read. But we break it down chapter by chapter. But the reason why I bring that up is the way
Jeremy Gilbertson kind of the rules of the road as we called it in book club for AI was like, hey, we just got to get more GPUs and we just got to scale, scale, scale, scale, scale. And then we'll figure out all the shit when it comes down and you know what to do with that and how to manage it. Are we doing it the right way for the right reasons? Talk to me about what the quantum industry is doing. Right. Is there a big push to drive and scale?
Jeremy Gilbertson Qbits like there is GPUs. Is there any correlation there? Are you seeing a different approach a better approach a worse approach?
The two camps in quantum
Brandon Severin the industry basically splits up into two camps, which is...
Brandon Severin I can build my quantum computer one qubit at a time. Or I can think about, okay, we're going to need at least a million qubits to do anything useful. and by qubits, mean physical qubits due to this.
Brandon Severin error correcting problem we have. So how am I going to build a foundations in a way such that I can, build for a million qubits? it's in a way it's like, I want to get to the moon, do I take a, do I try and build a rocket ship or am I there, you know, assembling ladders in a sense?
Brandon Severin What camp you sit in I think heavily depends on who you are actually as a person, in my opinion, and your thoughts of what's possible in terms of scale.
Brandon Severin There's a typical sort of thing where like often the people in the first camp of one qubit at a time is often very much like a university spin out sort of mindset where, and this is really how quantum started, you know, it's like at least physical quantum computing, which is I'm doing an experiment in my lab. it just so happens you can actually build a quantum or a qubit out of this. Maybe we could turn this into a quantum computer. And that's really how a lot of the early quantum computing started.
Brandon Severin came to fruition. And I was like, okay, but hold on, we need millions of these qubits. Just how do we get there? What platform will actually get us to millions and millions of qubits? How will we make them? How will we have the power to cool those chips down? Where will that power come from? Where will we have enough water to keep these chips, basically your...
Brandon Severin gosh, these cryomech pumps to keep your pulse tubes cold, all these things you need to think about when you're thinking about like million qubits sort of scale is very different to, I just want to get three qubits next year, then four, then five, and then so on. And I think that's really what makes spin qubits very unique. It's manufacturing-wise, we know we can make chips with the potential to house millions of qubits. But then at the...
Brandon Severin initial Lego block stage of, okay, but how can we get our one two qubit gates fantastic? How can we deal with this control and calibration problem that we're solving, which is just a huge bottleneck to scaling because we know we can make excellent one two qubit gates and so they can we've broken, you know, the required thresholds and records for that. The problem as well, okay, if I give someone a four qubit system, which is a spin qubit, that's like 20 individual parameters they need to optimize to just get the thing working and on. And it's like, how does that scale to millions of qubits? So we're very much in that million qubit thought process of that's where you wanna go, let's build for that.
Brandon Severin when you see the pictures of the beautiful Google Willow chip and they hold it in their palm, right? And it's like, this is the Google Willow chip. It has 105 qubits, or 103 qubits. It's like such a feat of engineering for them to be able to do that. like, in a chip that size, for spin qubits you can have like billions they're just a million times smaller but that's also the hard part is they're so small they're to control and so on and so
The vibe in quantum, and the vibe in AI
Mark Fielding referring back to what Jeremy mentioned about the empire of AI.
Mark Fielding You went to Y Combinator, you work in Silicon Valley in a startup, you are living in this world.
Mark Fielding Is it a different kind of person who works in quantum that works in AI? But now you're bridging those two worlds. Do the two groups of people come together? What's the vibe? What's the culture of quantum?
Mark Fielding Honestly, the vibe is very... I'd love to know what the vibe of AI was like, let's say 10 years ago, because it feels like quantum is kind of like at that stage, where we have these amazing demonstrations, and we can kind of sort of see this pathway to, oh my gosh, this is going to change everybody's lives, but we're not there yet. Because AI in 2013, 2012, and cats and dogs, and captioning images, and things like that.
Brandon Severin Obviously there are always going to be personalities, but the culture is very, very collaborative and supportive. And to give you an example,
Brandon Severin During my undergrad, I mean, I did it at Oxford and so like just across the street from my building, which was in material science, there was a physics department. And in the basement, somewhere in that building was Chris Ballance, who then went on to found Oxford Ionics two years later, that recently got acquired by IonQ for like, you know, don't know, one billion or something like that.
Brandon Severin I met Chris Ballance because I was working on a quantum project for like a few weeks, which was what would a quantum computer server farm look like? Who was the best person to talk to? Chris Ballance came across the road. Hey Chris, you know, what do you think this will look like? He goes, well, you know, like it's going to be tricky. We have all these problems, you know, we started chatting and me and another group of students. So just to give you an idea, you know, like, and I, and I think we're kind of entering this new wave where before it was very much like professor of X group founded quantum company and academia is always going to be rivalries, right? Because of effectively, you know, defined as an academic is you being first to do something, your publication record and so on and so forth. But I think now, you know, the students of those professors and the students of those students are now forming quantum companies and we're seeing like this huge community sort of grow and everybody support each other in different ways. There's many quantum founders in regular touch with, and I'm like, you're just seeing the age drop and the community build at the same time as well. It's very small. I'm always surprised at how small it is, but because of that, it's very beneficial and very.
Brandon Severin But you know, everybody kind of knows, right? Like, is Iron Trap's gonna win or Super conducting qubits gonna win or Swing Cupid's gonna win? I think though the smart ones among us also know like, there's probably not going to just be one modality. And we're already seeing that. Google quantum, for example, are doing great stuff with superconducting qubits, but so is atom computing and Quero with neutral atoms. And you can use their computer on Amazon AWS. Literally, you can log on and start running quantum algorithms. And arguably, they have more qubits. So we're already kind of seeing that. Okay, it's not really winner takes all. And I think because of that, the community is still very...
Kevin Kelly's question
Mark Fielding So our last question left, which asks all our guests from Kevin Kelly, what should humans be and how does technology help us get there?
Brandon Severin humans should just be like, obey the golden rule. You know, treat others as you want to be treated.
Brandon Severin I think having that role is very human in itself.
Brandon Severin because it gives an idea of long-term thinking.
Brandon Severin It definitely leads to it. How do we get there? I think technology helps us get there by...
Brandon Severin Providing insights to other people's experiences So whether that's literally like media discussions podcasts like this ⁓ or understanding of suffering and how to cure that suffering quantum computers, classical, whatever, you know, solving technological challenges.
Brandon Severin Uh, like honestly, I really see it as technology is kind of like this eye opener, uh, for humankind. Um, because, because you kind of see it already, right? Like you can know what's going on on the other side of the world where you weren't able to do that, where you were just, you know, like hunter gatherer tribe in South American jungle.
Brandon Severin But then it's like, okay, what do you do about that? Does it affect you? Do you care? And I think, yeah, as technology progresses, you're actually able to impact other people in ways that you just weren't able to before by flicking switches and playing with transistors and whatever. That'd be ideal. Literally treat others as you want to be treated and leverage technology to do that, whether it's through spreading the world, education, whatever.
Jeremy Gilbertson The next best startup that will help that is Empathy Activator. ⁓
Jeremy Gilbertson Quantum powered empathy activator presented by your friends at Thinking on Paper. No, that's great. That's great. What about a carryover question for the next guest? Can be about anything,
Brandon Severin It's kind of related to the first question, which is like, do you really care about what you're doing? particularly like other technologists and founders and also related to the previous discussion of like the community of scientists building this technology. Why are you doing what you're doing and do you actually care? I mean, I know for us at Conductor, it's like, I couldn't think of doing anything else with
Brandon Severin I really see a way that this is just how we have such a huge impact on the world. But I've also worked on projects where I'm just like, geez, man, can this be over already? Like, why, why, why?
Mark Fielding Unfortunately, unfortunately though, Brandon, this is thinking on paper and all our guests care.
Brandon Severin Good. Wonderful. Honestly wonderful.
Jeremy Gilbertson man, what a conversation. Brandon, thank you so much for chatting with us. Anything you wanted us to ask that we didn't?
Why is it taking so long?
Brandon Severin why is it taking so long? Actually, I think is a good question.
Mark Fielding Why is it taking so long, Brandon?
Brandon Severin Honestly, I think it's concentration of human beings and resources. I think it's this approach where either you want to build a rocket ship or you want to build ladders. And we don't have enough people building rocket ships.
Mark Fielding There's a finite supply of smart people. It was quite funny early, you said, the smart ones among us. And I was thinking, hold on, aren't you all incredibly smart?
Brandon Severin You can only know so much about like your particular, like I know spin qubits, but it's like, I'm like, can I design a quantum algorithm from scratch right now? Probably not. You know, the layers of the stack goes on and on. know, could I run an eye? Well, I know it's possible, you know, cause I've seen other people do it, you know.
Jeremy Gilbertson I love that you said probably not. We would say definitely not on our side.
Brandon Severin a years of hard study, right? But that's the thing, it's a few years of building the tool set and exercising that part of the brain.
Brandon Severin I don't know, look, it's different strokes with different folks.
Brandon Severin I do think that the progress we've made as a community is just comical in terms of how good it's been. 20 years ago, people just didn't think it was possible to have a qubit. They also didn't believe you could rotate the spin of an individual electron. During my PhD, I saw teenagers do it in their lunch break as part of their labs. Right? ⁓ on that, what do people think's impossible now that in 20 years, your look back go, well, they're doing it now on their lunch break.
Brandon Severin Yeah, the scale, the scaling argument of getting to millions of qubits. I think every quantum modality kind of has that like, you know, we can make one, we can make two, maybe we can make a hundred, but how do we really get to a million qubits? What does that actually look like? Everybody's like, And some people generally believe it's just impossible in certain modalities because they're like, the cooling requirements, the power. But I know, I do think, you know.
Brandon Severin within 20 years, you will have millions of qubits and it'll seem like, okay, cool. to one, wen billion qubits.
What one algorithm could change
Jeremy Gilbertson What's the likelihood I don't know, seven years from now, someone figures out how to make one qubit so freaking magical and so useful that they debunk this whole scale theory? No, not at all.
Brandon Severin There's a great graph and study done, literally by DARPA, the US Quantum Benchmarking Initiative, where they basically say, okay, if you get the qubit fidelity to 10 nines of fidelity or whatever it is, you will still need a million qubits to run RSA or whatever, to correct for those errors. Because to really give you an idea, in terms of fidelity, the bits in your computer, might...
Brandon Severin you know, make a mistake like one in like 10 to like the 23 times, you know, it just never happens. Whereas the threshold for quantum computers is one mistake in a thousand, you know, like 99.9 % fidelity. Big difference. And so you really have to be able to take out more errors than you are putting into the system.
Brandon Severin What could change if someone comes up with an algorithm that can take advantage of, let's say, lower fidelity qubits and, you know, Google's done work saying, okay, well, in theory, you could run RSA if you have like this fidelity qubits and with a tile.
Brandon Severin implementing that physically and experimentally, good luck, know, IBM have the TORI codes, it's, look, it's, it's once again, it's that rocket ship, it's like, if you can make many qubits, you don't need to worry about that problem anymore. And you're going to need that number of qubits to do error correction in the first place. And that's why, yeah, we're like very, very focused on leveraging silicon technology, because we just know you can print these things.
Jeremy Gilbertson There you have it.
Mark Fielding answer. Thank you. A lot to think on. Thank you, Brandon. It's been an incredible conversation. More layers of quantum for us to unpack at a later date.
Brandon Severin Anytime. I love talking about quantum I mean, we went down the rabbit hole, guys. Thinking on paper.
Why quantum feels different
Mark Fielding we speak about a lot of technologies, but it just seems there's something about quantum that...
Mark Fielding is different to the other technologies we speak about. when we have quantum, maybe it's just because we know so little and it's just so nothing spikes our curiosity like this. because it's so foreign really to us, but at the same time it is us.
Brandon Severin think this is one thing I want to do as a company is really demystify this foreignness. Because there are certain ways of thinking about...
Brandon Severin the quantum world that actually do make sense. know, it's like, you know, you have a kid and you offer it two Mars bars, right? One in the left hand, one in the right hand. It's like, okay, which Mars bar does the kid choose? The kid wants both of them.
Brandon Severin Right? That's effectively your electron in superposition. It wants to be around this positive hydrogen atom as well as the other one. Right? It wants to be in two places at once. And it sounds ridiculous, but hey, this is normal.
Mark Fielding I love that kids are just every time you speak to a kid, they're literally running through Schrodinger's cat experiments in their head because they want it all.
Jeremy Gilbertson That's awesome.




