 And I think the big thing you wanna know in this conference will be, will quantum computers break the internet? Okay? And through, you know, factoring them, be able to factor numbers and breaking RSA. Now, I'm gonna give you a quantum answer to this. And the answer is both yes and no. And the yes is because the experimental groups are making real progress now. And I moved to Google because we were able to get some things work and we think we know how to scale up. But the no part is that this was really, really hard, way harder than building a classical computer, way, way harder. And so, you know, the quantum nature of this is actually very applicable because this is interesting but hard, but the only way we're gonna know whether we can do it is measure it. Okay? We're gonna have to build it and see how it works and see if we can make progress and kind of measure whether nature is actually gonna allow this. There's no fundamental reason to think we can't do it but there's lots of practical things we have to butt up against. And I'm also gonna say, and you guys in a crypto conference knows this way better than I do is that I don't know everything that's going on. There's a lot of things going on in secret and we don't know everything. So there's a level of uncertainty there. And so what I'm gonna do here is I'm gonna try to tell you enough about how this thing works that you can kind of make up your own mind to some degree what's going on. And then as things progress in the future and read things in the news, you might be able to tell the difference what's going on. So I'll talk about all these things here, quantum computing, how an error correction, what a factoring machine gonna look like, 100 million to billions of qubits, okay? That's not as many as in here, okay? But right now we're doing experiments at 10, okay? So, can we get there? That's the question. Some general comments in the end and then hopefully I'll have time for a lot of Q and A. I'm sure there's a lot of questions here. Okay, so why is quantum computing so interesting? Normally we think of bits. It's either in the zero or the one state. In quantum computing it can be both zero and one. And the advantage of this is you can basically run through your quantum computer the case of zero and the case of one at the same time in a parallel computation. Okay, that gives you a factor two. That's really great, but by itself not that interesting. But of course what's interesting is you then put in two qubits. And that now runs through four states, zero, zero, zero, one, one, zero, one, one, which is double the number for two qubits. You add one qubit, you double the states. So in three qubits, that's eight states, four qubits, 16 states, scales exponentially, the parallelism. And that's interesting because by the time we're at 50 qubits, that amount of parallelism is kind of the size of the memory of a supercomputer. And we're actually thinking of doing experiments starting this year to be testing this kind of thing. Competing our qubits against the supercomputer. At least have the supercomputer check. It's not doing anything useful yet. At 300, two to the 300 is more states than there are in atoms in the universe. And clearly, even Google's not gonna be able to do something like that. And clearly anyone can't do that. Okay, so that's the idea. And then of course a big thing is the shore algorithm. And the shore algorithm takes a few thousand what we're gonna call logical qubits. And basically we start out with a bunch of qubits in the zero plus one states of this huge parallelism going on. And then you do your arithmetic computation where you modulil a exponentiate all these different numbers. And then to find the factor and when you modulil exponentiate it, there's a algorithm, I think it's Euler's invented it, where there's going to be a periodic repeat of the numbers. Again, and to find that tone, to find that frequency, we do a quantum Fourier transform and we do the period finding and then you could figure out what the factors are. Okay, that's all I'm gonna say about that because this is the easy part of building a quantum mechanic. So though it's beautiful theorem and beautiful result and only takes a few thousand qubits, okay? This is actually really hard to build, okay? And the reason is, is you're doing a bunch of arithmetic operations and the number of operations is kind of typical to what you might find in a cell phone application, but these are done with quantum bits and quantum bits are not stable like classical computing. Okay, this is stable, you know how it's gonna work, it's great, quantum bits are fundamentally unstable and you have to deal with that at the very fundamental level and that's what we're doing right now, we're doing experiments, the world's doing experiments, they're doing that and you need to understand how hard it is to do error correction, to know whether this algorithm will ever be done, that is the easy part. So let me explain a little bit about quantum, why there's errors in quantum systems and I'm gonna use the example of storing information of a coin, okay, as a class, first do a classical bit and of course who has coins in their pockets anymore so everyone has a cell phone, so treat this as a coin and I'm gonna treat this as a classical bit lying on a table, say, where it facing up is a zero and facing down is a one, okay? Now this zero or one state is very stable, I can jiggle this around and this state's not gonna flip over and I can in fact jiggle it hard so it's robust against errors and of course what happens is there's a force holding it down and if it does happen to pop up a little bit there's some dissipation in the system that keeps it either the zero or one as long as you don't jiggle it and force it too much. So classical information is inherently stable so you can make devices with very, very low error rates. Now in quantum information you don't have that stability anymore so think about this coin just kind of floating in space and the way to think about it, again this is zero and this is one but rotated 90 degrees pointing this way is zero plus one and in fact there can be a complex phase so this is zero plus one, zero plus I one, zero minus one, zero minus I one and that's kind of how quantum mechanics work and in fact there's a very good analogy of what's called the block sphere and block vector between information, classical information and quantum information, quantum information just lives in this 3D space. Now you have this and you have this in some state just a single qubit and you can see now that any little force on this is gonna kind of cause this to rotate and give you some kind of error. There's nothing holding it in some particular direction and this makes it intrinsically unstable and it's kind of worse than that when you go through the quantum physics and you can say well okay can I measure it somehow and any kind of measurement here it turns out with quantum mechanics it's basically coming from the uncertainty principle any kind of measurement that tells you it's zero or one this so this direction puts error in this direction and any kind of error to know the phase in this direction causes it to become uncertain in this direction. You can't kind of measure both those quantities. You can measure if it's zero or one to get an answer but by doing that you just don't have any information in that direction and that means you know any kind of attempt by itself to try to stabilize this is doomed for failure and you're gonna introduce errors. So that looks pretty bad and the beautiful part is is you know you measure amplitude random phase and vice versa. The beautiful part is if you take two qubits and then go ahead and measure their parity. So you measure the two of these but measure their total parity it turns out that you can measure their total parity amplitude or parity phase without screwing up amplitude screwed up phase or phase screwed up amplitude. This is not this is very strange and I'm gonna have to show you some quantum mathematics to explain that. So the way this works in quantum mechanics to try to figure out if one's gonna other you have to do with these operator algebra to have very upper weird operator algebra where if you flip the bit and then flip the phase that's differently than flipping the phase and flipping the bit. So let me do the thing I have this in the zero state and I'm gonna flip the bit while rotating it over and then flip the phase and then I have the cell phone pointing in that direction and then if I'm gonna take this and flip the phase and then flip the bit I'm claiming it's in another state. However, you probably couldn't see this but in both cases I have the top of this pointing towards you. So classically you would say well it's the same but in fact it's not. Okay and the way to do the mathematics is just to just right here it's simple mathematics a bit flip turns a zero into one and one to the zero and a phase flip puts a one puts a minus sign instead of the one. So let's just start with the state zero plus one. Okay we do a bit flip that's one plus zero we do a phase flip and there's a minus sign here. Zero plus one do a phase flip that's zero minus one and then do a bit flip and that's one minus zero. Note that this is the minus of this. It's not the same state. That minus sign has important implications in quantum mechanics, okay? So this bit and phase is not the same as phase and flip and in fact, and this gives you the quantum mechanics behavior, okay? That's what you need but notice it's what we call anti-commute. There's a minus sign here. So here's the trick. The trick is we measure parities, okay? So we're gonna do, now we're gonna talk about flipping two bits x one and x two take two quantum bits flip both of the amplitude or flip both of the phases and now we're gonna see if this commutes if the order matters. So I'm gonna write this out x one, two, z one, two, z one, two, x two and now I'm gonna flip the x and z's here using this relationship and I get a minus sign here and a minus sign here and because of the very complex mathematics of minus squared is equal to plus one. I think you can handle that. Okay, this is the same as this and it's zero, okay? What this means, this is really cool. This means that if you worry about the parities of quantum bits, okay, which are very familiar, it's phase and amplitude is a little weird but the parities of quantum bits because it commutes, it obeys completely classically. So all of your classical intuition can be used for example about ill building error correction codes and then people in quantum error correction codes have used all these ideas of classical error correction to build up some way to do it with the caveat that you have to be dealing with parities and not any other thing. You have to always deal with parities, okay? So you can use your classical intuition, it's pretty cool that way. So what I'm gonna do, okay, is now what happens, the fact that it commutes means that if you have two bits, okay, one and two and you measure this phase parity and then the bit flip parity, and you're gonna measure it over and over again because one is not gonna screw up the other which I said the quantum mechanics does. However at some point here, let's say this plus one goes to the minus one, you're gonna say, aha, I detected an error, okay? Now of course you obviously see, you know that one of the qubits was an error, which one, okay? You can't tell from just two. So the way to do is now you have three qubits. And again, if you're doing parity error checking, you know all this, this is what you do. You have three qubits, now you have a Z12 parity measurement and a Z23 parity measurements. And let's say you start these all in the zero state and the parity of course is zero, zero. Now if one of these qubits changes, you have a unique decoding with the parities. So if it's one zero, it's the left qubit. If it's one one, it's the middle. If it's zero one, it's right. So you can decode it, okay? Now of course if there are two errors between the fact that you do the parity measurements down here, then when you decode one zero and you think it's one zero zero you're gonna get the bit opposite. So your decoding will be an error, okay? And that's bad. So how do we get around that? How do we make sure that if there happen to be two errors we won't decode, okay? And the obvious thing is you just don't make three qubits. You make four qubits or five qubits or six qubits. You keep on adding up so that in order, so then with four qubits and three parities you'll be able to correct for one and two qubit errors but not three. And then you add more qubits and you can do errors. And then at some point it's just gonna be so unlikely that you're gonna have 10 or 20 errors, okay? That you can say you can correct for them all. Okay, and that's the standard way. This is all classical. I'm not telling you anything new here. Now this is for string of qubits for the Z, let's say bit flip errors. What do we do with a quantum circuit? We have both bit and phase and it turns out what you do is instead of a one dimensional for like bit errors what you do is you make a two dimensional circuit and essentially one direction is bit and one direction is phase, okay? That's the basic idea. When you have to work it out and carefully that's good enough for now. And then instead of two bit parities we actually do four bit parities both in terms of X and in terms of Z, okay? And then from this you can figure out what's going on. So for example, you're running this thing and then in this particular parity measurement here there was a change in parity which said that one of these four qubits screwed up and then let's say at the same time we also saw a change in parity of this one which says one of these four screwed up and the natural decoding would be it's the shared one right here. So you qubit errors identified with kind of adjacent parity measurements that come in pairs and in fact one can do some kind of minimum weight matching because it comes in pairs for very complex patterns and it's very efficient and you can build software to do that. Okay and then of course as before what happens when you have lots and lots of errors then you're gonna decode it in some way and you could be an error, okay? And basically the error goes as how error gets less as you make this bigger and bigger like I said before, according if this is dimension D that goes as some number to the D plus one over two and as you make D bigger this gets to be bigger and bigger so it's exponentially suppressed and this is the key thing is that in order for this to get bigger as you make it better the physical error rate has to be below about 1%, okay? Or else it just doesn't get any bigger. So there's a threshold and I think in classical error correction codes there are I think similar kind of thresholds. Again, this is nothing new, okay? And it just says you have to make good qubits, okay? That are better than you typically less than 1% error per operation. That's not easy but you can do it and then if you make it bigger and bigger you can do it. So typically we're gonna try to shoot for about 0.1% error so this lamb is 10 so if we make this 10th order this logical error would be 10 to minus 10 and then we could do something, okay? But you can see if it's a 10, it's a square ray there's a lot of qubits, okay? But that's the basic idea here, okay? So this is just a plot of how many qubits you need. Here's the 1% error. You wanna be below that and then what we wanna do is we want 10 to minus 10 to 10 to minus 20 errors, okay? And that requires an order 20 or 30 and that's gonna require about 1,000 physical qubits. So you need a lot of qubits, okay? So it's just what you need to do this and people are working on it to understand it better and making better qubits, wanna make better qubits and we wanna maybe make better codes and get this number down but that's the idea, okay? So we have made a lot of qubits, okay? So I wanna talk about what will a factory machine look like? So it's gonna be big and scary but I just wanna get you thinking about this as a hard problem. What would a classical factory machine look like? Just knowing what we do now looking at technology. So classical factoring of 2048 bits you have a data center right here, big data center and it's gonna be the size of good fraction in North America and excuse me for our Canadian friends, I put it up there because there's natural cooling going on to make it easier and you see there's even a little pipeline to Prudhoe Bay to make it easier on you and I'm sorry for the polar bears but what they're gonna like it because there's gonna be a lot of people to eat, okay? So if you did it this size, it would be about a 10 year run time, okay? And cost maybe it'll be a little bit less because of scaling but pretty expensive. I think the real problem is it's gonna consume all of the Earth's energy in about a day so I'm gonna say you just don't wanna build that thing, okay? So this is much more reasonable, okay? So we have a quantum computer, 100 million for the super-conducting qubits and that's gonna factor a number in about 24 hours. If you were to make it, oops, if you were to make it 100 billion it'll be about a thousand times faster. There are other qubits that are slower and they're gonna need billions of billions of qubits to make this thing fast, okay? So it's hard, okay? And some big refrigerator with all this stuff, lots of money, whatever, it's gonna be hard, okay? But okay, maybe it's not the North America. Okay, so let's talk about encoding of the quantum bits, physical encoding. You naturally would think about, well, let's use atoms. You take the hydrogen atom, you have the various orbitals, the 1S ground state, there's some two states that are S and P. You'd probably encode it in terms of zero as being that this hydrogen atom and one is this item, this is a stable state but this state will decay but it decays kind of slowly so that would be okay. And then you would put light between these two frequencies here to cause it to change between zero and one. And then you wanna interact them together so you're gonna make some kind of molecule and the natural interactions are gonna do something here. Okay, so let me tell you what the problem is with here is that if you build some kind of molecule the problem is you come with the light, the light is about 1,000 times bigger than these atoms so how are you gonna come in light and affect that atom and not the one next door to it? Okay, that's gonna be really hard and it's a technological problem, okay? It's just so, atoms are so darn small. You know, how are you gonna build a complex system out of that? I know with electronics we wanna make them smaller and smaller but it gets really hard. So what people do, physicists do is they make big atoms, they put ions that are a micron apart that still talk to each other, they build neutral atoms that are in some kind of light lattice where they can talk to each other and you can still manipulate them. There's a lot of clever things that are going on. What we do is we build integrated circuits. We call them quantum integrated circuits. In this case, they're out of superconducting metal, aluminum, they're not too hard to make. They're big size, okay? This is 100 micron, thanks Gail. And this is, the light here is the aluminum that gets reflected and where dark is where the aluminum has been etched away. And basically, these are just microwave oscillators. This little square here forms a capacitance between the cross and the ground and then this is a little Joseph's injunction non-linear inductor and you basically have an inductor capacitor resonator at six gigahertz that forms our qubit. And it has special quantities that look like a quantum system but in the end that, in particular, you have the current that's flowing between the ground plane and the center island through these Joseph's injunctions that's in a wave function that's both going up and down at the same time in a quantum wave function in the same way that this electron is up here and down here at the same time, that's the quantum. But the big thing here is, this is relatively easy to control because it's a large size. Here are little wires that come in, they're way smaller than this. You can bring them in, you can print this with integrated circuit fabrication, not too hard. Okay, so let's just talk just about how qubit operation is. What we do is we just let this relax to the ground state. So we're in the zero state and then what we're gonna do is we're gonna put on some microwaves at a resonant with this oscillation frequency and that's gonna cause a zero to one transition and then back to zero again. It's gonna cause this state to rotate, okay, like that. And then after a certain amount of time, which is known, we then measure whether it's a zero or one and we get a zero or one and we have to repeat it about a thousand times to get a probability. So this is a plot of the probability versus the time of the microwaves for a fixed power and you see you start here at zero at probability zero, one and then at 40 nanoseconds it goes to a one state and then another 40 nanoseconds it goes back to the zero state. So 40 nanoseconds is a zero state going to a one state and if you remember from your Boolean logic, that's a knot gate and then another 40 nanoseconds takes it back to the zero state and we know from Boolean logic that a knot times a knot is equal to the identity which is what we see. However, we can do things like pulse it for 20 nanoseconds which takes a zero state in fact to a zero plus one, okay? And if you do 20 nanoseconds and 20 nanoseconds you get to a knot. So that you're gonna call a square root of knot gate and the algebra gets obeyed anyway. So I understand if you have classical intuition of what a knot gate is, you have no idea what a square root of knot but it turns out the quantum dynamics of this, you can find a very nice algebraic form to describe it where everything is very precise and you know exactly what's going on. In fact, you can form these continuous gates which is what you need for the quantum system. So you see this is doing, it's nodding and it's doing all these things versus a various amount of times. You start seeing this pulls away a little bit from zero and one and these are the decoherence effects and basically the lifetime of this qubit is around 10 or 20 microseconds so we can do about 1,000 operations before it kind of loses its memory but we can get that to work pretty well. So I'm just gonna show you some simple results where we did this quantum error correction but making it for the full 2D array was really hard. We're gonna start attempting that at the end of this year, next year, it's doable but it was just too hard a couple of years ago so we just did a linear chain and we're gonna do the measure of the qubit, bear parity over and over, detect big flips. It's a realistic test of a surface code, kind of simple though and then we wanted to say is it possible to extend the lifetime by doing this? That was the big question, okay? So here are the nine qubits. You see the X's like we had before, the control lines are coming in the bottom and then these little microwave resonators are what we're gonna use to measure it and we bring in microwave tones and we measure the reflected phase coming out of all the qubits to know their states. I didn't go into that, it's a little complicated but you can build up the chip. Here's the chip in a lumen amount with a bunch of microwave connectors coming out of it. This is in a dilution refrigerator at 10 milli Kelvin because we want the thermal energy which is K Boltzmann by T to be much, much less than the quantum energy which is H bar omega so you have to go to low temperatures at six gigahertz. Fine, it's just something we buy. Okay, we have a bunch of filters and this is all done right and then that's down here and you bring up all the wires to a rack of a control electronics and then we sit here and program for days and days to figure out how to get this thing to work. So I like this picture because it looks staged and you can tell that because the lab is so clean. It's normally, okay. So okay, we do this experiment. I just wanna say there's a bunch of complicated waveforms and microwaves putting in or whatever. We have a bunch of qubits. We do this over and over again. This is a complicated sequence kind of starting to look like a real digital circuit. Okay, not as complicated as this but it is kind of complex and we can get it to work very well. Okay, and this is just the basic idea about showing how it's going. This is we repeat these parity measurements over and over again, eight cycles and if we don't do anything to the qubits, they relax in about 30 microseconds given here but if we do the error correction protocol and measure errors and we can correct for it, if it's five qubits, which is only corrects for first order, that's the blue line here and then if we use nine qubits, the whole sequence and correct for first and second order errors, that's right here. And indeed, we see that things get better and in fact it's the difference between first order and second order that tells you what that lambda factor is that I mentioned earlier, it's about 3.2 and it says as you make it bigger, the errors go down and down so that in principle, if we were to just add more and more qubits to this, we could make the error exponentially small. Okay, and in fact that's one of the things that we're trying to do now is make this much, much bigger so that we can make the error rates lower. So this number, this improvement here really shows that at least just for bit flips, okay, this linear chain, things are working well, okay? That is kind of the state of the art. I mean, other people are doing error correction and there's a bunch of other experiments but kind of doing it in a way that really simulates the full error correction, that's pretty good, okay? People have done some other things, proven some other things, but we're at a very rudimentary stage and we're working to make it better. So in talking about the future, I wanna mention a couple things that we're doing is besides trying to get error correction to work, we wanna do some tests on our qubits and we want to both make the qubits better and make more of them. There's an experiment we're doing right now. Okay, so first of all, let's talk about the exponential complexity. As I said in the beginning, the computation power, the parallel computation power goes exponentially with the number of qubits. This parallel state space we're looking at is to the end. That's the basis for all this. We all believe this is true. All the physicists believe quantum mechanics but this has not really been tested experimentally yet. And right now there are, I'm gonna say three billion dollar government programs looking at quantum information and more money than that, doing that. And we're spending billions of dollars and we don't even know if this idea that the state space grows exponentially is correct or how correct it is. We're all assuming it's gonna be correct. So we know it for about, let's say 10 qubits to a state space of a thousand, okay? Which is good. But will it go to two to the 50 or two to the 300? We don't know yet. So one of the experiments we wanna do is demonstrate this exponentially growing computational space. You have to do more qubits but they have to be qubits and we're gonna do that through a quantum supremacy experiment that was kind of motivated by John Preskill at Caltech. So for a well-defined problem we're gonna show our more computation power for qubits. Basically we're gonna run a 50 qubit experiment and we have to check where the output's right by doing a supercomputer calculation. To do that you need good fidelity like I've been talking about here for doing error correction. In fact you need about 0.1% errors per gate and that's good because that way we can validate our control and then go ahead and build a general purpose a quantum computer with error correction. So we're working on this and we think there's some interesting thing. So here's a chip we have in here. I'll show you the latest hardware. This is the chip right here that we talked about before and now there's two rows of 11 qubits. So it says 11 by two and that's a chip and then we take that chip and there are little indium bump bonds, little indium 10 micron spheres essentially on here and here and then that chip gets flipped over and does a bump bond press against this which is our wiring chip that brings the wires out to the edges of the chip and then that's brought here and then we wire bond to a circuit board and to a bunch of connectors that go up to our electronics. So we're at 22 qubits right now and we bring up a bunch of microwave lines and then we bring them down to a bunch of electronics and this is a crate that's gonna control about 20 qubits, 22 qubits or so. So if we wanna do, if you go look at a high energy physics experiment we can build racks and racks of electronics so we feel we can scale up the way more than 22 if things work right and we know what we're doing. Okay, so I'm near the end. So you wanna go away, hopefully you've learned something about this and you're gonna wanna judge whether our field is making progress and there's a big problem and the problem is fake news. Okay, no surprise. Just to give you an example when we did the mechanical oscillator showed quantum mechanics, the headline on one of the news organizations was physicists show time travel. Okay, but I would say most of the time it's pretty good but the problem and what happens in our field right now is we're trying to build systems which is really cool, systems that work and the journalist, and this makes sense, the journalist kind of think of this as a horse race. Who's ahead, what are people doing, whatever. And the natural way to think about a horse race is to say, well, what's the number of qubits, okay? And that makes sense because when you talk about some person, you have something here you wanna know how much RAM it does and how many CPUs. So it's clear what the metric is but that's totally wrong because this thing we all know is 100% reliable in terms of the computation where in qubits, one thing you learn from this is the qubits are intrinsically error prone and you have to deal with that. So I think the right way, if you wanna get a little, we're building systems, it's very complicated, it's more than two things but if you really wanna simplify that, the problem is it's a two-dimensional horse race where there's obviously the number of qubits but you need to know what the qubit error is too because that's something that's not zero. Okay, that's some finite number. So you have to be tracking this and you have to be very careful on this axis because the physicists will tell you their best error rate, okay? Because they're proud of it. But if you build systems for a living, you know it's the worst error that always kills you when you spend all your time doing that and that generally is how two qubits interact. I didn't talk about that in my talk but that's the hard part. So it's the two-qubit error. So where we are here right now is where we, I show data, nine qubits error is about a little bit under 10 to minus two, it's below the threshold, okay. And what we wanna do is get to 10 to minus three or lower and then we wanna scale up, okay? So that's what we wanna do and that's what we're working on right now. So here's like a competitor of ours, they have 15 qubits but it's three to 8% error, it's way up here, here's another competitor, it's two qubits to 10 to minus three but it's not done in a full system way. So it's interesting but whatever. So you know, we're still working on it but we have to go down this way and then learn how to scale it up and it's these two things that you have to talk about doing and we're moving, we're really working hard to move in this direction, I think other people are, it's clear you have to do that but you have to understand those two things. So where are we right now? I think we're closer to Kitty Hawk than on the moon, okay, we're trying to develop the technology going from there to there but okay, ooh. Okay, now for your audience, okay. Here's my, I would say it's a personal message, I don't, this is it. There is, okay, is the internet gonna get broken? RSA gonna get broken? Okay, there do exist and you know this more than I do, there, it does exist classical cartography methods that are known to be quantum resistant or at least thought to be known, okay. So RSA is used because it's known and reliable, everyone can trust it, okay. But there's quantum resistance code and not thought to be broken by quantum, that's great, okay. People are working on this and in fact NIST has some testing and standards program that's about 10 years, that's gonna take about 10 years where people can get together and test it and do it and that's really great, okay. And we need more research and we need to do that and let's make sure that if RSA happens to be broken in some future, we'll have something to stand. Now what's interesting is I hear comments that this is a nice program but this kind of 10 year timeframe is too slow, okay. Which I understand, you wanna make sure that you have a replacement, you know, as soon as possible. Now let me, sorry if this is gonna sound a little bit like giving you a hard time, but I kinda am. Building a quantum computer is really hard, I hope you see that, okay. And we have to build all this technology to do that. And people are worried that it's gonna be faster to build a quantum computer than it is to be write some software code. You know, thousands aligned to software code, okay. Is that really, is that what people are worried about? I understand writing the code and doing it right is really tough because you have to make something work and not break something, that's always harder. But I really think it may be a challenge to the community that, you know, come on, you guys have time to do this. You should focus on it. Let's not, you know, let's save the internet. Let's, you know, save the world, okay, with their data. Let's get this to work. And, you know, really try to work on this and solve this as fast as we can. I don't know if a quantum computer will be built in 10 years, but you can, I think you can certainly, you know, make sure that their alternatives are really good. You have the time to do that, okay. So that's kind of the conclusion of this. I now just, just a couple minutes, talk about why I am working on quantum computing. I mean, given the fact that, you know, factoring can be circumvented, the problem of quantum factoring can be circumnavigated by classical computing. I said, why do I want to spend my whole career on doing something that you guys can fix for me? And what we want to do is work on quantum computing to do something that'll benefit science and humanity. And I think there's something there. So it's not your field, but let me talk about what's going on. And a natural application is using a quantum computer to solve what the chemistry and physics is of other quantum systems. In fact, this was the original idea of Feynman way back in the 80s of why you would want to build a quantum computer. So a lot of people have been thinking about it, but I want to show you there's something really new and exciting going on here. So the idea is a bunch of the DOE computers are used for quantum simulation, okay? Things like the Haber process, which is creates ammonia fertilizer, consumes something like 2% of the world's energy, and yet there's bacteria that's known to be much better than that. Lots of other problems that one can solve if one could do a simulation. Of course, what's hard about quantum simulations for molecules is this exponential blow-up of the state space you need, okay? So a quantum computer has that resource. So let me tell you why I'm excited about this. And this is a little bit technical, but you guys know all about this. There's been huge progress in algorithms for this quantum chemistry thing. And I want to just briefly talk about that. In the 80s, Feynman came up with the proposal. Why don't we do it this way? About 2015, people said, look, it's not exponentially growing. It's a polynomial problem. We didn't know what the polynomial was, but we knew it wasn't exponential growing. And then various improvements over the years, and finally in 2013, they came up with a specific enough algorithm that it grew as size order n to the 11. Okay, so you're a computer savvy audience. When you hear something's n to the 11, you naturally say, well, that's not gonna work, okay? Which is true. And for example, if you have the number of orbital basis functions, which will be, let's say 50 to 100 or more, and the circuit depth, which scores scales as n to the 11, and you put on some rough scales here, n to the 11 is 10 to the 21 operations. It's the lifetime of the universe not gonna work. We know that, we knew that. But what's great about being specific is that people work on it. And I'm gonna say a lot of the work was done by the Microsoft group and other people too, and bringing this number down. And then what's nice is there's been some nice results down here recently coming down to something that was published on the archive about two months ago, where now we have an n to the 2.7 for an exact solution or order n for an approximate solution. So that's down here, and now you see, maybe solve the problems in less than a second, way more efficient, okay? And for this particular case, it's 100 logical qubits, you need error corrections, so maybe a million qubits, physical qubits, this can be hard, but this is kind of in range a little bit. This is approximate, it might work with 100, even 100 physical qubits. There's some insensitivity to errors here, so we have to try this. And that's something we're gonna build pretty soon. So this is very exciting. I wanna note that there's this guy, this Babish guy who's been doing a bunch of stuff here. Google hired him last year, so we saw what was going on. Okay, so that's the end of my talk. Here's our quantum hardware team, and they've been working very hard to get this all put together. And we've been Googling about three years now, and they were kinda just been kinda working on our infrastructure, and starting this year, we're starting to get things put together, 50 qubit machine scaling up from that, and we hope to show some more results in the next year, and see what we can do in terms of error correction also. Okay, thank you very much.