Mentioned in the episode
Global Risk Institute 2022 Quantum Threat Timeline Report
The Center for Quantum Technologies in Singapore
Wikipedia page on 2 nanometer process for microprocessor fabrication
Creators & Guests
What is The New Quantum Era?
Your hosts, Sebastian Hassinger and Kevin Rowney, interview brilliant research scientists, software developers, engineers and others actively exploring the possibilities of our new quantum era. We will cover topics in quantum computing, networking and sensing, focusing on hardware, algorithms and general theory. The show aims for accessibility - neither of us are physicists! - and we'll try to provide context for the terminology and glimpses at the fascinating history of this new field as it evolves in real time.
Sebastian Hassinger 0:02
Hey Kevin, how are you?
Kevin Rowney 0:32
Hey Sebastian, how's it going?
Sebastian Hassinger 0:33
Very well, thank you. I'm very excited about our conversation on this episode, we talked to Joe Fitzsimmons, who's the CEO and founder of Horizon quantum computing, which is a quantum software company based in Singapore is really struck me as very intelligent, which is not unusual in the quantum space, but also really good, well reasoned sort of thinker, he thinks things through very thoroughly. And he articulates his thoughts and really clear manner. And we actually reached out to Joe, because there was a recent some recent coverage of quantum computing in the mainstream press that generated some sort of, you know, hyperbolic responses on both sides, I think, positive and negative to the, to the coverage. And Joe had a Twitter thread that, I think laid out a very well thought through position, that balance the the pros and the cons of, of that type of mainstream article and just the overall perception of quantum computing in a really interesting way.
Kevin Rowney 1:47
Yeah, exactly. I felt like he has just a nice CEO, sober, balanced tone of both being, you know, skeptical about some of the hype, but also he's, he's a long term, you know, within the next decade or so optimist. That's a hard balance to get right. I mean, if you're somebody of, of rigor and authority in the field, I think he does it quite quite well. I just just for the sake of helping our audience a lot. Some of the controversies related to this time, cover story, I guess, called what quantum computers can solve countless problems and create a lot of new ones. That first hit the wire back in January 26 of this year. So yeah, that was this the kickoff of some, you know, kerfuffle, within the Quality Committee space. Yeah, and
Sebastian Hassinger 2:32
some of it was warranted. There's definitely some factually incorrect claims made in the article. But some of the some of the criticism, I think, was over the top because and it happens whenever there are clear sort of statements that quantum computing has this enormous potential for disruptive change or transformative change. They'll whenever that that type of statement gets made, you will find people sort of throwing stones, I think, and just saying like, you know, this is purely smoke and mirrors, it's all theoretical, you haven't proven anything, it's never going to work all the various good arguments, and neither of those reactions is is fully unwarranted or fully justified, that delicate balance that Joe really struck, and we just wanted to talk to him to give him a chance to talk about you know, what his thoughts were behind that Twitter thread in more detail. Since you know, no longer how I know no matter how long the Twitter thread is, it's always still a Twitter thread.
Kevin Rowney 3:36
30 Yes. Okay, so here's, here's the interview with John.
Sebastian Hassinger 4:12
Okay, we're here with Joe Fitzsimons, CEO and founder of Horizon Quantum. And Joe, you're currently you live in Singapore, right? Yeah, I do. Thank you very much for joining us at this time. How long have you been in Singapore?
Joe Fitzsimmons 4:28
I moved out here at the end of 2010. So this is I think, the start of my well, a little over 12 years.
Kevin Rowney 4:36
So you must feel like a local button. Oh, wow. Yeah.
Joe Fitzsimmons 4:38
Well, I mean, it takes it you know, it takes some time. But it's you know, it's probably one of the easiest places to live and to fit in.
Sebastian Hassinger 4:48
I've been very impressed with Singapore when I've when I've gone there. Just went there last several weeks. We saw each other last summer when it was there. So So I would love to hear how you ended up starting a quantum Startup acquiring computing startup. Can you take us through that history? Yeah, sure.
Joe Fitzsimmons 5:04
So I mean, I guess different people have come to this industry through different parts. Mine, I guess I see as the kind of most traditional path in some sense, even though it's a very young industry. So I, I did my undergrad in theoretical physics and went straight into a PhD in quantum computing. So I was exposed to quantum computing in the last year of my undergrad, which would have been 2003 2004. To what school I did the undergrad in University College, Dublin. Okay. So I applied for PhD programs. I got accepted to one in NUI Maynooth, which is a small university in Ireland a little bit outside Dublin. So I went there to start my PhD. And then, you know, significantly less than six months in my supervisor met with me and explained that he had accepted a position in Australia, and I could move over there, that there'll be a place for me there, but you know, the way things were going, I was going out with a girl at the time, we're now married. So I mean, I guess I made the right choice to not be I
Kevin Rowney 6:20
guess you did way to go. Yeah.
Joe Fitzsimmons 6:23
Yeah, so I started looking for new positions, found another opening and Oxford moved over there. I already started the PhD in Ireland. So, you know, I had some experience from there. So I went through the PhD, while the D fell in Oxford pretty quickly, so 2005 to 2007. And then as a fellow in Merton college for up until the end of 2010.
Sebastian Hassinger 6:49
That's pretty unusual for an undergraduate program to have quantum computing. At that point in time, that's very early adopter kind of mentality. They're in Dublin. Do you think that's typical of the I mean, is there some reason why there is that sort of advanced thinking of the university?
Joe Fitzsimmons 7:08
You know, I can't, I can't necessarily explain the logic behind it, because I was, obviously I wasn't part of those decisions. You know, I've been involved in curriculum decisions since I was a professor after I moved to Singapore. But, you know, it was an odd, it was kind of an odd arrangement, because it was the only course that we didn't take in the university. So we have to go to the Dublin Institute for Advanced Studies, which is somewhere between UCD where I went to university and Trinity College Dublin, people from both, from both universities, were going there for this one quantum computing course. So, you know, there was a professor there who had worked on some quantum information problems, quantum computing problems. So it was, you know, it was an interesting course. was interesting to get exposed that early on. I mean, I say early, I mean, it was a decade old at that point. Yeah. But still, there's still
Kevin Rowney 8:15
that's way back could be impressive. Yeah. Yeah. stuff. I'd be early start. No doubt.
Sebastian Hassinger 8:19
Absolutely. And defining your your PhD, when you said you did a PhD in quantum computing, was it sort of formally recognized as being specifically quantum computing or quantum information science or
Joe Fitzsimmons 8:33
Oxford details? Don't say departments on them, they just say the title of the thesis, they say. So mine is like architectures for quantum computing under restricted control. Interesting, very cool. So it was mostly about you know how to build a quantum computer, if you can't get access to all of the operations you may want. So if you have to entangle qubits by making optical measurements, for example, like now, or if you have to do the same operation to every qubit, because, you know, perhaps you can't resolve each individual qubit. Interesting is out, it's possible in both cases, you can make universal quantum computers. Right.
Sebastian Hassinger 9:17
That's interesting. So I mean, at that point, so that what was your again, when you when you were doing your PhD by the time you got to Oxford is 2000
Joe Fitzsimmons 9:25
tests? No, five to 2005.
Sebastian Hassinger 9:27
Right. Yeah. Right. So very early days of experimental realization of qubits. And I think there's some superconducting early experiments NMR primarily, and some early trapped on at that point, right.
Joe Fitzsimmons 9:41
Well, so NMR and trapped ions have been around for, you know, close to a decade at that point of experimental demonstration of of quantum information processing. Superconducting qubits were at a much earlier stage of development. So if you look through the kind of history of superconducting qubits that coherence time on superconducting qubits has changed dramatically over over the last 20 years. So ions less so. Right. But ions are the same. Now, as they were, you know, as they've always been this. They weren't the Big Bang,
Sebastian Hassinger 10:19
presumably, as they always will be, right.
Joe Fitzsimmons 10:23
Yeah, exactly. So, you know, so you see less change there than you do in superconducting qubits. But the net result of this is that superconducting qubits were not the most promising system right at that time. So yeah, my knowledge of superconducting qubits significantly less then other other kinds of systems that would be like NMR, or ion traps or optical quantum information processing, or like, NV centers, things like this.
Kevin Rowney 10:51
Right. That's an interesting perspective. Again, I wonder I do do would you classify as an optimist or a pessimist on the trends Bong architecture, the superconducting qubit architecture? What do you think? Is it still in your mind the plausible leader to get to quantum advantage?
Joe Fitzsimmons 11:07
Oh, I mean, you know, it's certainly picked up steam over that intervening 17 years, there's been, there's been a lot of development. So what you know, what I would say is that I'm not, I'm not an experimentalist, I'm not the best one to tell you what the best way to build a a quantum computer is. And I am occasionally reminded of that. But if you look, you know, if you look at quash the community kind of judges. One way you can do this is there's a report that's written every year for the global risk risk Institute, Marco piani, and McCallum, Oscar radish. And they asked about 50 people in the community, generally experts in different areas, algorithms, error correction, fault tolerance, and, and experimental systems. And they asked them, basically questions that are used to estimate a timeline to quantum computers posing a risk to cryptography. And they were particular to RSA 2048. And one of the questions that you're asked, as a respondent on this survey, is to rank the systems as most likely and least likely, in terms of getting to this threshold. Now, it's not asking you about 1000 years from now, what will quantum computers look like? You know, it's about this one particular threshold area. When you when you build a system with about 6000, logical qubits that can, you know, that can support, you know, several billion or 10s of billions of operations. Without error, that's, you know, that's a particular size system. And consistently, the top two systems that come out in that ranking are ion traps, and superconducting qubits with photonics coming third. And then you see spin systems, you know, silicon, quantum dots, and so on coming behind.
Kevin Rowney 13:15
useful overview. Thank you. We'll probably just post a link to that. The reports or the notes, no, no, thank you. Definitely.
Sebastian Hassinger 13:22
So then, what was the the the transition from from Oxford to Singapore? You said, Actually, you were a professor, I think at NUS Right? Was that what brought you to Singapore initially, so
Joe Fitzsimmons 13:32
what happened was, I was coming to the end, I was a fellow in in Merton College, which is really nice gig, if you can get it straight out of your PhD. You know, essentially, you don't have a boss. You do your own research. And you may not be paid terribly well. But it's a very nice lifestyle of being able to have your dinner in college and go punting on the river and stuff like that. But yeah, it's coming to the end of that three year appointment. So I was looking for other positions. Arthur record was a professor in the in the same Well, he was a fellow and Martin as well. So he had also been the one of the examiner's on my PhD thesis. So he was my internal examiner. And he suggested that I should apply to Singapore, because he was at that point director of the Center for quantum technology, which was kind of getting up and running at the time, it had only been running a few years. And I visited once or twice, I think two times in late 2009. So it seemed kind of attractive. My wife is Chinese. So actually, culturally, it's a good fit as well, in terms of location. So you know, she can get the food she likes and so on.
Sebastian Hassinger 14:55
So yeah, it is very good in Singapore.
Kevin Rowney 14:59
It goes until I've just for the food. Oh my god. Yeah.
Joe Fitzsimmons 15:03
Yeah. So I mean, it ended up, it ended up being a good good fish I moved over, as I said into 2010 initially as a senior research fellow in the center for quantum technologies, okay. Then I moved to Singapore University of Technology and design, I got a National Research Foundation Fellowship, which is quite a big personal grant that they give you in Singapore, or in your career, or I mean that they give some people early in their career. So it's about three and a half million dollars. So it's quite a lot of money for starting a group. So I built up quite a large research group at that point, I had already been supervising a postdoc in CTT. But that kind of expanded the group grew to about 14 people got tenure there in 2018, and then resigned to start horizon. The company existed on paper before that.
Kevin Rowney 16:05
Okay, since but still very impressive. Wow. It's just a great anecdote. Wow. Good ride.
Joe Fitzsimmons 16:15
Kevin Rowney 16:22
Exactly. Thank you, for the for the overview of this is these are always fun conversations. Just a little warm up to hear your your pathway. I mean, one of the the key issues that the themes that we've been talking about on this podcast for some time is the balance between, you know, how do you get to a clear sighted, you know, conversation about the future of quantum computing? Right, it feels like there's, you know, a lot of hype that I think it comes from either people that are not well informed about it, or you feel like they can apply the history of Moore's Law to quantum computing, or, you know, also some pessimism which we think is sometimes warranted, and sometimes a little bit too much. How do you find a balance that really sees both the potential and acknowledges the significant problems. So that was one of the bits of your writing that we saw that really drew us to seek you out for an interview. That was I mean, that feels like the substance of a lot of what we're trying to delve into in this very podcast. So we were hoping we just do a little back and forth with you on that on that topic.
Joe Fitzsimmons 17:20
Sure. Sure. So I mean, I should say, sometimes I write these Twitter threads when I've become a little bit frustrated with something. Something else I've seen on Twitter or something like this very often. But
Sebastian Hassinger 17:35
it was a very gentle rant. If you're saying it was a rant, it was very, very
Joe Fitzsimmons 17:42
well, you know, there's a couple of things you need, at least from my perspective. And to be honest, I think most people don't think about it clearly. I think that there has been a bifurcation in the community that pulls people towards two extremes, and neither of those extremes is correct. And I guess you can see how happens, right? That, you know, one thing when you come up to the academic system, that is very important to you is, is let's say intellectual honesty and rigor. And it's, it's really important not to be telling stories about the world as it is supposed to be that are not accurate. So, if you, you know, like, if you look at it this way, you can look at some of what has happened in the quantum industry. And you can see a lot of hype, you can see things, you can see plenty of statements that are not true. So, unfortunately, in the in the article that led to that. There are a number of things that are factually incorrect.
Kevin Rowney 19:11
And shortly to run through just for the benefit of our audience, you're directly referring now to that time cover story, Time Magazine, and it would just be helpful for our audience who may have read it or may not have I mean, a quick if you could a concise itemization of the the key ideas, it could go on too long. I'm just trying to again for the benefit of those who are listening, get us.
Joe Fitzsimmons 19:35
Sure. So, I mean, what I can say is I thought it was cool to see quantum computing on the cover of Time. My colleagues disagreed. In terms of the in terms of the criticism that has gone around of the article, essentially there is several aspects of it. First of all, it gives a An incorrect statement about how the how the world currently is. So it talks about how quantum computing is currently being used to optimize logistics to make healthcare decisions and things like this, and are really known as using quantum computing to make healthcare decisions.
Kevin Rowney 20:23
Our global supply chain, right.
Sebastian Hassinger 20:26
So they cause
Kevin Rowney 20:29
Joe Fitzsimmons 20:31
That is, weirdly obfuscated, from the from the point of view of if you're, you're an external observer looking at quantum computing, it's not clear how useful quantum has been a quantum computing has been to this point, those two different industries and so on. Like, there's a lot of, there's a lot of claims out there. What's really important to understand is it has not been the there has yet to be a clear demonstration, that quantum computers can solve a meaningful problem more efficient than conventional computers, would existing hardware, that milestone has not yet been reached. We know it's possible, we know you can factor large numbers, and you can break codes in ways that you cannot do with conventional computers. But we haven't reached that point yet. Right. And we there's lots of applications that we believe will be useful. But one of the things that's happening at the moment is as quantum computing becomes more commercial, there are quite a lot of industrial engagements for at POC projects, where people are applying quantum computing to various different kinds of commercially oriented problems, writing up results, putting out press releases, because I guess, for many of the companies, part of the value of doing the POC is the is the PR value. So this leads to a situation where people believe, potentially that quantum computing is being widely used to address all of these problems. But it isn't yet. It's not yet at the point of actually offering an advantage in production. In fact, most of the hardware that's out there, you're not allowed to use for production.
Sebastian Hassinger 22:17
It's interesting, you started by by framing this as you know, when you're coming from academia from scientific research, there's, you know, it's considered not good for him to represent things as as you want them to be, rather than as they are. It's interesting to me, because I think, in part, what's happening is this is this is tech marketing, right? It's actually it's a collision of two different mindsets and tech marketing, you know, fake it till you make it sounds sort of sinister, except that you have to bear in mind that the pace of technological innovation was sufficiently is sufficiently fast in classical digital technology is where you can sort of go like, Oh, we're going to be able to do X, because six months from now, you will be able to tell. And
Kevin Rowney 23:10
those early movers those people that made the original bet, I mean, they it turned out great. Right. Yeah. Right. In that expectation is gonna go that way is pervasive. Yes.
Joe Fitzsimmons 23:18
Yeah. So I mean, I agree. I agree that culture is that, that contributes to this? What I would say is that, at least for us, in terms of Horizon, we've been, you know, revisiting our values. At the, you know, actually, we're going through the process at the moment, but one of the core values for us, is integrity, both in science and in business. And, you know, you have the potential issue of misleading people about hard sciences, that being the, you know, seeing men to be a violation of what you might think of as a kind of scientific integrity, but you also see it when it comes to companies like thoroughness, where that also leads to massive issues, in terms of business integrity, and like, growth cannot be said in the same sentence as, as companies like that. Right? So quantum computing has an issue that is not yet fully realized. And if you add on to that the misleading statements, then you get you into a real into a real troublesome area. But but the other side to this, so you need to be clear about where you stand. And we should be able to agree on where we stand now. And if we're not able to, there might be a problem. Now, maybe, you know, something, I don't know. Maybe someone knows about progress within a company that is not public, or you know, within a government job or something that is not public, and they know Oh, something's been achieved that I don't know has been achieved yet. That's so that's completely possible. Absolutely probably wouldn't expect it to be a massively rare in history. So we should have a fairly, we should have a fairly good estimation of approximately where things stand at the moment. You know, then there's the question of where are things going. And that's where I think particularly the clash of cultures can also go really different directions. So, you know, an issue here has also been that there's a culture when you're coming from the academic side of there's certainly pressure to not do something that will be bad for the community. And one of the things that has been seen historically, as being bad for the community is saying that you're going to build a quantum computer. Because the idea is, well, I mean, at least my sense of where this is coming from, is that if you say you're going to build a quantum computer, you draw in funding to do it and you fail to deliver, then those same funding agencies will be reluctant.
Kevin Rowney 26:22
Right, right. Right. For other teams that might be able to do it. Yes. Right. Yeah.
Joe Fitzsimmons 26:26
And to be, I mean, to be brutally honest, every experimental effort believes 30 own experimental effort that can work.
Sebastian Hassinger 26:33
And notice that it is a bit of a religious landscape in a way, it's like, my qubits are the only true qubit.
Joe Fitzsimmons 26:44
So I mean,
Kevin Rowney 26:46
it's just I'm sorry. For the backstory of this would be how many research projects tried and crashed and burned on the prominent successful examples, but the ones that are dead or survived, forgotten? Do you know, how many have been shuffled? I
Joe Fitzsimmons 27:01
mean, if you look back, even before I entered the fields, there were attempts to build quantum computers that just turned out to be much harder than much harder than people thought they were going to be. Certainly, in the, you know, I've gone through several phases, I have this. I have these slides that I show sometimes in public talks of the the number of qubits with time, and it kind of increases up to about 2005, and then kind of stagnates and then starts increasing again, around 2016. So, you know, there have certainly been situations, I don't want to get into this too much, because, you know, sometimes these are attributed to particular individuals and so on. So I don't want to Sure, there definitely been situations in the past, where people have felt that this has happened. For sure it, there was a very strong backlash against D wave in the mid 2000s, for making claims that they were building quantum computers, right, fairly demanding off. Right. And that, you know, that led to, you know, that there's many there's many examples on this front, but it happens in that academic realm as well. So there's this notion of endangering funding. And I think one thing that's sometimes missed is that the funding won't continue forever. The funding is now at a higher level, because people sentence closer. That's right. If we don't deliver now, that fundings going away, you can't preserve you're actually have to use it to make forward progress, right.
Kevin Rowney 28:51
A bit of a temporary a temporary winter. At the end, maybe quantum funding would return to me just like artificial intelligence and others there. Were there ebbs and flows of Spike
Joe Fitzsimmons 29:00
spikes above the ai ai winters long.
Kevin Rowney 29:03
Yeah, it was not fun, right. For a lot of people. Yeah.
Joe Fitzsimmons 29:07
So you know, there's, there's a couple of sides to it. But yeah, I mean, that I mean, you could burn funding for 15 or 20 years. Right. Light. You sure. Like if the current moment, if it fails, you know, we just stole out a qubit, you know, a couple of 100 qubits that are noisy, and we cannot get beyond that. Then it's, it's going to put everyone off quantum computing for quite a long time. Right. Now, I don't think that's going to happen. But, you know, there is
Sebastian Hassinger 29:42
a danger. What you're saying is that that is in the back of the mind of people who would basically we have a negative opinion. He said to me harm the community that taking that funding is setting up a failure condition that'll that'll spoil the funding. Well for a long
Joe Fitzsimmons 29:59
time Yeah. So, I mean, the other part of this article, though, is forward looking statements about how disruptive quantum computing and quantum computing may be, and various implications of that. But that's where I feel like we really, you know, get into a situation where it's not about being honest about where we currently stand. It's about different possible futures. And realistically, I'm not sure. If you come up through a path like I did, you're all that much better placed to talk about where things are going than if you came up through the tech industry, for example.
Kevin Rowney 30:44
Sebastian Hassinger 30:45
You know, it's speculative. It is your imagination,
Joe Fitzsimmons 30:50
missing things. So right, you know, if you come up to the academic path, you don't necessarily have the same perspective in terms of how capital can change, you know, can lead to rapid change, because mostly you're talking about small research groups doing things. And right now, okay, they, I mean, they can have more funding or less funding, but it's not, it's not at the point of like TSMC, building a fab, you're right, kind of level, right? So if you look, right, the projected costs of two nanometer fabs, they're about the same as all of the money that has been spent or committed in quantum computing over all time. In terms, including forward commitments, right, so. You know, I don't think people in the field really have that sense of the scale of the scale. Yeah, what happens, and I also think they're a little bit more distant from what has happened in the history of computing. I looked it up the other day, and the first computer I had had in the processor had three and a half 1000 transistors, right. Right, you know, my laptop definitely does not have three and a half 1000 transistors, probably just
Sebastian Hassinger 32:15
probably hasn't run that in every key has three and a half 1000 transistors every human sport. So
Joe Fitzsimmons 32:22
we're, you know, we're getting to a point where, actually the, the number of qubits we have, and so on, they're starting to look not that dissimilar from, you know, reasonably early PCs and so on. We lack memory, though, right? It's an issue. And we haven't really gotten to a point where we've componentized, the right systems. So at the moment, everyone's trying to build these giant, monolithic processors that are everything. They're all of the memory, they're all of the processing, and there is no structure to them necessarily. There's no ALU, there's no fpu, you just have essentially like the equivalent of an FPGA that you're that you're trying to, you're trying to program
Kevin Rowney 33:10
a sequence of gates out and run where you go exactly like everything is
Joe Fitzsimmons 33:15
software defined, because it's defined in these pulse sequences that are going to the devices. For most for most systems. I guess optical systems are a little bit different, in that sense. But yeah, so you're, you're in the situation where, actually, the systems are starting to look like systems from maybe the history of computer noise. A bit high. But we've had noise before in computers. Yeah, you know, error correction is used now in, you know, whether that be on large scale systems, or what are being communications links, and so on.
Sebastian Hassinger 33:54
Do you think part of the issue might be like, over indexing on you know, Shor's algorithm with sort of the, you know, the quantum quantum shot around the world? And as you say, You referenced before you get, you know, 6000 fault tolerant qubits, and you can crack most codes that exist today? Are we like, expecting too much? Are we expecting to get to there like a big bang, where we get a system that does that, and that's where everyone's sort of anchoring on that level of functionality, maybe not specifically around cryptography, but that level of functionality of a very robust, mature system. When, as you said, your first personal computer at three and a half 1000 transistors, there was a whole era of that we forget about now of even pre microprocessor electronic computers that were doing real work in specific use cases that might actually unlocked that industrial interest and skin therefore scale that you were describing before.
Joe Fitzsimmons 34:55
I guess the answer to this is that I don't really know for sure are smaller than Shor's algorithm, you can do it with smaller than 6000 cubits. I think there's estimates that some of the chemistry algorithms start to work around 200 cubits right. But these are low noise qubits. Now, if you asked me, what, what path does quantum computing take from here? I'm not sure that it's, you know, each qubit encoded in 1000 Physical qubits. I'm not sure I believe that that's where we go. We might demonstrate it, you might see it demonstrate without
Sebastian Hassinger 35:33
being the IDs, that being the leading sort of strategy for error correction is having a logical qubits made up of a physical collection of noisy qubits.
Joe Fitzsimmons 35:46
Yeah, so quite a few efforts, look at surface code encodings, for example, where you're imagining encoding a single, logical qubit, into maybe 1000, or even 10,000 Physical qubits. And that's very wasteful. We don't do that in classical computing, right. And there's not really clear reason why we would have to quantum mechanically, we know that it's possible to do many too many encodings, or a few too many encodings, rather than one to many. And that would certainly save a lot. But it's less explored from the point of view of for target sets, and so on. Interesting. But really, these things are very sensitive to the noise rates of the individual qubits. If you bring that noise rate down, you need less qubits for the encoding. And that's clearly what has to happen. And that's, you know, that's an argument for why ions are good, because actually, there's not that many different sources of errors, and you can start knocking them off one time. And as you knock them off, you just gain more and more nines. Now, I'm not saying it's easy to not often sources, but least finite ly many of them superconducting qubits is a little bit different, because they're manufactured. They're not, they're not perfect by by their nature. But still, you know, the performance of superconducting qubits has increased, and increased and increased and increased. And yet, we're still a long way from having essentially everless qubits. But you can see a path where actually the control over the systems get significantly better so that you're well below the photon threshold. And at that point, the effort you need to put in is much smaller when he was doing so if you tell me that you are going to try to make a full tolerant quantum computer and your individual qubits. They haven't they are good to, let's say, you know, seven nines or eight nines, then that's not gonna be so hard for you. If you tell me they're good, too, it's gonna be really hard. Yeah. Enormous. So
Kevin Rowney 38:11
interesting analogy or comparison you're making between you know, how, at some point quantum computing might take on the shape of maybe a Moore's law, you know, kind of like growth curve, as you say to students silicon for the past many decades? Well, I mean, well, it is already. Yeah, maybe it was starting to I mean, the data perhaps seems limited, but it looks interesting. I mean, do you feel then pretty, pretty certain that that curve is going to continue? Are you are you firm on that prediction?
Joe Fitzsimmons 38:37
So what I would say is, if you look back to the start of, you know, first experimental demonstrations, and you plot the data points for the largest number of physical qubits in the system, I'm only talking about systems where you can perform all of the gates, if you plot it on a log log scale, from 95, or 96, whenever the first one is up to around 2015. At 30 is a straight line. On that log it sorry, on that semi log plot. So there's exponential growth, right? What happens? The problem is the doubling time is every five years compared to Moore's Law of about every two years to 18 months. Right? If you plot it now, from kind of 2015 2016 to the present, there is also a straight line. But now there's a kink at that point, so that the slope of the curve has changed. But it's two straight lines with a slight interpolation between and that is going around, you know, doubling times more like once a year. And that's faster than Moore's law. Wow. Faster Yeah. So, you know, it means that we could get to large scale systems significantly quicker than we might otherwise expect. But at the same time, this doesn't really track the error rates, because as you build larger systems, you tend to, I was gonna say, a bit of a dip in performance before they come to the same level.
Sebastian Hassinger 40:27
Yeah. In the vendors where you've had multiple generations of every certain chip design, it's really interesting seeing like T one and T two rates and gave Fidelity's with first gen to second gen to third gen. It's I mean, it's, that's when it, it feels more like engineering than science is like, Okay, we got the science to build the thing the first time. And now, they just had to turn the engineering cranks to get better at fab and get better at, you know, preparing the material and all of the things that we know from, you know, that's where sort of the broader scope of technology can bring brought to bear. Right. I mean, fabbing is something we've been doing for a long time now with semiconducting. So there's lots of techniques anyway, yeah.
Joe Fitzsimmons 41:13
This one of the things that I think can be misleading to people that you can think, well, if I can't do it in my lab, it can't be done. And I don't think it's necessarily seen. Well, what if I tried $27 billion stuff the problem? TSMC does, right. Like that's how they get to modern processors. So, yeah, I mean, it's a it's a different, it's definitely a different way of thinking. So at least from my perspective, I think there's, there's going to be more of a shift in that direction as we get to larger scale systems. But you're right, like, it's not surprising that the bleeding edge in terms of qubit number, that you see a performance drop, and then as there is time for those to mature, you see the performance increase slowly, right. So our largest systems will always be our least performance. In terms of fidelities, I'm sure that'll be true for, you know, essentially forever.
Kevin Rowney 42:23
I mean, I guess, another dimension of performance improvement. I'm not sure if you agree, but I've been several people that pointing at the last 12 months have pointed towards really significant advances in quantum error correction. Like, some people even think it's been one of the most notable past 12 months in quantum error correction. In a very long time. I mean, this, perhaps could set up a further acceleration or further arcing up of that of that curve. What do you think?
Joe Fitzsimmons 42:51
Yeah, so I mean, I guess what I would say is that there's a huge amount of work to be done on quantum photons, we started to see a lot of physical demonstrations. Lots of efforts are getting to the point where they're able to demonstrate full truncates. That's where you're starting to see suppression of, of noise, and so on. So we're getting closer to that kind of demonstration of sustained, sustained error correction, sustained, full tolerance operation,
Kevin Rowney 43:22
we're not reaching that threshold, as mentioned in the threshold theorem. Yeah. So
Joe Fitzsimmons 43:26
I mean, you don't want to do it just a single shot, you want to be able to repeat it, right and suppress it further and further and do your operations in that in that setting. And, you know, there can be not all systems are capable of this, because many systems don't allow mid circuit measurements. Or the you know, what is necessary to make a mid circuit measurement is highly disruptive to the rest of the system. So you end up introducing extra errors. But the reality is, we're clearly getting closer to that. We're getting closer and closer. At the same time, from the theory, you're seeing improvements as well. So you're seeing that, I mean, there's, you know, even really, quite simple things. For example, if you take a surface code, and you pick your stabilizers, slightly differently. So it's still the same surface code, it's just the local basis is rotated. You can take advantage of bias and noise. So frequently, the noise affecting qubits affects the more in one axis than another. And, you know, this, this one simple change can lead to you being able to exploit bias to get much better performance. So yeah, there's quite a lot happening. For a long time. The focus had been on proving that full tolerance was even possible. And you know, under what assumptions that was possible. And it's only, you know, it's quite recent that the focus has been on concrete numbers on like, how many physical qubits do I need to do X, and so on. I think, as we have moved towards that we've seen we've seen more changes, like more changes in our understanding of fault tolerance. As the goals have shifted, they've kind of pointed us in new directions. And we started to see new. Yeah, we've started to see progress, for sure.
Kevin Rowney 45:36
got started by Sebastian.
Sebastian Hassinger 45:40
Oh, it's gonna say, do you think part of that the skepticism from the quantum community, you've already called out the sort of perception of risk to the community by taking money and then failing? And then do you think that and then you described, you know, not understanding what industrial scale investment can do to get you over the threshold? Actually, if you get, you know, some limited success, that it's attractive enough to capital capital will mobilize and make it into a large success? But then that what you just described there about the landscape of fault tolerance? Do you think there's an effective almost siloing of people seeing, you know, let's say gate Fidelity's, and not tracking the, you know, the evolution of thinking around surface code, or only looking at surface code and not understanding the game fidelity or? I mean, that's a simplistic,
Joe Fitzsimmons 46:31
I don't think there's I don't think there's any problem and how people think about this. I think we've just had different goals over time. I mean, actually, if we want to get to a fault tolerant quantum computer, like to an essentially noiseless quantum computer, there is in principle, a path. The problem is it well, problems that have occurred over time for you know, why we are not further ahead on this problem, for example, is that, you know, like, one example, is just noise models, right? So, you know, you may be proved full tolerance is possible within some set of assumptions. But that set of some options does not mirror how real devices work. And we didn't necessarily know that that wouldn't be an accurate description of the real devices, or wouldn't be a reasonable idealization. And we see it now more. So in general, we know that correlated noise is bad, or non Markovian. Noise is bad, because we haven't proved the theorems in that regime. But actually, non Markovian noise you can often deal with by kind of spin echo techniques and so on. It's like, it's an it's a memory effect, for the most memory. And you can often use that memory effect to cancel out the noise that has happened. You know, so these things which are bad in terms of the theorems, because you've entered a regime where the theorem, assumptions no longer hold. They're not actually necessarily bad. And figuring out how to take advantage of some of these things, is important. So even brass noise. I mean, if I remember correctly, the first time I really came across that being examined properly, was I think I saw a talk at QVC in 2007. When I think it was Bilasa, for us, and John Prescott had a paper that was that was showing that you could get much higher thresholds if the noise was biased. But, you know, getting to that point, you know, not having, you know, like, you can prove different things if you have to assume it's like adversarial noise, or if you have to, or if you're assuming it's like stochastic poly noise or like, whatever your assumption on the noise is, they lead you to different conclusions. But actually what we care about is the real devices. And as we start to see more of the constraints and more of what's actually happening in those devices, it will lead us in new directions. Yeah, for sure, because we'll be getting closer to what's actually possible
Sebastian Hassinger 50:00
Well, we were having such a great time talking with Joe. And he's got a ton more to say. And we decided to split this episode into two parts. So it was really terrific. For me to hear him sort of expand on that Twitter thread explaining his reasons for being really optimistic about where the industry and the technology is today and where it's going. And I thought in particular is really interesting. Just sort of pointing out how academia may not be fully grasping the what you can accomplish when you really can mobilize mountains of capital to get something done. Yeah.
Kevin Rowney 50:41
Yeah. I mean, yeah, we try this podcast, I hope what's coming off it to send us sober, balanced, right, between, you know, the optimism, but also being resistant towards the marketing hype. But I think he's got a valid point there. I mean, look at look at the huge transformation that happened in the 20th century around, you know, space and orbital technologies. I mean, there was this lead up for like, over 100 years of like, huge theoretical breakthroughs and experimental results. I mean, the likes of Robert Goddard, and all this amazing rocketry did back then, rather obscure at the time. But still, I mean, it wasn't really a lot of effort. But you know, good, powerful contributions are also the mathematics of have to have skis equation for, you know, orbital calculations, and how much fuel you have to have to get up on orbit. I mean, really great stuff. But it was just early phases of research. And then nation state competition and huge amounts of nation grade resources came pouring in armies of people starting to work on this, and it was a complete phase change. Right, right. In the in the race to the moon. So I think he's got he's got, I think, a valid point here.
Sebastian Hassinger 51:45
Yeah, I agree. And I thought that, you know, that's a great example of the the race to the moon. And I thought his example, his current example of sort of like, the capital it takes to build a two nanometer microprocessor fab is something on the order of $27 billion. Which teams TSMC is currently investing in building fiber, that it's justified because the size of the the opportunity for the advanced technology, crazy chance right there. Yeah, many times more than has been invested in quantum computing as a whole.
Kevin Rowney 52:22
today to talk about
Sebastian Hassinger 52:25
it. So yeah. So join us. If you enjoyed this part, join us for the second part, we'll be releasing in two weeks. And Joe is going to start digging into the technology that he and his team are building in horizon, which was also extremely interesting bread. So really, I found it really eye opening in terms of the potential quantum advantage being much broader than most people tend to view it in the current context. So yeah,
Kevin Rowney 52:53
a real optimist. He is this really? Absolutely. Okay. That's it for this episode of The New quantum era, a podcast by Sebastian Hassinger, and Kevin Roni, our cool theme music was composed and played by Omar Costa Homido. production work is done by our wonderful team over at pod phi. If you're at all like us, and enjoy this rich, deep an interesting topic, please subscribe to our podcast on whichever platform you may stream from. And even consider if you'd like what you've heard today, reviewing us on iTunes, and or mentioning us on your preferred social media platforms. We're just trying to get the word out on this fascinating topic and would really appreciate your help spreading the word and building community. Thank you so much for your time.
Transcribed by https://otter.ai