Kevin Rowney 0:07
All right, welcome to the second episode of the new quantum era. Hey, my name is Kevin Maroney. I'm
Sebastian Hassinger 0:35
here with Sebastian Hassinger.
Kevin Rowney 0:38
And so yeah, we're the we're the CO hosts of the podcast, we, for some time now have been watching this fascinating area of quantum computing. We think it's an arena that's got, you know, many fascinating aspects of it about it in terms of its enormous potential economically in terms of that possible impact on future computing architectures. But also, we see there's numerous significant risks in terms of unsolved, massive unsolved engineering challenges left to be done before it can actually be a practical thing. So it feels a high potential and high risk endeavor with with Mitch, much potential and many ways, a very compelling topic.
Sebastian Hassinger 1:18
That's right. Yeah. Kevin, it's, it's, I find it fascinating. For the reasons you said. And it's really, it's difficult to navigate compared to prior sort of emerging technology areas like mobile, or the internet, or the web, or even AI and ML, because it's not an incremental advancement on the current paradigm of computing. It's it represents in quantum computing an entirely novel way to compute. And so there's, there's a significant learning curve, as you said, that involves understanding more about quantum physics than I ever thought I would. And then mapping that to slightly more familiar terrain of technology development and market technology market development, which you and I are much more familiar with, but at a very, very much earlier stage. And in fact, our interview today with Cisar Rodriguez from strange works, is he'll touch on sort of the analogy to the development of the classical computing world, at the stage of vacuum tubes and the invention of the transistor. So I think that's a really good calibration for people who are trying to make sense of the topic to realize that we're not at even the PC era. In quantum we are still looking to the transistor.
Kevin Rowney 2:41
And I think maybe many people even forget what it must have been like to be alive right during the onset, the dawn right of the IT era before became standardized and rising fast through Moore's law. And so back then I bet you, you had that same mix of both optimism and reason, skepticism, and that's, that's really the reason we really enjoyed this interview with Cesar Rodriguez is he's a great example of somebody who is knowledgeable about the space and sees his possibility. But also he's got, you know, a decent level of guarded optimism and even skepticism on some of these results, which, in many ways, they run fits and starts and sometimes even go backwards. There's a lot of of chaos in the in the messaging and the progress.
Sebastian Hassinger 3:26
Yep, absolutely. Right. Well, let's get to the interview. Let's dive in.
Unknown Speaker 3:55
Sebastian Hassinger 4:05
all right. Cool. So today we're gonna talk to CES are who we've known for a little while. But yeah, Sir, do you want to start off by introducing yourself?
Yes. Hi. I'm, Emma says, everybody gets Rosario. I'm a quantum physicist, and I'm the Chief Science Officer trying to work.
Sebastian Hassinger 4:22
Excellent. And so I thought maybe just start off, you had a really interesting path to get to quantum computing. So can you can you fill us in on how what was the circuitous route that you took to get here?
Yeah. I'm reading from Puerto Rico. I always like computing in general, the history of computers, very passionate about it, and so on. And I ended up studying computer engineering then. And, and, and while I was doing that, I got very interested in quantum computing. This was very rare and it was 20 years ago. And because I was interested in the history of computing, and I love the early days, I was like, oh, that's another early days. That's going to happen. I want to I want to do that. So I had the choice of staying in Puerto Rico programming, human resources, software, and surfing or going to learn how to put them up on a computer. So that's what I did.
Sebastian Hassinger 5:13
Do you ever miss surfing? Yeah, exactly.
I haven't done surfing like many years. But I liked it. I liked it. Yeah. Right now I'm in Germany, so I cannot serve it. There's no,
Sebastian Hassinger 5:23
there's very little German serve. So you got your PhD in quantum physics specifically, or quantum refresh?
Yeah. So So I went to I went to the University of Texas in Austin. And, and when I went there, because I was not a physicist, I did not have that background. And it was actually very difficult with the GRE ease and all that. But when I was an undergraduate student, because I was so interested in reading physics in general, I did a lot of research jobs in the summers and such. So I think that really helped me with like to land like to get into grad school. And University of Texas is a huge department. It's like it's, it's brutally huge. And so often, students feel very disoriented in grad school, because you don't know who you're going to work with, and so on. And so I had to do the thing of knocking on doors talking to professor's I was very open minded about what I was going to do, I thought I was probably going to be a experimentalist because my, my background as an undergrad was experimental physics, mostly because my engineers because I could fix stuff basically, that was really, like, you know, as an undergrad, the main thing you do is like, you clean stuff and like put it back together. But when I was talking to professor's, I ended up talking to George Sudarshan, when there will be my PhD about Beiser. George passed three, three years ago. He was a very senior senior scientist, I think it was nominated for the Nobel Prize eight times. And, and he, he knew firemen, he knew government. He knew like all this other strands of physics, he was one of the giants too. And, and he did really important work in quantum optics, in quantum information, and quantum decoherence. And a lot of quantum stuff. So at this time, he was getting interesting in quantum computing were more more focused at this stage. So I talked to him. I really didn't think of myself as theorists. But he was very funny. And we bonded talking about plantains and bananas, because he's from Indiana. So I kept going into the group meetings. And then finally, I
Sebastian Hassinger 7:30
got a PhD just by showing him for the meetings.
Yeah, that's what I supposed to do. That is the biggest mistake. Just like it just like if you want to come again next week, and I just kept going like, I was never like, officially admitted. They just get bored. Really? Yeah. Well, there was no official procedure. It was just like, Yeah, sure. Let's do it.
Kevin Rowney 7:47
You just keep showing up. Yeah, that's cool. That's amazing. 90% of life right there. Yeah.
Sebastian Hassinger 7:54
So that that's sort of how you found yourself more on the theory side, the experimentalist side.
Yeah, it was very hard for me because I, my, in my engineering school, the math was very limited. And so I had to, I had to learn a lot of stuff. In fact, I think I learned quantum like completely backwards, like from everybody else. Because I usually at University of Texas, people have two years of quantum before they go to the ground level Quantum. And they have about a year of classical mechanics, which is usually a prerequisite. But I went to University of Texas, and I had not and then I didn't even take classical mechanics before to quantum because of the way they were offer ended up taking quantum first. So So I am quantum native in the sense that I see the world as quantum like, really, then then when I look classical, I'm like, oh, it's like that. But then you change back.
Kevin Rowney 8:45
rather impressive. You climb that steep hill, you know? You have that help? Yeah, my God.
Yeah, it was really brutal. I really struggled. I didn't do very well, at first. Mostly the books that made no sense to me. And then I discovered jumbo Norman's quantum mechanics, by this is this a really old boy? He's from the 50s.
Kevin Rowney 9:07
Oh, yeah, it's open sources out there. But the Celtic website, great, great resource. Absolutely. Yeah,
yeah, it's, but it's, and then I read that, and it's actually focused a lot on the density matrix, and not sort of wavefunction etc. And then it made sense to me. I'm like, Oh, it's just probabilities. So I have learned quantum the wrong way. And I think everyone should learn.
Sebastian Hassinger 9:33
It is funny how, you know, there is kind of this, this on ramp that's kind of been defined over the years. But it also it I think, in many ways, generates those multiple stages of going like this is totally non intuitive. This doesn't make any sense because it's like, it's like challenging what you learned in the previous stage at each stage. It's funny,
and this is because a lot of books they're very historical driven. Like they show the historical way. But at some point, somebody has to be brave enough to say, Okay, we've had 100 years of quantum mechanics, it's okay if we learn it
Sebastian Hassinger 10:07
the way we like it, just assume it.
And then you like you, like scratch everything and start from information theory, which is which I think should be done.
Kevin Rowney 10:15
I agree. I mean, because there's such chaos and uncertainty in the minds of the early scholars that they've struggled with mightily. So trying to recapitulate that it just feels like it would definitely induce like tough pedagogical outcomes, right?
For me, for me, like very simple things like I usually, like in quantum, you first learn the wave function. And you do everything with a wave function, and so on. And then like you some people don't even learn the density matrix. And there's a big mystery of how do you get probabilities out of wave functions, it's such a weird thing is the only thing that kind of ruins the theory. But if you start with the density matrix, it's all like, it's the same as classical probability theory. It's really It's beautiful. So so so that's what I think it should be. Like, it just makes more sense. Information Theory. Yeah.
Kevin Rowney 10:57
And I recall vividly you on the Slack channel for our reading group are urging all of our some of the students to focus on that density matrix. That's fundamental. So appreciate the coaching.
If you get used to doing things that way,
Kevin Rowney 11:13
everything goes easier, right?
And then also, you're like, oh, it's probably just provenance.
Sebastian Hassinger 11:21
So do you think coming from a computer engineering is because software engineering perspective, do you think that and and also having this this sort of unconventional journey through quantum physics, do you think that sort of gives you a different perspective on on the field today? And like, what, what might be interesting or productive? You know, sort of PAs research paths to go down?
Yeah, so So I can say that in like, gives me like, there's two angles? When I I sense, I think a little bit differently. One is this historical perspective on the development of, of classical computing, classical view, like for me, it's often the question everybody asked is, Can Can I can I get some bitcoins tomorrow, using a quantum computer, this is like, last week. And of course, this is very disconnected from where we are historically, which is a very, very early stages of one year. So and there's nothing wrong with that. So I always throw the parallels like this is where in the back into era. Like that's, that's what we that's what we are, we're in the back with I don't have the peering, it's vacuum tubes. Why is it that so when people were building the first I can go back into its computers, they may have niches that they had, where that the back end does just blow out. And you have to learn, which was more than the others that tried to work around that.
Sebastian Hassinger 12:42
Error Correction and error mitigation.
Another issue they had was rats eating the cables, yeah, rats eating cables, and, and like, when you read the list, all these things that you realize, like, Oh, this is like really down to earth, like real engineering problems, and they're really struggling. And that's kind of where we are now. And so, so. So for me, which one is the best quantum computer right now? And he makes no sense. Because if you were in the back into where they all kind of sucked in. And he's nothing bad. He's just, you know, it's just,
Kevin Rowney 13:12
it's just your to be fair,
Sebastian Hassinger 13:13
they we haven't had the we haven't had the Bell Labs transistor moment. Yeah, right. Yeah.
Yeah. So so so of course, like, that's what we're waiting. It's like when when do we get into a sister. And when I when I call, what the way to think about the quantum position will be, of course, this is a requirement, but in quantum computer, well will be is that you have you can you can use some qubits that they're very, that the quality is really, really good. You can connect them very, very efficiently. And you can connect them as many as you want, in a way that this scales, like we have to do all those things. And nobody has cracked the code for all this bullet point.
Sebastian Hassinger 13:49
No, no. And in fact, in classical terms, what you're describing is both the transistor from Bell Labs, but also the monolithic solution from TI and Fairchild, right, because it until we figured out how to print circuits, like with transistors on boards, we were trying to hand solder components together. And you could only get to a couple of 1000 components before there was definitely going to be a crack solder joints. Yeah, exactly.
Exactly. Exactly. So that's really when once once you reach that stage, then is when you can start making real predictions of this is the rate of growth, then the computers are going to have but until you reach that, everybody tries to guess they're just making up stuff.
Sebastian Hassinger 14:30
Right? Yeah, Moore's law did not kick in until until the monolithic solution was really sorted. And we had microprocessors. So yeah. But
Kevin Rowney 14:39
it's so refreshing to hear a voice that is able to sort of both see the possibility but also is turning you know, a skeptical eye right on me as some of the vendor claims because it does feel like there's just a lot of spin in this space and that the press releases are so shallow compared to the I don't know the complexity and depth of the real story. It's very difficult space in which to like discern where is the Real progress. Yeah.
I always say I don't read press releases. And I don't because I get
Kevin Rowney 15:08
nothing from them. Yeah, yeah, it's written by somebody who doesn't even understand. And and I'm
always so poor, like people laugh. I'm like, I'm like, What do you say?
Sebastian Hassinger 15:17
So given given that sort of stage of the technology, what do you think the role is for foreign theorist? Right? Because in some ways, you know, Quantum Information Sciences sort of, in some ways won't fully be realized until we have that that world where you can connect as many qubits as you want to in a scalable sort of, you know, error resistant way. So, you know, the role of theory now, do you to think that there's sort of almost like a co design between experiment and theory to try to figure out how to get there?
Yeah, that's an excellent question. So I was saying like, right now, if you want to program quantum computers, which are, you know, very noisy, the connectivity is really bad. The gates are not great. You do have to know the physics very well, of what's going when the computer like and so for that, you have to be a theoretical physicist, you have to really, really hear it. And you have to be not a fair. And this is similar at the beginning, when you were programming, classical computers, you really, you have to go with the oscilloscope and stuff, you really have to understand the test. So that's kind of where we are. Ideally, what's going to happen is that, that once we have the scalable arrow, correct, corrected qubits and all that, then you don't have to be a theorist anymore. And so my so my, my, my theories had is I'm not going to put it again, I'm going to be a full time quantum engineer, it's, and that will be healthy. Like I want that to happen. That means that that the the industry succeeded. When that happens,
Sebastian Hassinger 16:49
then there, there can be quantum script kiddies, when I'm writing my equations, we have not seen it.
Kevin Rowney 16:55
And so how do you how do you view then the short to medium term kind of like inniscarra landscape because there's a lot of people point here new algorithms that might perhaps be enough of a level of resilience with respect to the current limits on these these machines to still do something useful. But yeah, there's I still feels like there's a lot of even Niska algorithms, which were published up on archive and have been de quantized. I mean, there's, there's a lot of uncertainty, but how do you how do you view the reliability and the breakthrough potential of the current crop of algorithms you have given given current limits,
then I will make it with the early days of aviation like brothers, right? So so if you ask me, if he is American, so who made the first player raises the brothers, right? If you have proceedings, by the way, they have another guy, of course. And the reason is that it depends how you how you define the thing. Like brothers words cannot use a catapult to launch the plane. And you can think that's cheating. So if you think that sitting there, we're now the first it really depends, like, when do you make it? Oh, this is this is like, legitimately, like on its own standing. And it was it was like, the it took like, more than 30 years before you had the first flight with passengers and luggage.
Sebastian Hassinger 18:13
And then another 40 years, we're on the moon.
Yes, exactly. So, so so we are right now the near Scarah, which is the noisy intermediate scale, compute computing, quantum computers. So this one is, it's really where the early stages we are brothers right? And and what we are trying to do is everybody wants to claim we were the first ones to do something that has never done before. And this is known as quantum advantage. And to be more to define your life, concretely, quantum advantage means that you have a quantum computer that can solve a problem that no classical computer right now could solve. That's it. That's that's really, and so so. So that that sounds kind of us right. Like suddenly, this is like a real superpower. Right? However, so there's many ways to go about it. And there's been some claims like Google had a big announcement a couple years ago that they achieved this quantum supremacy
Kevin Rowney 19:14
a rather controversial announcement. It feels Yeah,
no, well, I don't think it was, it was so controversial, but But it shows how ill defined this question is is that I will say so so so what they did was they did show an algorithm that could outperform any any high performance computer we have in the planet to do a task that nobody could do before. Like I think it took them like like a few minutes and it will take a computer they estimated like 10,000 years so that's you know, that's that sounds very convincing. Over the task was completely toy task. That was really like you define it as for the computer that you have, and it's not the other way around like you you make you don't make a computer to solve the problem. You make up a problem that fits a computer. But that's how you start like this is not like take it down. This is err, absolutely, that's how you have to start there. And of course, what happened after was that a lot of researchers found ways to improve classical algorithms. And and for me, like, that's, that's legit. That's good. It benefits everybody. Right?
Sebastian Hassinger 20:15
Yeah, in many ways. I think the distinction, you know, classical information science and quantum information science is an arbitrary one, right? It's like, there's just information. sighs You just have different tools?
That is all it is. Yeah. The world is made up of, of information. So it's right. That way, the distinction I make is that quantum information, quantum computing uses all the information in the world. And classical uses, I just eliminate it. It's just,
Kevin Rowney 20:44
it's just a great way to sort of see with more clarity the the current landscape as you comparing it to the, the the early experiments in flight. I love that analogy. But I mean, if if you were trying to sort of see from the landscape of the current algorithms out there, I mean, do you feel like there's an imminent breakthrough in this scare algorithms? That shows that supremacy that has commercial relevance? I mean, I guess that's a lot of our audience. They wonder about this all the time. There's so much like emphasis on heavy theory, it for a lot of people who are more practical or you know, maybe want to make a bet of VC watch one of these plays, it's hard to really feel like it's not just a big science experiment.
Yeah. So so so far, it is that makes sense. Right? Right. That's gonna do that's okay. Brothers, right. They had no space for a long time. So that's, that's just like we do because we can that was a point. And I think that's fine. I think it just shows like where we are in history. And for me, that's why it's exciting. Like, I don't see this as a bad thing, though. It's,
Kevin Rowney 21:42
it's amazing stuff really beautiful. It's just Yeah, it's also there's the practicality of like, so where is where is the outcome that would really make a difference in the world?
So one way to think about it is that what, if you want to get into quantum computing right now? You don't do it? Because you want to get the bat, you don't want to extract the value from it, you do it? Because you want
Kevin Rowney 22:02
to do right, because yeah, you're passionate about the subject. Absolutely. And for
me, that's, that's very exciting. Of course, other people, which are, why don't you like Hurry up? I'm like, Hey, I'm been hearing. But, um, another thing to take in consideration is that, for example, that Google and Amazon was one, but there have been others after that by some Chinese groups and other research groups. And they and the same thing happens, like they make a breakthrough, it seems actually supremacy, then somebody catches up in some other way. And I think that's for me, that's extremely exciting that that's, that's the engine of progress, is that it's gonna keep happening. And it's gonna be a little back and forth and so on. And this is because supremacy, like we like to think like, oh, you're rigid, then you're done? Like, no, no, no, this is like a very loose definition. And everybody is playing games to like, barely make it. And of course, you realize it didn't make it, or sometimes they made it, but it's not so exciting.
Kevin Rowney 22:55
So you're pretty confident at this point, there's going to be more G quantized. Algorithm sounds like,
now I'm sure I'm sure. But this idea of the quantum center algorithm, so it's sometimes it's very hard to know, what makes the quantum algorithm tick,
Kevin Rowney 23:07
or even even classical algorithms, right? I
mean, and also, sometimes people are like, Oh, why don't you just make up a new quantum algorithm? And my response is always like, why don't you make up a classical? See how, right XOR quicksort? Come on. You just look it up. And there's a reason for that. And it's because it's very difficult to do anyway. And quantum of course, it's very counterintuitive. A lot of this, the classical algorithms, you're there, you already have really good computer. So you did the my trial and error with quantum, it's a little harder. And there's nothing wrong with that. It's just just shows where we are in the industry, and where things are headed. And, and I will say like, it's okay, that things are not useful, yet. There's nothing wrong with that, because we're still working towards that. And the way I see this progressing is like this supremacy claims are gonna keep happening, that you guys are gonna get bored from the press release. And you're like, Oh, but I hear that
Kevin Rowney 24:04
a level of fatigue and cynicism might sit in some minds. Yes,
yeah. But it's gonna be very nuanced. And what's going to happen is that they're going to keep having more kinds of supremacists, until maybe at some point one barely scratches the surface of it being useful for somebody, because you're not useful. Yeah. And so and that's going to be a big deal. And so I that time, where a quantum algorithm becomes useful, like, I don't because this advantage supremacy these words are really overloaded. I like to call it quantum value. And for me, what quantum value means is like, when I when you can convince someone to use a quantum computer because it saves them time, saves them time or money. Like that's, that's it, that means sensor doesn't have to leave the call to explain to people at technologies like look, it's gonna save you this much money like this is very, very concrete, right? But we're not there yet. And it's gonna be a while but that's fine. And hopefully, in the next year Aren't we find something that is correct, it's close enough that is valuable even if it's for a tiny problem, that's an application for very few people that will be there'll be a success.
Sebastian Hassinger 25:09
If you if you look at the theory landscape right now, what are the things that are the most exciting to you?
Kevin Rowney 25:14
Yeah, we'd love to hear your your if you're a betting man, I mean, which of the different categories of current news or algorithms aren't likely to sort of show that kind of
breakthrough. So So of course, I the whole world of, of variational quantum algorithms. This was this was work. It was started by a group of analysts progressive, I used to be a postdoc in his group when I was at Harvard. And the same week, this paper came out. Eddie fire, he also came up and difficulty and came up with another one to say the same way. Like those groups, I know for a fact they they talk regularly, because we used to take, we used to take the team Boston and go to meet them, and they will come up here with us and so on. So I'm sure this was like, it was something like, you know, let's, let's put them up at the same time. They're different. But and the idea of variational quantum algorithms is, is the idea that quantum computer might be very good at quantum computers are very good, of course, at solving quantum problems, right? So if you give it a quantum problem, can can you like, have it come up with one answer? Maybe it's maybe it's just like, if you give it a quantum gas, let's say, and he tries to tell you, Well, this can give you this answer, then you take this input this output, and then use it, how can I update my gas using classical computer, so I'm alone, let me change the parameters of it to jiggle things, maybe again, I guess, and you keep iterating this way. And all these variations are going on. And this is sort of the spirit behind behind them. They're very, very intertwined, the classical and the quantum part. And this is just really exploiting the, like the fact that quantum computers are very good at Quantum simulation and linear algebra, it's just, you just doing that and then uses the classical computer, of course, to steer it the right way to make it faster. The benefits of this algorithm is that because you're iterating a lot, you don't have to have a very long cigarette like lime in terms of death, which you have a lot of gates one after the other, these tend to fail more often because the noises add up. So this one, like the clever thing is you break it up into small chunks, and you you get an output before everything falls, like falls apart? And then okay, they live up, let me update my guess. And you keep extracting amount of little information, so you steer what is likely answered the door. And this is a huge world right now. Like this is basically what everybody's doing. It's impossible to keep up with that there's experts at how to do this guesses is that they answered like how to handle the updating, what are the different techniques
Kevin Rowney 27:47
and also the the architecture of the quantum hybrid kind of setup in your what kind of algorithms might exist on the classical side to help evolve the the guests, there's a ton of research there, my God.
Like, like, like, I it's impossible to keep up with all of them. But a lot of this, a lot of these, like take the data, when it's a clever way to update your guess these are classical algorithms are doing the guessing. Right. And this is, this is like a gigantic world. This is primary existing classical computer. So so they intertwine between the assumptions you have to make. You know how to cook the recipe just right. So you get like, barely make it work. I think there's a lot of noise, a lot of like details that matter for that. But it seems to be very fruitful so far. So and of course, this it was designed for this kind of computers that are that are that don't scale. So well, this it was designed for this. So I will say that's the most thriving area. And it's very exciting. And I don't think we're gonna get rid of them, to be honest, like the variational quantum algorithm, they're here to stay.
Sebastian Hassinger 28:50
I mean, it looks it looks like like a coprocessor. Right, which makes a lot of sense until until you get to that that you know that that really scalable architecture where you can get, you know, to the point where you've got something that can be a general purpose computer, in the variational stuff. It does everything still fit into the sort of general buckets of V QE and QA Oh, a, are there other categories that are emerging that are distinct enough from those two approaches that they they have their own acronym yet?
And so QA for example, is a cannon will be key? Yeah, okay. Okay. He was it was shown that maybe at first it was unclear, it seemed more limited at se Khan is basically you take variational quantum eigen solver, that's what it stands for. And all of them kind of look some one way or another like a variational quantum eigen solver. And the reason is because, because the word states like so it's very so non variational means you jiggle the guests until you get better, right? And then the quantum is I use a quantum computer and the eigen solver is to find an Eigen value.
Sebastian Hassinger 29:54
That is kind of the most generic description isn't it?
Yeah, but that but yeah, so then okay, so you you come up with some gas or something and you find an eigenvalue, then you check, is this eigenvalue as low as I want it? No. Okay, let me do it a little bit more in a smarter way
Kevin Rowney 30:10
to any questions on quantum state. I've seen
Sebastian Hassinger 30:13
sort of the the, whatever the claim that actually things that are being called Quantum machine learning are actually also under the category of V QE as well, right? I mean, there are those are also variational. Approach. Yeah.
So some, some are, some are. And, and so and a lot of the machine learning things depend on algorithms that are just really good at linear algebra. Some have been the quantize, which is something that we mentioned before. So by the quantize, the quantizing, what does it mean? It means that the algorithm was invented its environment, so on and somebody realized, you don't need entanglement. That's what I mean. So So going back to what we said before, is, if quantum information is using all the information on the world, then somebody came up with an algorithm you just saw, and then somebody figures out like, do you really need it? Oh, you just need the old stuff. But this is that is very nuanced. Yeah, because it's only been shown for us some classes of problems only. So it could be, you still need all the one for the bigger ones, and so on. And I think this is gonna be it's gonna be we're gonna give exploring what these are. And it's gonna have an impact on quantum machine learning. Like, it could be that we don't know, we don't know if if quantum machine learning is better than classical. And, and this is because the way we usually think about algorithms is using complexity theory. And complexity theory is, if you have ever met one, they're the most unreasonable people on the planet. Because that's their job is to be extremely unreasonable. So they come up with the problem, you come up with a problem, and they're gonna come up with the most unreasonable expectations of like, what about this really case? How fast can you do it, and that's the job, they had to come up with this insane scenarios. And you have to show that this works for this instance, and nobody in the right mind would ever try. And but that's how you can be really darn sure that this algorithm is better, always right, by machine learning doesn't work that way. Because machine learning is you're interested, that means like, you kind of like try it. And if it works,
Sebastian Hassinger 32:10
like that's good enough.
And so comparing classical algorithms to quantum machine learning algorithms, you already you already don't have the tools of complexity, and you cannot do that. And so it's all heuristics. And so the question is, will always, will always be there. And I will say, if I were, if I was a betting person, I will assume that if you were doing machine learning for quantum problems, you can guess that the quantum computer is gonna do better, right? That's a really safe bet. What is the area in between? No, I'm not, I'm not placing on
Sebastian Hassinger 32:43
it. It's interesting, because I like that framing, you know, that you use of, of Quantum Information Sciences uses all the information because it really does underline the fact that, you know, in a classical simulation, HPC simulation, let's say, of, you know, a chemical reaction or a seismic, you know, simulation, that's those are materials simulation, there, you know, the, there are a number of choices that the researchers or the experimentalists are making to downsample, essentially, to throw out a bunch of information, because they can't use all of the they can't actually, they don't have the capacity to use all the information there calculation. And I guess what's interesting is that, that implies that quantum computers will allow us to test those assumptions about what information is actually, you know, okay to throw out and what's not. Okay, so it's super interesting.
Yeah. In fact, in fact, like one way to see why quantum computers are fundamentally better at some problems, and classical is exactly what you said, which is, how do we solve a quantum mechanical problem right now, most of the scientific computer and is used for doing calculations of the sky, because they're essential for material science, for medicines. For all those you have to fit, you have to find out any of the molecules is a quantum problem. And it's really, really hard to do, like even tiny, tiny molecules, we cannot have the biggest supercomputers cannot solve them. So this is why when you're coming up with new materials, except for solar panels, or new medicines, or something, you end up with a lot of people with lab coats, mixing liquids, and and in the lab, and this is try this very smart try on there. But but it's because it's super expensive, because you need a lot of people with PhDs with a lot of like, pets by pets and all this stuff, mixing it all up. So and this is because we don't know if we could up if we could a priori know, these are like good candidate molecules, then you can just aim for those and it will save a lot of time. And but we cannot do that. And it's because you start with a fundamental quantum mechanical problem that is too hard to solve. So you do approximations just like you said. And then you take this really approximated thing which already like you know, you threw out a lot of good stuff. Then you put in a classical computer, which is not suited for it, and then you run it, but you have to stop the computation times and turn, you can run it until the end of the universe, so you have to stop. So you already have more errors, because you stop at some point. And then you take all these errors, and then you have to put it back into the quantum problem. And so you've already lost so much in this process. And this really shows like, what are the limitations of, of computing, as we as we, as we know, in now for for this kind of problem. And this, this was the original motivation for quantum computers was like, instead of doing all this, can we do better. And however, like, of course, there are some problems, like we like one of the bidders are not magical, they're not good at everything. So there will still be problems that they cannot solve. And we still approximate them, of course, maybe they're better in approximations, we'll figure it out. But there are some, and they bet they can handle bigger datasets, not right now. But in the future, there is another there is a tiny bit, they will be able to handle better, because the quantum information is exponentially larger than classical information. So so you can fit more in them a lot more. But it's gonna be, it's gonna be a while I think before before we can exploit that kind of comparison, not so good. But it's actually give you a sense of like, okay, there will be more problems that this approximations were bad that they will, we don't have to worry about them. And for sure, the obvious ones are everything that has some quantum in it, it's gonna it's gonna transform things like that. If we if you want to think like very future wise, like, when quantum computers I scalable, pharma, pharma and chemistry companies, they're gonna look barrier free, the people that work there is gonna be less people will have coats
Kevin Rowney 36:40
me for the right for this molecular geometry problems. Me It's just there's such huge value there. But I mean, you could tell me from a lot of the people that are actually practitioners on current molecular geometry, struggles, they they're not, they're not embracing current contemporary quantum solutions. Right now, they're, they're still doing the pragmatic, classical thing with its with its many limits, it still feels like that that's one area where there could be a big breakthrough, as far as we, as far as we could, as we could tell, many of the big software startups in quantum algorithms have done a fairly significant pivot many of them in that very direction of towards the molecular geometry use case. So maybe that's the first big break breakthrough outcome in this space, who knows.
And to define the molecular geometry problem. What that means is, if you imagine like that, right now, what we do is like, like those toes, you have like little balls and sticks, and you put them together and you build something, that's what we think of particles, right? Now, that's a classical way to think about molecules, right? And that most of the techniques are that with some corrections. But in the end, they take these things, they jiggle them, and now look at the molecule he moves when you shake it this way, therefore, he has his properties. But this is wrong. Because I was on a boss there, there were functions that are very fuzzy, the connections that connect interconnectivity between them, and these are quantum quantum operations, essentially. And you have superposition and entanglement, this is what keeps the molecules together. So they have like, you have to have an entanglement that was there's no molecule. And, and so and, and so these are very like highlights, like, okay, clearly, like even like the fact that you can put like the, when you see the pictures, even the fact that they're drawing this picture, and you can see, there's a lot of stuff missing. And of course, the idea will be like, in the end, you have to do a quantum mechanically the problem, ultimately, we have to remember quantum mechanics was invented to solve chemistry problems. That was really it. Like, back in the day, there was a debate in science, which one is the most fundamental, like the chemistry or physics and people had no idea at the time, and this was a huge debate and, and, and then quantum quantum physics was developed. And then turns out, you can explain that or electron orbitals. So that means you could in principle there the whole periodic table, like the principle, and that was only when I get physics is the most fundamental one that was like that was like, when, when physics like they say that we were the big daddy. And but before that, there was nobody knew. And so, and this and this, and so it's gonna be like that, too, like, this is gonna have the same impact, not from like, very fundamental, but in the way chemists do the work that he's gonna, he's gonna change,
Sebastian Hassinger 39:16
it feels like there, that's likely the first practical impact from these systems we're building is going to be insight into physical sciences into chemistry and of material science. And obviously, quantum physics itself, like condensed matter physics, there's things that you can start to simulate on these machines in their current sort of imperfect calculator or even from, you know, information calculation state. The noise is not necessarily a barrier to scientific insight, right?
Yeah, absolutely. Absolutely. And this was always from the from the beginning, of course of quantum computing, because the very first proposals by Fineman and others of quantum computers was the idea. If you have a quantum problem, maybe you should embed in a quantum system
Kevin Rowney 40:00
simulated there. Yes. Right. Yeah, the very use case, right of molecular simulation. Yeah.
And so it was like that was really the inspiration. Then of course, Peter Shor came and he transform everything we think about, about quantum because he came up with a factorization problem and came up with efficient algorithm to do it better than anybody else. And this was like groundbreaking to me. Like, we still we still do not understand. Like this, this gray area like Peter, if you think about what it can you do stuff that's not like, that's really, really good. That's not Quantum. And I'm going to spiritual because there's like, very few that's there. Nobody else stands at the same level. It was like that groundbreaking
Kevin Rowney 40:36
research a thunderbolt of a breakthrough, but also be it's just set so many expectations, I think, arrive for people from the outside of this industry, because they're like, Wow, there's this huge breakthrough. So where are the results? I mean, we've got we're a long, long way away from anything that could run Shor's algorithm. Hmm. Decades, maybe lifetimes, right? I mean, who knows? Right?
For useful problems, right? Yes, but
Kevin Rowney 40:58
at least for that particular algorithm, right. Yeah. Right.
I think that's we can just I can probably factor the number in my head.
Sebastian Hassinger 41:04
Yeah, exactly. Right. Like the the NMR experiment that IBM Research did Almaden where they used. They used Mr. and Mr. qubits to factor 15. That was the first implementation. Sure.
That's right. Yeah, that was the very first one. And it was like, it's kind of ridiculous, oh, three times five. equals 15. That's like, who knows? I know that my son knows. But, but it was important like principle. And so it could be it could be that, that this landscape of class of classical art, or like algorithms, or classical problems that are more powerful, it could be that it's huge. And we don't know that. But I'm not surprised. Because if you look at the history of classical computing, people didn't come up with the algorithm until they actually had computers to run them.
Sebastian Hassinger 41:50
Yeah, sure, is really unusual in that way that that he came up with an algorithm. There's absolutely no way to like how where it came from. I have no idea. I've talked to him. He doesn't really know.
Yeah, I think that I had maybe a theater and that's what I think that's what everybody everybody like always asking how you come up with this. Yeah. And Scott aaronson, when he was at MIT, he I saw him in a seminar really going into like the technical details exactly what makes this there's complexity, like the complexity, despite what makes it so special for Shor's algorithm. And he made up this was a talk he made in there it was was Peter Shor. And he made a point that in every slide transition, there was the head of Peter Shor, like flying and crashing and stuff and like pictures of fish or like raining. Like otters was like laughing hilarious. Hilarious to see Peters, you know, shorthand comes and destroys them of the old the old views. And he was like, it's just got to wait to say like, this was like showering. So
Kevin Rowney 42:56
I love his talks up on YouTube. They're so fun, fun,
buddy. Like I made sure every slide have Peter shores headed.
Sebastian Hassinger 43:07
So you mentioned you mentioned just before we started recording, that was some major update recently, or a major breakthrough. We were progress recently. But what were you referring to?
Oh, yes. So So in terms of hardware, like I follow very closely, the existing hardware and so on. So So of the Dominion, which we have right now, of course, I get some breaks and superconducting qubits that will be like think of IBM has is investing a lot in this technology. Yet, he also has invested on LinkedIn noise, and D wave has also invested in this technology. And the way you just superconducting qubits in a very different way than IBM started getting, because they're not gate based their handlers. And the wedding at the web does a very single purpose machine. It's not like universal. But this is, it's been very valuable right now, because they can have, they can run stuff with a lot of data points more than other people that maybe the problems they can solve are more limited. And then of course, you have the benefits of superconducting qubits is that they're very fast. Most of it works on silicon, but not everything. But you still need superconductors to make them work. And to cool them off, which is a big, big, big challenge is to call them up. But you can use the fab techniques that we all know and love from Intel. Nothing has changed there. And, and so there's a promise that the scalability like there, however, the quality of the qubits is not so good. So once they've been improving consistently, they still have to pass a threshold, such that once you pass that threshold, you know that once once you have qubits reliably beyond that, you can take many, many qubits to support between that support one, one cubed n is known as error correction. So it think of it like one of those like, cheerleader pyramids, when you have a lot of people holding on one on top right, so it's a bit tight. That's kind of like ever correct And there's many techniques on how to do it or correction was like, there are many shades of the pyramids. So the other ones are ion traps. So I interrupted, it's like you take, you take an atom extract, an electron is charged, you can use like electromagnetic forces to put in a vacuum. And you can use electromagnetic forces also to compete between them, then the challenge is that, that you have to isolate one atom, so you have to put in a vacuum bagging chamber. This adds a lot of complexity. And also that the, the way you connect them with this radio frequency signals, and it's quite slow, actually. So it's big, it's slow. And sometimes even the scalability of it is limited. Like you don't have enough places to put more atoms,
Sebastian Hassinger 45:40
right? I've heard the traps can't really scale beyond about 40 atoms. Yeah,
so yeah, exactly. So and then the other one, that's and there's an atom atom competing, it's similar, but it doesn't use ions. And it's just a lot of lasers instead of lasers, but he has a lot of similar similar challenges. The other one is photonics, quantum computing, photonics is kind of like a very different beast, mostly because
Sebastian Hassinger 46:06
nobody has desert Rosa. But so like this, just very simple if like, you need to make light interact with light. And if you take two flashlights and put them on each other, they use like Lego toy, right? Like, doesn't want to interact with them. So that's a big challenge. That's it, it's very hard to make the gate. And but there's many people doing them. Of course, the benefit of that is like if they get them to work, right? You don't have to worry about decisions of calling them or backing chambers. But nobody has nobody has shown either scalable way. And so these are like the like roughly the three categories or more known ones than that. And but then something that was missing in the new industrial players was something based on silicon. So something is hoping that it's really like solid state devices, like more similar. And so of course, like, as I started engineering, so I had to learn a lot of pn junctions and so on for your diets, which is you extract some site, you extract some electrons from one place, you dope them in the other, and you use n and so the idea was, uh, can you do that? Can you can you just set your dope up, makes a big silicon wafer one way, and then you dope it the other way. And you glue them together to make diodes and transistors? Can you like, take one atom and just like, extract one electron? And that can you use that to manipulate? Yeah. And the answer is, yes, of course, you can do that. But doing operations is quite, quite challenging still. And a few weeks ago, in nature, there were three different independent teams in nature in nature, of course, is the biggest science journal, there were three independent teams, and that show that they had this kind of file called, like, envy centers, this is for for vacancies, because they extract, like trying to create a hole. And it's the biggest thing. So they showed that using this kind of qubits, that they were able to surpass this threshold to get to error correction like this. And this is this is really, really big. Because maybe there's a way that nobody has done computer computers right now. And there's an industry to make scalable computers look
Kevin Rowney 48:13
like a like a whole a whole new architecture. Right, a whole new branch of the tree. Yeah,
that whole new architecture, completely new architecture that. And so So wait, so the question is, is this the transistor moment? And and I will say like, I think we don't know yet. There's still a lot of questions about interconnecting them. And the actual practicalities of scaling.
Sebastian Hassinger 48:36
What isn't there a speed issue as well. So as I when I saw it, like, in the diamond envy centers, the coherence times are really long, but but to to interact with them was very, very slow, if I remember correctly.
That's correct. Yeah. Yeah. Yeah. That's, that's it. That's it. So they were able to show they wanted to do two, they were to do to me centers in one and like, in one place, so those were fast. But beyond that, the question is out there Absolutely. Is the question. I will say they're still once if that is resolved, then we will look back. I think this was a register moment. But But until then, we should not claim that yet.
Sebastian Hassinger 49:15
Right? It could be that it's it's flash RAM, maybe quantum RAM, right? Yeah, maybe.
Unknown Speaker 49:22
Kevin Rowney 49:23
that's really interesting news. I hadn't seen that flyby on the newsfeed. So that's, that's a good one to know. But I will look up those future papers. Yeah.
Yeah. So it's from it's very exciting because somebody's doing something new. And it's like he has some promise. We'll see. We'll see. We'll see what he gets. Of course, the science takes a while. It's not gonna be like in six months. It's now a company's doing this. Yeah, he's sending more papers before people have confidence. But but it to me like this is this is exciting that I'm like, we're looking at news articles and thinking like what is this transistor moment? This is such a huge moment. Are we seeing it in front of our eyes? For me, I love that part about the industry.
Kevin Rowney 49:59
This The coolest thing about this reading group we did with the study and on this podcast is so you feel like you're being you're getting a ringside seat
Sebastian Hassinger 50:08
view of history or not. It could be not history. We don't know.
Kevin Rowney 50:12
Yeah, you're right. Yeah, exactly. It's just revealing itself.
Sebastian Hassinger 50:16
All right, so we've got just a couple minutes left. Sorry. Any anything you want to touch on before we wrap up?
No, I think this was, this was a really fun conversation with you guys. Yeah.
Kevin Rowney 50:27
I hope you had fun. And yeah, we had a blast. I really appreciate your time. Thank
Sebastian Hassinger 50:31
you very much.
Kevin Rowney 51:13
Well, Greg, that was I thought of a fun interview with CSR. He's a fun guy to hang out with. And he's been really, really, I think, a helpful mentor and guide. And for me, and I think you said that, yeah. And also in terms of grappling with, with understanding quantum mechanics, I hope it's clear that you what we're trying so hard to do with this podcast is deliver to people who are non physicists, right, some deeper level of understanding of it from the perspective of you know, technology and those, those trends. So anyway, I, I really enjoyed his reasoned and balanced view of both the potential applications on one hand and the frank admission of the of the serious current limits. It's hard, I think, to have both of those views, in your minds at the same time, it's a little bit fatiguing, for some people, but I think it's the appropriate way to see this face.
Sebastian Hassinger 52:03
Yeah, in particular, I liked the discussion of variational approaches, because those are generating the most sort of interest in this NISC era with these intermediate scale machines that are quite noisy, there's still some hope that marrying classical compute resources with quantum machines at this nascent stage will yield some kind of advantage. And also the the discussion about quantum supremacy and quantum advantage was really, really great.
Kevin Rowney 52:32
That that was interesting. And that might have for some of the listeners have flown on by rather quick, it was a just a short set of remarks there. But it turns out that there's numerous examples, recently, even right of especially prominent advances that got a lot of hype and quantum computing, that were later retracted. It's called a de quantization event where the the perceived advantage of the the quantum computer or the algorithm is found to be false. And we have to reconsider.
Sebastian Hassinger 53:03
The turns out it doesn't need to be entangled. of quantum information science is is information science, just with all of the information, right. And it turns out that so far, most problems that we've looked at, don't actually need all of the information they there you can find a subset and a narrower approach that yields a sufficient answer. So it's a really fascinating exploration.
Kevin Rowney 53:33
Very interesting and treacherous mathematics. So watch out. Yeah, I guess that was really interesting. We had that little back and forth about the the quantum supremacy result that was announced by Google all the way back in I think 2019. Yeah. And so we touched on them. But since that recording, I mean, just the last couple of days, while we've been in production on this on this podcast, it was it was well announced, probably announced that a team in China had found a way to exactly replicate the precise same quantum supremacy results that Google did, but on a classical computer with classical algorithms and just an amorphous a supercomputer. So it's got to be actually probably the most pronounced D quantization event. I'm not sure if most people have been watching the press on this one, it the news of the quantization is less greeted with optimism and
Sebastian Hassinger 54:24
attraction is never as big as the headline, right?
Kevin Rowney 54:27
Yeah, exactly. Right.
Sebastian Hassinger 54:29
It's still as scissor pointed out, though, it's still you know, I think we we, and maybe the presses is somewhat to blame for this, we we over index on the, on the announcement on the the, the whatever the significance of the actual advantage and, and don't recognize in the larger context that this kind of exploration is is necessary, we won't find real advantage unless we find all the ways in which you might manage That's the self that don't turn out to be true, right? That's the scientific method. You got to disprove everything really? Yeah, so
Kevin Rowney 55:07
the adventure that doesn't involve stubbing your toe. It was also, I thought really cool to close in, in our remarks around, you know, CSR gesturing towards, you know, new innovations that were happening, you know, while while we were near the time that we're talking around brand new quantum computing architectures, which appear to represent possibly new breakthroughs that could provide massively more scalable manufacturing techniques that you know, not maybe full, zero Kelvin, frigid temperatures. So I, I wonder, Sebastian, do you think give a firmer command of of
Sebastian Hassinger 55:44
what he was referring to is nitrogen vacancies in diamond. It's a technique for popping, popping electron out actually. And using that that hole as the qubit. Essentially, the advantages very long coherence times the T one times are quite long. That can be you know, seconds, or even in some instances, longer like minutes, which is kind of amazing to think about when you think of the you know how hard it's been for superconducting qubits to get to one millisecond T one times. And the challenge has been manufacturing, and the performance. So it takes a long time to interact with these types of qubits. There's been some suggestion they might be better suited for RAM, or memory storage. But I think there are also people working on techniques for increasing the the the gate readout speed. But the paper that that says I was referring to was I think he said three different independent groups all came up with a benchmark that showed that they could maintain coherence longer than than required for error correction. So they can they can implement error correction effectively in these diamond vacancies, which is really exciting. That means, essentially, if they can tackle the fab challenges and the performance challenges, we have a qubit. That's, that is amenable to error correction out of the gate, which is a big, big jump.
Kevin Rowney 57:17
And so we don't know for sure, but we could be living through these days, right, that transistor moment. So to strike for that. I think that framing
Sebastian Hassinger 57:24
is really that in the Wright Brothers phone, which was delightful as well. It's that's the reason why I, I am compelled by this field, right? This is why I keep coming back and trying to figure out exactly what's going on even, even despite all the complications and the difficulty that entails is because we have the potential of moving through another moment that is that transformative as transformative at the moment when Bell Labs invented the solid state transistor, and, and we were off to the races. And Moore's law really started to kick in. That's right,
Kevin Rowney 57:59
a ringside seat on these developments. I mean, that's what really, really attracts us to this whole area. So I think it would be useful for our audience Sebastian to go back and just review this absurd idea of of coherence and T one time, that's a trade term that her perhaps some of our listeners need more context on, would you be willing to take a go at better explaining,
Sebastian Hassinger 58:26
as a total non expert, T one is, is a time as the time that a qubit maintains a coherent state. So the superposition that's put into the entanglement that it's involved in other qubit with other qubits that is effectively in existence for a very, very short period of time. Because, you know, the universe does not like to to have little tiny bits of it isolated, it likes to be as I said, like to be a molecule or to be an atom, it has to be fully entangled. All the pieces of it have to be fully entangled. And so what we're doing is splitting off tiny quantum subatomic particles and isolating them in time and space, so that we can interact with them in a controlled way and perform calculations. So that time window is vanishingly small. As I said superconducting qubits only recently reached the one millisecond T one time,
Kevin Rowney 59:26
Sebastian Hassinger 59:30
It is so I mean, that is a among the many many many many fundamental challenges that exist for computers work. Yeah, yeah, I
Kevin Rowney 59:40
mean, there's so much potential but also these these quantum states with so much power, I mean, they can be so easily are so fragile can be so easily disrupted by the slightest intrusion right of a bit of Brownian motion for Pete nearby or who knows a cosmic ray. There's all these ways that the whole calculation can be done. wiped out by by ambient conditions rattle. So there needs to be increasing coherence time shows that there's actually novel new architectures that could perhaps be way more resilient for those fundamental challenges that nearly the entire space face. That's right. I will there it is. I hope I overlooked the transistor moment but you know, we're skeptical. We'll we'll see. Okay, that's it for this episode of The New quantum era, a podcast by Sebastian Hassinger and Kevin are cool themed music was composed and played by Omar close to me. Production is done by our wonderful team over at pod fly. If you're at all like us and enjoy this rich, deep and interesting topic, please subscribe to our podcast on whichever platform you may stream from. And even consider, if you 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 we'd really appreciate your help spreading the word and building community. Thank you so much for your time.
Transcribed by https://otter.ai