The New Quantum Era

Sebastian Hassinger: 00:01

The New Quantum Era, a podcast by Sebastian Hassinger.

Kevin Rowney: 00:06

And Kevin Rowney.

Sebastian Hassinger: 00:31

Welcome back to the podcast. This is Sebastian again. I've got a couple of episodes for you. Kevin will be returning. We actually have recorded an episode this week, that I'll be cutting together and posting early next week.

Sebastian Hassinger: 00:48

But in the meantime, back in April, I went to a workshop at the Simons Institute For the Theory of Computing at the University of California, Berkeley. They had a week long workshop on near term quantum computers, fault tolerance and benchmarking, and quantum advantage, and quantum algorithms, which was fantastic. There were a ton of really good talks, great presentations, great speakers, and the Simons Institute was kind enough to provide for me a space to record, and I grabbed a couple of the guests, the invited guests who were giving talks and recorded these interviews. Today's is gonna be with Jens Isert, and he is a professor at Vrije University Berlin. And you'll hear, again, a little flustered with trying to set up the space.

Sebastian Hassinger: 01:42

I didn't do a proper introduction, and I apologize to Jens, but we had a really great, conversation about his work. He's been, really a a pioneer from very, very early on. He's been in the field for something like 20 years. He's been doing a ton of work, with his group at the Freitas Universitat, and we talk it's a pretty wide ranging conversation around the, the work that he's been doing around benchmarking, verification, sort of the limits of, the variational and NISC algorithms and other topics. So I think you'll enjoy.

Sebastian Hassinger: 02:55

Thanks for joining us Jens. Can you start just by sort of a brief introduction?

Jens Eisert: 03:00

Well, the pleasure is on my side. Thanks for inviting me. My path, I mean, my you mean my biograph Sure. Yeah. So in a way, I'm a veteran of the field, you can say, and that I did already my PhD in quantum information science on the role of entanglement in quantum computing and quantum communication.

Jens Eisert: 03:20

And when I started, the field was still Right. Quite young, although not that young. I mean, some of the big results, had been Of course. Achieved, but, I was doing, like, other work on open quantum systems before. And then when I found this field, I was, like, super excited.

Jens Eisert: 03:36

It was, like, hiding in the library library for weeks reading all the papers I could find for, the at at the time. And then I was, like, hyped and wanted to work on this. And then I, arranged things. And since then, I've been in the field. That was, like, more than 20 years ago already.

Jens Eisert: 03:53

And since then, I've been, like, thinking about the subject and and now also leading a team that's, in a way, rigorously minded, working with ideas of mathematical physics, but at the same time being pragmatically motivated on the protocol, on the phenomenon, on the on things that that that that exist. And Right. Sometimes you work with experimentalists, and that's kind of our place in the community.

Sebastian Hassinger: 04:18

Yeah. So, I mean, is it fair to say that you are more I mean, you're trying to establish theoretical sort of limits or guidelines to what we can do with the hardware that exist today? Is that a major area

Jens Eisert: 04:32

of concern? My my recent interests are pretty broad, but, yeah, I mean, my talk was about that today. Yeah. That that's right. So I'm interested in what we can do with quantum computers in the long run, but also in the near term.

Jens Eisert: 04:48

What what, like, what goals we could find, like, what things we can do that may have practical applications on the one hand. But I'm also interested in finding the limitations, like finding the fine line between what's just not possible and what's possible Right. And, also to see the fine line be with the classical world and the quantum world. And, yeah, the development of the last years have been extremely, like, exciting Right. And and and productive.

Jens Eisert: 05:13

And, yeah, I wanna see how things go forward, staying away from the hype because it's also, like, a bit of maybe exaggerated, expectations. And, yeah, I'm interested in finding out what's, like, honestly possible in the next years to see what expectations we can raise.

Sebastian Hassinger: 05:32

Is that in part motivated by the fact that, you know, sort of the the most, I would say, popular or best known approaches to hopefully get some sort of value out of NISQ machines, out of these noisy machines are variational algorithms or techniques, VQE, QAOA. And those are they're heuristic. Right? There's no there's there's limits to what you can provably do with or accomplish with those types of of approaches.

Jens Eisert: 05:55

Yeah. That seems fair to say. I mean, this is a hugely exciting idea. No doubt about this. It's really difficult to make strong claims on the utility.

Jens Eisert: 06:04

I mean, there's no proven separation. Well, there's some, but under, like, somewhat funny conditions, it's also easy to, like, make strong statements on their computational power, but it's it's worth finding out. And, what, like, one of the research themes that we've been involved in is to kind of find out what these devices can do and also, as I said, find limitations more often, understand the the line of how classical simulation methods can go and on the other side where a regime could be identified where you can do interesting things. And, even though it may be difficult to formulate, like, the the holy grail proof of really a full practical advance of a type on a on a near term machine, There's kind of proof pockets. You could Yeah.

Jens Eisert: 06:55

Move a couple of things that are interesting that if put together as a as a kind of puzzle, they give rise to a more complete picture. And I see my job in parts on providing puzzle pieces to the to the big picture. Hoping that if you have enough puzzle pieces together that the a more coherent picture emerges at the end of the day.

Sebastian Hassinger: 07:14

Right. Right. And is that in part I mean, I I see that in 2 lights. 1 is to try to, as you said, sort of provide more puzzle pieces to potentially get more value out of these current devices. But it also seems like it can help inform sort of future direction of, it's almost the co design idea of like, how the next generation of hardware, like you're, you're feeding ideas and observations and theory into the experimentalists' hands to try to help them and their, their progress on the on the hardware side.

Sebastian Hassinger: 07:45

Is that fair to say?

Jens Eisert: 07:46

That's absolutely fair to say. I mean, it's actually it's it's more than that. I mean, it's true in 2 senses. I mean, in on on the one hand, we are actually providing, lots of tools to kind of benchmark quantum device. That was actually the topic of the session today even though I decided not to speak about this, but that's a topic that's pretty present in in in in my team to, like, think of, like, randomized measurement techniques, randomized benchmarking techniques, and so on to find out what's happening in the device.

Jens Eisert: 08:12

But not only that, but use that information to kind of then feed it back into the design of the device to see how you can improve things with this actionable advice acquired by the benchmarking technique on the one hand. On the other hand, I mean, I've even I've been mentioning, like, some no go results. And, I don't see no go results as, like, killer statements. Right. So they're not meant to shoot down things.

Sebastian Hassinger: 08:38

Right.

Jens Eisert: 08:38

I mean, early on in my career, I I I I, formulated a no go. See, there was, like, then much cited and so on. And people complained about this for some time, but then they realized, wait a minute. You can overcome this by clicking outside the box.

Sebastian Hassinger: 08:50

Even today during your

Jens Eisert: 08:51

time, you were saying, like,

Sebastian Hassinger: 08:53

so is where's the loophole? You could tell people were sort of looking for the loopholes around your no go statement.

Jens Eisert: 08:57

Yeah. I mean, and also some years ago, I had a result on quantum computing with Gaussian states. It was a kind of idea of using, like, continuous wave of quantum states to produce to do measurement based quantum computing. And for a large class of state of that kind, I could show that they will not work in the sense that if you do something on one side, then the impact on the other side will necessarily decay exponentially with the distance so that you cannot even transport information. And people got, like, very upset about this saying, oh, is this now showing that the idea doesn't work?

Jens Eisert: 09:31

I said, no. No. No. I'm just saying that for arbitrary projective measurements, this is not going to happen.

Sebastian Hassinger: 09:36

Right.

Jens Eisert: 09:36

But these are the rules of my game. And then a couple of years later, somebody found out that if you project onto certain code words of quantum error error correcting Mhmm. States, like GKP states, then you can actually do this, which was slightly outside my theorem. Right. And then it could be done.

Jens Eisert: 09:54

So Right. Right. Right. That was precisely the, development I had hoped for and that I wanted to invite people to think a little bit outside the box. Yeah.

Jens Eisert: 10:01

And this also did happen. Yeah. Yeah. And, yeah, that I mean, I like, you know, theorems, but, again, not to kind of produce roadblocks, but to motivate people to think in slightly different directions.

Sebastian Hassinger: 10:14

Yeah. I mean, the the that's sort of the core of the scientific method anyway. Right? It's it's it's disproves to the then provoke new theories that try to get around this principle.

Jens Eisert: 10:23

It doesn't always work out well, but

Sebastian Hassinger: 10:25

Yes.

Jens Eisert: 10:25

I mean, if it works out, it's it's a nice development.

Sebastian Hassinger: 10:28

Yeah. Exactly. Exactly. And so returning to what you're talking about today, I thought the benchmarking, aspects were really interesting. And also, you know, you were mentioning, you know, those, being able to predict the outcomes or the distribution, sorry, of the outcomes.

Sebastian Hassinger: 10:51

With Clifford Gates, you're able to easily predict the the, the distribution or or to model the distribution. But as soon as you add one Toffoli gate, it becomes really difficult to to to forecast. Is that No.

Jens Eisert: 11:03

No. No. It's it's it's even more subtle than that in the sense that, I mean, there's a forward direction and a backward direction. Mhmm. I mean, let's not forget, there's a circuit.

Jens Eisert: 11:11

You make measurements in in a fixed basis, so you get a classical distribution from, quantum circuit plus measurements. Okay. And, I mean, these are, like, sampling schemes and and, like, to to sample from the output distribution is a specific reading of a of a similar of a classical simulation of a quantum device, and that's well understood in what precise way that Clifford circuits would be efficiently classically simulated. Right. More than one way.

Jens Eisert: 11:34

So in in in in although known ways, it's a classically assimilatable thing. And you can also add t gates. If you add not too many of them, like, logarithmic, you can still do that, and you can kind of basically keep track of the of the branches, and you you can still do that. The the converse problem is you have a distribution given and you wanna learn it in a kind of machine learning sense or in the in the PAC learning sense to put it in fanciful terms. But this is, like, really, really the the the kind of approach of computational learning theory that given samples from a distribution, can you actually learn what the distribution is like in in certain fine prints of the of this of this question?

Jens Eisert: 12:11

And then it turns out that why Clifford circuits can be efficiently learned, this is no longer true if you add 1 t gate, which would be strange and striking because the the forward direction is borderline trivial. It's I mean, clearly, the classical simulation is efficient because you just keep track of the of the of the branches of the quantum gate. The the the backward direction is not possible, which is I mean, you're not forced to find this counterintuitive, but it's interesting. It is

Sebastian Hassinger: 12:39

interesting. So I mean, does that I mean, is it possible that sort of lends clues to possible areas of advantage? Does that that sort of suggest areas that need to be investigated for for potential quantum algorithmic advantage?

Jens Eisert: 12:55

Well, yes and no. I mean, literally speaking, I'm not sure. Like, whether this specific result has any implication, I might be somewhat doubtful about this. Of course, that that result is part of a program that is, of course, directed towards finding advantages

Sebastian Hassinger: 13:13

and types.

Jens Eisert: 13:14

So, like, also some of the I mean, I mentioned this briefly at the end of the talk when we looked at random pool features. I mean, there's a lot of activity also in my team to think about new, like, classical simulation methods that are not not completely standard

Sebastian Hassinger: 13:32

Right.

Jens Eisert: 13:32

And that kind of identify, like, a classically accessible regime, then, like, hoping to delineate the boundary of the of of the quantum side so that so to say, where you can hope for quantum advantages of of a kind. And this bigger program is surely directed towards finding advantages. Mhmm. I mean, I even mentioned this at the beginning of my talk when I mentioned, like, short circuit learning advantages, which is a question where you wanna see whether constant depth quantum circuits have a superior power concerning certain well defined learning task compared to well defined log depth classical circuit with a meaningful locality structure. And this seems to be the case, which is, if you want a quantum advantage of a kind, it's still not the big answer because it's not very practical.

Jens Eisert: 14:26

Right. And the kind of distribution the the the concept class are a bit crazy and so on. But, like, lacking the the big proof of the big statement, I think it's sensible to try to approach the question from all possible directions so that you come closer to the real thing in the middle by Right. Providing sounds.

Sebastian Hassinger: 14:46

And that's honestly along

Jens Eisert: 14:48

the way.

Sebastian Hassinger: 14:48

That makes a lot of sense. And that's that's what I've noticed from from talks that I've I've heard of yours before and looking at papers as well. It it does feel like, you know, part of what you are trying to do is break up preconceptions in all of the various approaches. So there's there's the, you know, the error correction class of folks who are looking at known, you know, surface code, color codes, etc. There's the, sort of the experimentalist turned hardware engineers who think they're on the right path to building, you know, their transmons or whatever there to a particular type of cubits.

Sebastian Hassinger: 15:21

And it, you know, and then there's the people who are trying to squeeze, you know, performance out of the hardware that happens to be available today. And it it feels like what you're trying to provoke is is all of those individual groups and others, there is clearly others, to think about what they may not be considering in their existing approach or challenging the, the, traditional approach, if you will.

Jens Eisert: 15:48

Yeah. Yeah. That's, yeah. That's a good way of of putting it. I mean, when when I have, like, applicants to the team and they ask ask me, like, how the team functions, and and then I'll often say that there's a kind of group style of a type.

Jens Eisert: 16:04

And even though this may not be verbalized or even not even known to people Right. There's there's some things that that are in in like, just happening in a certain way. And what I often say is that we are, like, mathematically minded, but still pragmatically oriented. Mhmm. But another thing I tend to say, and it's not completely untrue, is that I personally like and that's also a sentiment shared by many people in the in in the team is that I like to write, like, first papers on a subject or last papers on a subject.

Jens Eisert: 16:31

And by that, I mean, like, the first paper is like something crazy. Wild idea that you throw to the community and it may start a new field Right. Or not do anything at all. Right. But it's like a completely fresh, like, approach to a problem.

Sebastian Hassinger: 16:46

Yeah. And

Jens Eisert: 16:46

this can be, like, wild or or whatever. But at the same time, I also like last papers on a subject. For example, I mean, in the morning, we heard a lot about randomized benchmarking. And this is a field where we, like, try to produce the last table of a kind where we have, like, lots of different schemes and all of over 30 different flavors of randomized benchmarking with different groups, like the Clifford group, the dihedral group, real randomized benchmarking, cycle randomized benchmarking, linear cross entropy benchmarking. Your head explodes with a different variance.

Jens Eisert: 17:18

Yeah. And then it was also not so clear, like, in what way the the type of data you would expect, namely linear combinations of decaying exponentials would really come out of the data naturally. And we sat down to formulate, like, a a framework theorem that encapsulated the known schemes of randomized benchmarking and showed in what precise way, it's true that if the implementation map is close enough to the anticipated setting that you really get the anticipated data of exponential decaying functions. And then as a as a bonus track, we also found an optimal way of getting the classical post processing together. And why am I saying this?

Jens Eisert: 18:01

Why that's not front page news because it's not something new. Right. But it's still valuable because, it kind of formulates a framework. And once the framework is settled, you really know what this kind of machinery is doing, and then you can also progress from there.

Sebastian Hassinger: 18:15

Right.

Jens Eisert: 18:16

This is kinda funny because we submitted this to a good journal, and then it was, like, desk rejected on the 1st day, saying that we don't publish review articles. I said, no. It's not a review article. It kind of aims at solving the problem. Yeah.

Jens Eisert: 18:28

And then, oh, they apologize. Now it's a it's a well cited article.

Sebastian Hassinger: 18:32

Yeah. Yeah.

Jens Eisert: 18:32

But it's kind of, well, the point I'm trying to make is both kinds of work are good work. It's good to be creative and and and and think outside the box Right. But also to sit down carefully and and think an idea until the end.

Sebastian Hassinger: 18:47

Yeah. And tie it all on.

Jens Eisert: 18:49

Both things are like characteristic for our

Sebastian Hassinger: 18:51

And and what was sort of the what's your set of conclusions out of that randomized bench? Mark? There's limits to how much it can scale. I remember you were saying earlier today that, past I think 50 cubits or so, there's no real value to the randomized benchmark. Is that?

Jens Eisert: 19:07

Well, I mean, said something about this in this intro tutorial today. Yeah. Scaling is an issue. Yeah. That that's right.

Jens Eisert: 19:25

But there's there's more pressing questions. Like, what really the the diagnostic information is you can get

Sebastian Hassinger: 19:30

out Right. Right.

Jens Eisert: 19:31

Of these data and, like, what you can infer about the the setting you have, like, beyond getting, like, average gate fidelities or something or whatever. You can kind of Right. Squeeze more detailed diagnostic information out of a randomized Mhmm. As a kind of increasing an increasing understanding of what we can learn about quantum devices meaningfully by suitable randomized measurement schemes that go beyond what the original applications are. Okay.

Jens Eisert: 19:55

I mean, scalability is an issue, but that's not the the core issue, I would say.

Sebastian Hassinger: 20:00

I mean,

Jens Eisert: 20:00

depending on I I like to say that you can say a lot about little and little about a lot. I mean, depending on what you want. If you think of a, like, a huge scale system, then randomized benchmarking would not be the right tool. Then you have to think about other methods to to kind of certify the function.

Sebastian Hassinger: 20:15

Is there is there any broader relationship between randomized benchmarking and just, validation of quantum results? I mean, there's a lot of, work going on anticipating, you know, systems that are too large to actually, you know, efficiently simulate. So how do you validate the the work that's being done by that system? A long long

Jens Eisert: 20:33

and and Not really. Lengthy question. I mean, you wrote a

Sebastian Hassinger: 20:35

review article on on benchmarking certification that kind of,

Jens Eisert: 20:42

tries to approach the question, we have a kind of a mind a mind map on the first page that kind of classifies known methods according to the assumptions you are willing to make and the, and the level of detail you wanna learn. And many of the known methods can be kind of placed in that two dimensional diagram. And then what you said, I mean, it's it's a bit fits into the scheme because, I mean, some I mean, if you have a large scale system, then you might think about, like, direct certification schemes. Right. And so generally speaking, I mean, in this whole business, you want to there's a couple of, like, properties you wanna see, like, there's a there's a data that you wanna see fulfilled.

Jens Eisert: 21:23

And, of course, for example, the measurements you wanna do or you you think you you must do should be, like, experimentally feasible, then you would want to have a kind of understanding of the sample complexity, like the number of shots. And depending on the experimental platform, this is, like, super relevant. You think of, like, cold atoms. You you make an experiment, and one click costs, like, 15 seconds. Yeah.

Jens Eisert: 21:46

You think twice what you do because, I mean, getting good statistics means many runs of the experiment takes a lot of time. And, also, what the community has, like, painfully learned, I would say, I mean, I have painfully learned it, is that robustness is is key. Mhmm. And people speak about spam robustness, like state preparation and measurement Yeah. Robust ness, in that the preparations are wrong and like, everything is a bit wrong Yeah.

Jens Eisert: 22:10

But you still want to, reliably learn about the property of a of a quantum system. And that's nice to have, but this is not so easy. And many of these randomized techniques, they are aimed at being, like, robust in a certain way. And if you can do it, if you then you should have it. It's not always attainable.

Jens Eisert: 22:30

I mean, sometimes you have to be a bit more forgiving, but there's increasing the portfolio is growing of, like, randomized methods that would also work in a in a scalable fashion so that you can assess, like, the functioning of a of a big scheme, like, of an analog device. And we heard a bit about this this morning. Right. And we also have work on that where well, the last word hasn't been spoken in the sense that it really depends on what you are willing to in in willing to invest and what what you want at the end of the day. Mhmm.

Jens Eisert: 23:01

But, again, these randomized techniques are basically set up to be specifically robust, which is quite important. Like, we have, for example, calibrated like, help the Google team calibrating their Sycamore chip. That was a fun joint venture between our team and and and the Google team. And, and there, the initial ideas were, like, super we're quickly done. We had ideas of super resolution, many thought optimization.

Jens Eisert: 23:26

We're, like, super proud of learning the Hamiltonian of that system from data. And then it took us, like, 2 more years or so to make it properly robust. And once the data came in, the devil was in the detail, and we had to work very hard to to overcome certain limitations and make it properly robust. And now we've nicely understood it, and the paper will presumably also come out. We are very happy about this.

Jens Eisert: 23:50

But robustness is not only key, but it can be pretty nasty and painful. That's sort of that's the challenge in many cases.

Sebastian Hassinger: 24:00

Interesting. And so okay. So returning to your talk today, in a in a way, you were you were I wouldn't say you're, closing the books on the sort of potential from NISC machines and variational approaches to VQE or QAOA or QML. But do you have a sense, you know, are you I guess, put it simply, are you optimistic or pessimistic for for getting sort of valuable, functionality out of the the type of devices that we can build today? Noisy?

Jens Eisert: 24:34

This is an interesting question. I tend to be cautious. I mean, I see my main role as providing tools to fairly assess this. And in the last part of my talk, I have been talking about new work that we did on the impact of non universal noise on the functioning of circuits with interesting results in the sense that certain non universal noise gives rise to predictions that seem counterintuitive in in in a way. And then equipped with those tools, you can see how far you can push things.

Jens Eisert: 25:05

And I'll also talk about error mitigation. I'm not saying error mitigation doesn't work. I mean, I would wouldn't say this, but I we found certain limitations concerning their scalability. And you these are theorems. I mean, then the assumption may be wrong or inappropriate, but as they are, they I mean, they could be wrong, but I doubt that.

Jens Eisert: 25:26

So one has to work around them somehow.

Sebastian Hassinger: 25:30

Yeah.

Jens Eisert: 25:31

So my my professional job is to provide tools of that type. Mhmm. Now, of course, you can still ask me, like, what what is your personal way of doing it? I mean Yeah. At the end of the day, does it work or does it not work?

Jens Eisert: 25:43

I'm I mean, that's maybe surprising to hear given that I had this, like, very negative title slide of my talk. Yeah. I'm reasonably optimistic. I mean, we know that, like, we have locked depth available. That's not very I mean, we don't have very deep circuits.

Jens Eisert: 26:01

We are Right. Pretty much pretty quickly eaten up by by noise. Yeah. But locked, what does it mean? I mean, it depends on the on the on the numbers, like, how deep you can really go.

Jens Eisert: 26:12

Mhmm. And in particular, in the field of quantum simulation, I'm relatively positive that you could do something Right. With near term devices. And I I personally like to stay away from from hypes. Yeah.

Jens Eisert: 26:26

And if I may say that, to to my taste, like, the NISC era was a little bit overhyped Yeah. In recent years because then there's, you know, the the wonder weapon, you only have to do version algorithms, and then you solve, like Right. The the the the the world's big problems and and and, like, all optimization problems and routing scheduling Yeah. Whatever. And now suddenly people say, oh, we have to become fault tolerant.

Jens Eisert: 26:52

Before we are, like, fault tolerant, there's nothing you can gain. Right. So to my taste, the expectations were too optimistic before

Sebastian Hassinger: 26:59

Yeah.

Jens Eisert: 26:59

And are too pessimistic now. Right. And then I say, wait a minute. One must be wrong. And Yeah.

Jens Eisert: 27:05

Like, let let's see. So I'm I'm like, I was like, for many years, I was comparatively pessimistic Yeah. Compared to many others.

Sebastian Hassinger: 27:14

Yeah.

Jens Eisert: 27:15

And now I'm comparatively optimistic.

Sebastian Hassinger: 27:16

So you're a contrarian is what you're saying, Jansen?

Jens Eisert: 27:19

Well, a

Sebastian Hassinger: 27:22

challenge. You're challenging the the status quo.

Jens Eisert: 27:24

Yeah. I mean, I try to see things in perspective.

Sebastian Hassinger: 27:29

Yeah. Yeah. Yeah. Okay. So last question, you talked about having, you know, either the last paper or the first paper.

Sebastian Hassinger: 27:36

What are some of the wildest ideas that you're sort of contemplating or working on right now?

Jens Eisert: 27:45

Yeah. I'm not

Sebastian Hassinger: 27:46

Without giving anything away, I don't want to steal any of your publication.

Jens Eisert: 27:51

Yeah. I mean, I I I'm for example, I'm pretty I don't know. I'm I'm this is a lot of I'm not prepared for this question. Like, I'm I'm pretty convinced that for ex like, that's an old question, like, from the days of my PhD. Like, there's a notion of, like, a resource of entanglement.

Jens Eisert: 28:07

Mhmm. The so called distillate of entanglement that kind of captures a time as a resource. I'm pretty convinced that people tried hard to compute it. It was impossible to do for for for many people. I'm pretty convinced that that, like, finding out whether a state has a resource character or not, like, has the silver entitlement or not, that this is an undecidable problem, a Turing undecidable problem.

Jens Eisert: 28:29

So people no wonder people couldn't compute this because it's kind of basic and computable. But maybe more to the point, what I like being an academic and also what I like to like being academic outside the the quantum industry, although I like to work with the quantum industry, is that you can work on the the the hip topics of the time. You can work on error mitigation, error correction, fault tolerance, or near term devices and all the the hot stuff everybody else is working on. But at the same time, you can work on stuff that may not be crazy. It's not really qualifying for your question, but that's really outside the

Sebastian Hassinger: 29:08

Right.

Jens Eisert: 29:08

The beaten trails. It's like not top, not hit not a hip topic.

Sebastian Hassinger: 29:12

Yeah. Yeah.

Jens Eisert: 29:12

Anyway, for example, we had just 2 weeks ago a paper on a question that I have been fascinated about since many years. It was a core topic of my ERC grant, but not the one I have now, but of the previous one, like, from 10 years ago, like, how, how temperature comes about in nature. Mhmm. Yeah. I mean, we know that stat mech talks about maximum entropy ensembles, whereas quantum mechanics talks about pure states and evolving time.

Jens Eisert: 29:39

How are these pictures precisely compatible? And can you actually prove under meaningful initial conditions of low entropy, low entanglement, that they would evolve into states that for most times locally look like GIP states, that temperature actually emerges as a dynamical feature. And we made really solid progress towards a resolution of that question, which is fun. I I love it, and I'm super happy about this. Also, because I had this idea a long time ago, and then I gave this to to students at a postdoc.

Jens Eisert: 30:10

And they looked at it, and and and they they could make it work in the end after I've, like, tried a lot, some time ago. And I'm super happy about this also because it's a question that I find important, like, how does temperature come about in nature? It's really an interesting question. But it's like it's not a hip question. It's just not, the stuff.

Sebastian Hassinger: 30:32

It's no quantum gravity.

Jens Eisert: 30:33

It's not quantum gravity. It's also not, not a topic where the big players of the quantum industry are interested in.

Sebastian Hassinger: 30:38

Yeah. Yeah.

Jens Eisert: 30:38

But I think it's still an interesting academic question. And in academia, you don't have to decide to do one or the other, but

Sebastian Hassinger: 30:44

you

Jens Eisert: 30:45

can, in a way, do both or even get inspired. And, actually, we are now working on quantum algorithms used on based on these ideas.

Sebastian Hassinger: 30:52

Oh, very cool.

Jens Eisert: 30:52

You can even kind of combine the ways of thinking and, like, be a bit outside the the mainstream, which is interesting. Yeah.

Sebastian Hassinger: 31:04

I mean, that's that's what I find so exciting about, quantum information science, quantum computing, quantum technologies in general is that there is this very tight sort of, interaction between the sort of purely scientific and then the sort of technological development engineering side of things. It may be that this work on temperature, you know, quantum quantum or or, temperature in quantum systems has some kind of implications for a system engineering. Right?

Jens Eisert: 31:31

You started out by asking me, like, what the like, you you asked me about my personal biography. Yeah. And maybe I can kind of close the cycle in the sense that this is actually what fascinated me in the first place, that this is a field of research where you can think about the well, basically, the foundations of quantum mechanics, the the good old questions about what quantum mechanics is about on the one hand, but bring this all the way to technological implications. Right. And most of the time, you have to decide.

Jens Eisert: 31:58

You work either on technology, and then you already work on fine tune details that are to totally remote from the foundations in physics. Oh, you work on the foundations in physics, but then you are up in the sky and you are not connected in any way to technology. But in this field, you you can combine the worlds, and that's actually quite exciting.

Sebastian Hassinger: 32:16

Yeah. I agree. Well, thank you very much for joining us. It's been terrific.

Jens Eisert: 32:20

Thanks.

Kevin Rowney: 32:21

Okay. That's it for this episode of The New Quantum Era, a podcast by Sebastian Hassinger and Kevin Roney. Our cool theme music was composed and played by Omar Costa Hamido. Production work is done by our wonderful team over at Podfi. If you are at all like us and enjoy this rich, deep, and interesting topic, please subscribe to our podcast on whichever platform you may stream from.

Kevin Rowney: 32:50

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 would really appreciate your help spreading the word and building community. Thank you so much for your time.