qBraid on Quantum Computing: From Hype to Developer Reality

Welcome to Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome in the arena. My name is Alison Klein.

Today is a Data Insights episode, so I'm here with Jalees Dyrowski. How's it going, Janice? Hey, Alison. It's doing great.

How are you? I'm good. I'm excited for today's topic. Why don't you tell me who you brought along with you?

Yeah, I am very excited. You know, we just left Accelerated Compute and we've been talking all things quantum computing, which is far out there, but it's always exciting to learn from the folks and the experts. So today we have Kanaav Satya from QBrate. And Kanaav, welcome to the program.

It's so nice to have you. Yeah, thank you for having me. Excited to be joining for this call. Now, Knob, I was really excited to have you on the show, and then I started reading about you, but I got more excited.

You've spent your career working kind of at an intersection of plotted computing and software. What drew you to this space, and how do you see quantum ecosystem growing and how platforms will evolve to support this new disruptive technology? Yeah, I got into the field through my grad studies. I pursued theoretical physics at Dartmouth College for my PhD.

During that, I got exposed to quantum mechanics, followed by quantum algorithms. So as I developed through my PhD, I realized there was this upcoming field that could lead the foundation of the next wave of information technology revolution. I've been doing this for over 10 years now. And in the past 10 years, so much has changed.

When I started out my PhD, there were no freely available quantum computers. And then in 2016, IBM decided to put together their very small quantum computer on the cloud, made it available publicly. And then since then, there's so much has happened. As I said, when the IBM put together their first quantum computer, it was only two to five qubits.

And now regularly you see a couple of hundred qubits. And back then, there were only a couple of modalities like superconducting and maybe trapped ion. Now there's so many more different kinds of quantum computers. And there are more and more applications being explored, like where quantum computers could be useful.

And so much more has happened in this software as well. Now, Kunaf, one of QBraid's core ideas is helping developers kind of work across multiple quantum hardware providers and programming frameworks. In your opinion, why is interoperability such an important challenge in quantum computing today? Yeah, that's a wonderful question.

And for people who are not familiar with the field, it may help to know that right now, what hardware is most suited for building quantum computers is not even settled on yet. So for example, all of the classical computing, whether you're talking about CPUs or GPUs, they both are built on silicon transistors. So any kind of accelerators, any kinds of computers you see, phone, laptops, supercomputers, everything utilizes silicon. And you build transistors.

So all of the transistors lead to theoretical concept called bits. And you put together a whole bunch of transistors. Here you build these chips, which help you compute. In quantum computer, bit generalizes to something called a qubit.

And how you can build out qubits, you have so many different ways. You can actually build them using something called superconducting circuits, which utilizes somewhat similar semiconductor technology. But you can also build qubits out of neutral atoms. You can also build them out of trapped ions.

You can also build them out of photons. So which means if you are an end user utilizing quantum computers, you will be looking at different technologies. And this is a reason why you want to be trying all the different technologies available. And when you write your algorithms, you would want to run them on all the different kinds of hardware because you don't know which of the hardware is going to perform the best.

And this is the most important reason why the interoperability is super important. When you look at correct development, development, both from a standpoint of hardware providers and software frameworks. Where do you see the biggest barriers or fragmentation for developers trying to really build applications within these environments There are actually multiple challenges The field is super early right now where if you write an algorithm you kind of have to take care of the transpilation layer compilation layer, many times you yourself.

And this idea of also even just moving from one framework to another, that includes moving and swapping hardware, that is a big challenge, right? So you write your algorithm once, and then you run it through the various, let's say, one quantum computer pipeline. So you need to make sure that the algorithm is supported by that quantum computer throughout the various frameworks. And many times when you try out different algorithms, they have their own repositories, which are managed by certain companies that only support different hardware.

So you have to go zigzag to the end of the chain, which allows you to run on a quantum computer. So there's like a barrier at each level of the stack where your stack needs to be interoperatable at various different levels. And many times there are frameworks like Kubernetes SDK that we build that allows you to target various hardware level at a circuit level, but then more is needed on an algorithmic level where you can just specify the problem that you care about and then you can target at any level.

So that's one of the biggest challenge. And then another important thing that still exists, which is a nice layer of algorithms where you can specify problems and then it helps you break the problem into subtasks where each task is suited for a certain kind of processor and then that task is sent to qpu gpu cpu whichever is the most optimal for that task and then you need like a framework to combine those tasks and get the answer So there's another challenge, which is basically how do you distribute these subtasks on various kinds of different processors?

So a lot of exciting challenges. Definitely. And it's all new too, right? We're all figuring this out together.

But Qubrid really positions itself as a platform that bridges the gaps, right? That kind of brings things together in a unified way. From a product standpoint, what does it take to make quantum infrastructure easier for developers and researchers to easily use it? Yeah, it takes a lot of effort.

So the first thing that is required is you need seamless access to all the capabilities that any of the hardware companies provide, along with majority of the software that is available in the field. So what we strive to do is, because there are so many different quantum software companies, quantum hardware companies, we try to work with all of them to make sure that all of their software and hardware works out of the box. So what that means, if you're an end user, you can just come to the Qubrid platform and you will have super easy access to all the different hardware that are available and all the software available from various different providers.

So most of the entire industry is represented on Q-Brate. So we have to work very hard to make sure all the software and the hardware support is up to date. And then the next step we've been moving towards is, as I mentioned, the support of the classical compute that goes in conjunction with quantum computers. And this is where we've been working on something called Q-Grid OS.

For people who are not familiar, quantum computers right now, they're not perfect. So even to work on a specific problem, they actually end up consuming a lot of classical compute to work perfectly. So you need classical compute to correct errors that happen in quantum computers. And then there's other part of the algorithm that I mentioned before, certain tasks that you can run on CPUs or GPUs.

Now, how do you make both of these things come together in a seamless fashion? That end user can easily trigger, let's say, four to eight GPUs that can, let's say, two of the GPUs support quantum computers, and then the rest of the GPUs are available for other tasks. And all of those things need to be exposed in an abstract, swallowable function that any user can just call those functions and run their tasks seamlessly and get the answer. So it takes a lot of orchestration from all of these computes to make it happen.

And this is something that we've been actively working. The previous problem that I mentioned, supporting all the different software and hardware, this is something that we have a very good handle because we've been working on for the past five or six years. Now, I know that you're saying it's early days, but as more organizations start experimenting with quantum in their own environments, how will you see customers' expectations evolving from early research exploration towards something that might be inching closer to practical utilization?

Yeah I think as the field matures people are going to want things to be much more robust And that not only better performance from quantum processors but also better robustness from the entire deployment pipeline as well And that's where we come in. So once the hardware gets mature enough where you can run certain tasks and reliably count on it to address those tasks, the next step would be, okay, the entire pipeline needs to work very well. So then your software ends up getting standardized as well.

But because the quantum computers are not entirely there yet, the software stack has also been in flux. And I expect it will constantly evolve through the coming years. There's a lot of discussion around quantum that focuses on hardware breakthroughs, but developer tools, software and infrastructure are also super critical, as you just said, Kanovs. How do you see, though, that layer kind of shaping the pace of progress within the industry?

Yeah, software, I think, is quite critical because many of the breakthroughs for sure have come from hardware, but similar breakthroughs have also come from software. when both of those things combined delivered phenomenal gains. To give you an example, I think when the first ever evaluation of what size quantum computer it would take to break encryption or break Bitcoin, if you will. If I remember correctly, I think it was like a few billion quantum computers running for, I don't know, a couple of years.

and it was calculated, yeah, if you had this giant quantum computer, which people projected would take like another 50 years to build, then you would run it for like a couple of years and then you'll be able to break encryption. And then I think a few years later, more people, amazing people, they did many more algorithmic improvements. They brought it down to 20 million. And very recently, a couple of teams, I think as early as like last week, there were a couple of paper is that, hey, now we can do all of that with just 10,000 qubits.

All of those are algorithmic improvements along with a lot of software improvements. And what has been fascinating is even if you just look at the last week, there's two separate teams, one from Google and another from this new company that is building new neutral quantum computer using two different kinds of quantum computers and targeting the same problem. So now your software here needs to be able to target both of these kinds of quantum computers to leverage any improvement that these hardware deliver, right?

So this is going to be an evolving layer, but a lot more work needs to be put in to make sure this is super robust because as you move towards productionizing and actually using quantum computers to do tasks that enterprise relies on, this layer will constantly need more and more work to make sure that you're not making any mistakes and you're actually delivering the most performance you can get out of any quantum computer. Now, when developers first working with quantum technologies, what do you think surprises them most, either about the opportunities on the table or the practical limitations of today's systems?

There are actually so many things that most people get surprised by. So most people have the hardest time wrapping their heads around how a qubit could be zero or one or zero and one at the same time. And even if they roughly understand what's going on, there's always somebody going, it's actually not zero and one at the same time, it's more nuanced. And what is fascinating is this is something that most people coming into the field struggle with.

And then they spend 10 years and then they still struggle with what this actually means. It goes to the heart of quantum mechanics itself, like what is a wave function and so on. There's like a whole philosophical discussion around it. And another thing from the utility point of view that surprises a lot of people, on one hand, there is a talk of quantum computers breaking Bitcoin and encryption, right?

And then like, wow, these computers must be so powerful. And at the same time, they find it really hard to understand how come these computers are so fast and yet cannot do a single thing better than normal computers right now. So there isn't a single task right now where normal computers cannot outdo quantum computers, including encryption. Wow.

So breaking encryption, the whole thing depends on quantum computers being able to factor a really large number, which is incredibly hard for a classical computer. And yet to this day I think the record is like 21 Reliably breaking or factoring large number that a quantum computer can do I think that I could be off but my memory tells me it 21 which is far off from what we would need to do. And of course, there's like an argument that of course, the whole thing is about scaling, but to a late person just coming into the field, they find it incredibly surprising that on one hand, you're telling me these are the most powerful computers that we may have ever created, but at the same time, you cannot factor 21 hours as possible.

So there's a lot of nuance in understanding the capabilities of quantum computers and understanding what they can do well and where they are actually really terrible to use. And these are some of the things that most incomers struggle with, then get a better understanding. So looking ahead several years from now, what milestones would kind of signal, you know, that the quantum ecosystem has matured enough for developers to really just go in and take advantage of the stack? That's a good question.

I think of it in terms of two different sets of milestones. One is purely hardware. Once the error rates in quantum processors dip below a certain threshold, they become reliable enough to start doing various tasks. And below that, you can start trusting that, hey, this processor is doing what I wanted it to do.

Right now, there's like some finicky stuff where you have to actually post-process a lot of the results just to get a sense. Maybe it is doing things correctly, maybe it is not. Once it starts doing that, there will be two things that will happen. One, you will be sure that, okay, quantum computers are good enough to break encryption, which means you will need to update all of the encryption infrastructure that we use.

So that's one set of threshold, which is tracking the hardware and most important milestone would be error rates going below a certain threshold that you can be certain that you can throw any task at quantum computer and you can trust the result. Then the question becomes, parallelly, how fast can we come up with new applications that use those quantum computers, that use the power that these new kinds of computers provide us to do more useful tasks that provide economic value?

That, I think, is a small and open question. If you went back to 2005 or 2010 period, most people who were doing machine learning still were using CPU, GPU, and there might have been some tension which architecture is better. And as the time progressed, it became increasingly clear that machine learning was better suited for GPUs, right? And so now there is incredible promise in various different areas where theoretically it seems like quantum computers should be better.

But then you can be making a lot of progress, but you cannot predict that those certain algorithms while running on GPUs wouldn't also have incredible breakthroughs. breakthroughs, right? So you can deliver on all the promises on quantum computers, yet it just could so happen that people find new algorithms to do similar tasks better on GPUs. So only time will tell which of the applications ends up winning and whether we keep on finding new applications.

Intuitively, it makes sense that there are like certain tasks, which is designing new drug molecules for certain proteins and so on, where quantum mechanics comes into play. Those tasks intuitively should be better suited for quantum computers, but there's actually no proof that tells you that it should be incredibly hard for GPUs or CPUs. And all this conversation has been fascinating. I really appreciate your time today, and I'm sure that our audience is thrilled to hear from you.

One final question for you. how can the folks who are listening and Maureen engage with you and your team and continue the conversation and potentially engage around your software solutions? Yeah, absolutely. We are on LinkedIn X and if you ever have any questions, please reach out to us at contact.

cubere.com. Thanks so much, Kanabic and Janice. What a great episode.

This wraps another Data Insights episode in the arena. Thanks for being on both of you. Thank you. Thank you, Allison.

Thanks for joining Tech Arena. Subscribe and engage at our website, techarena.ai. All content is copyright by Tech Arena.