Error Corp: Shaping the Future of Error Control for Quantum Computers
Quantum computing promises to revolutionize numerous fields, from developing new materials and breaking encryption to solving complex optimization problems more efficiently than classical computers.
However, a significant hurdle remains: error correction. With every step of a quantum computation, errors accumulate, ultimately leading to inaccurate results. This limits today’s quantum computers to performing only shallow calculations with a few qubits and steps, significantly constraining their potential.
Error Corp pioneers a new approach to detect and correct quantum errors more efficiently, promising to reduce the number of qubits required for quantum error correction drastically. Founded by Dennis Lucarelli, Research Associate Professor of Physics at the American University, in early 2022, it has the potential to propel quantum computing beyond its current limitations and address some of the most challenging problems in science and technology.
Learn more from our interview with the founder and CEO, Dennis Lucarelli:
Why Did You Start Error Corp?
I had spent more than 15 years as a principal research scientist at The Johns Hopkins University Applied Physics Laboratory when I moved on to Sapienza University of Rome, taking a sabbatical to figure out what would come next.
Throughout my research, I have explored machine learning, autonomous systems, and especially quantum computing. I have been into quantum control theory since my PhD—it was my first love in research and has remained so until today. During the sabbatical, I spent time thinking about quantum algorithms, quantum machine learning, and quantum error correction—it’s fun to explore a new domain of physics and see what you can figure out on your own without asking other people or reading about their works. Just take some time to play around, make mistakes, and develop your own quantum algorithms.
As I was studying quantum error correction, I stumbled upon an idea that appeared to be a new way of thinking about a process called syndrome extraction. In quantum computing, you can’t measure and thus know a quantum state without destroying it. However, through partial, non-destructive measurements, you could still learn something about a quantum state and thereby detect where errors had occurred in a quantum computation.
It led me to explore these ideas around what is called syndrome compression. Eventually, it led me to found Error Corp, with the goal of reducing quantum errors in practice and unlocking the potential of quantum computers.
What is Quantum Error Correction?
Quantum mechanics is a fundamental theory that governs nature on tiny length scales. Once quantum systems interact with their environment, they lose their quantum properties, which is why our everyday experience isn’t quantum. Vice versa, we need to isolate quantum systems thoroughly to harness their quantum effects for computing.
Quantum states are very sensitive and can collapse when interacting with their environment. Think of the environment as another little quantum system sitting beside a quantum computer and smearing out its information. If its influence becomes too great, it destroys the quantum state of the quantum computer.
The crux is that you can’t fully avoid environmental influences. Quite the opposite, you must interact with the quantum computer whenever you input or output data or perform an operation. Even detecting quantum errors requires interacting with the quantum computer, introducing additional errors. It’s like you’re using noise-cancelling headphones that make a lot of noise themselves. Every operation on a quantum computer introduces noise.
Now, one important question is what we can do and how far we can go with noisy, error-prone qubits. The other one is about the best way to mitigate and correct errors.
In classical computing, error correction works by fixing unintended bit flips. Quantum computers face another challenge as they must deal not only with bit-flip errors but also phase errors, which do not occur in classical systems. Every error in a quantum computer is a combination of these two basic errors, so if we can correct both, we can correct any quantum error and drive the quantum computer back into the desired state.
Error Corp provides the infrastructure for performing quantum error correction at scale. Our approach comes in two varieties: syndrome compression and syndrome pooling, which are both ways of performing syndrome extraction differently. Ultimately, they allow us to detect and correct quantum errors efficiently while requiring much fewer qubits than other methods.
How Do Syndrome Extraction and Pooling Work?
Syndrome extraction means measuring certain properties of the quantum state through a partial, non-destructive measurement, which indicates whether an error has occurred and, if so, what kind of error.
The textbook approach to quantum error correction focuses very much on single logical qubits as perfect, self-correcting units. Syndrome pooling takes a different approach by spreading error correction across many qubits. It involves combining the results of syndrome measurements across several logical qubits to obtain a more efficient method to estimate the errors present.
By taking a global, collective point of view, we can utilize quantum resources better and fix errors more efficiently: There’s even a German saying for ‘the whole thing’ called ‘Gestalt,’ so the idea is to build error correction schemes from the Gestalt of a quantum computer rather than individual qubits.
The great thing about our approach is that it’s compatible with any basic quantum error correction code—methods for detecting and correcting quantum errors without collapsing the quantum state.
The most well-known quantum error correction code is the surface code. It has some beautiful properties but also some disadvantages, the main one being that it encodes only one logical qubit at a time, and requires many additional qubits. It’s not very efficient. More like a dog trying to catch its tail: You add qubits to correct the errors of other qubits, but these additional qubits are also error-prone, so you get more errors. The amount of computational power you get is dwarfed by the computational effort it takes to correct all the errors and maintain the computation.
But since we can work with any quantum error correcting code, we can use more efficient ones, such as other stabilizer codes like low-density parity checking codes. In addition, we can spread error correction across many qubits, not just by looking at single qubits but by abstracting many physical qubits into fewer syndrome qubits that produce fewer errors.
What Error Rate Will Be ‘Good Enough’?
That’s one of the big questions today—when should you start scaling the number of qubits in a quantum computer versus spending more time perfecting the qubits, the fidelity of quantum operations, and its overall architecture? Our view is that we need to do anything it takes to improve the fidelity, i.e., how accurately a quantum operation can produce a desired entangled state, as this improves so many other things downstream.
If you have two nines of fidelity, i.e., 99%, one way forward is to use lots of qubits for error correction. It’s inefficient, but as long as the overall number of errors goes down as you add more qubits, you can scale your way out of the problem.
Still, reducing the error rate and quality of qubits is beneficial. We may never achieve a physical error rate as low as 10-15, one error in a quadrillion operations, which would be comparable to modern-day transistors. But we need to get to the point where, as we add more qubits, they correct more errors than they introduce in addition. Getting there will be a gradual process, but it’s very exciting to see the progress in detecting and correcting quantum errors, which is a prerequisite to reaching quantum advantage and unlocking utility for commercial applications.
How Did You Evaluate the Opportunity?
The key challenge is that as a quantum error detection and correction solution, we sit in the middle of the stack, so we need information from the layers surrounding us. Still, we have a great opportunity to make a difference, as we have the flexibility to work with any hardware provider and for any use case. We don’t dictate how to build logical qubits; we just aim to provide the best fault-tolerance design and software infrastructure for quantum error correction at scale.
When I talk to customers, they’re always very excited about the prospect that we can accelerate their roadmap: reducing the number of qubits needed to implement quantum algorithms will enable those algorithms to run on current quantum computers earlier. If we can build trust, then our commercialization prospects are quite good, but we need to show our customers that it works.
As the quantum industry builds larger and more capable quantum machines, quantum error correction becomes more relevant to move from academic research projects to commercially relevant use cases. We might need it much sooner than we thought to unlock all the applications we’d like to see.
What Applications Are You Most Excited About?
A quantum computer can simulate other quantum systems better than any classical computer ever can. That’s why one of the most exciting applications of quantum computers is understanding other quantum systems. In the short term, this helps researchers discover and verify new aspects of physics. In the medium to long term, it will help with materials development and building digital twins for product development.
An analogy I like is how we use high-performance computing today to design new materials and products. If you’re an airplane manufacturer designing a new wing, you don’t start building a model and putting materials into shape immediately. Today, you build a digital twin of the wing and optimize its design before you start building. These simulations are challenging but still much easier than building many hardware prototypes. Quantum computing will dramatically increase our ability to build accurate quantum digital twins.
We know some applications already, but many more will become possible once we put quantum computers in the hands of the next generation of scientists and engineers. There’s a great opportunity to benefit society, for example, through the discovery of new materials and help scientists better understand quantum phenomena and, thus, our universe at the smallest length scales.
What Advice Would You Give a Fellow Deep Tech Founder?
First, if you have a startup idea that doesn’t go away, and you keep coming back to it, you should listen to it. It takes a while to develop the core idea, but if it doesn’t go away and you see potential in pursuing it, it’s time to give it a go. Trust your gut and listen to the idea that is trying to manifest itself in the universe.
You’ll need to be persistent and deal with a ton of uncertainty. That’s just the nature of a startup, and you’ll have to deal with it. You won’t have answers to all the questions. Your idea will develop and change. But you will also do things that you previously thought were impossible.
Then, find the way that is the best fit to build your startup. We’re fully grant-funded so far, as things always take longer than expected, especially in research-heavy startups. However, there comes a point at which grants can’t cover all aspects of a commercial startup’s growth. We’re now reaching specific scientific milestones that point toward commercial promise, making it a good time to consider raising venture funds.