By Dr. Chris Langer

In the rapidly advancing world of quantum computing, to be a leader means not just keeping pace with innovation but driving it forward. It means setting new standards that shape the future of quantum computing performance. A recent independent study comparing 19 quantum processing units (QPUs) on the market today has validated what we’ve long known to be true: Quantinuum’s systems are the undisputed leaders in performance.

The Benchmarking Study

A comprehensive study conducted by a joint team from the Jülich Supercomputing Centre, AIDAS, RWTH Aachen University, and Purdue University compared QPUs from leading companies like IBM, Rigetti, and IonQ, evaluating how well each executed the Quantum Approximate Optimization Algorithm (QAOA), a widely used algorithm that provides a system level measure of performance. After thorough examination, the study concluded that:

“...the performance of quantinuum H1-1 and H2-1 is superior to that of the other QPUs.”

Quantinuum emerged as the clear leader, particularly in full connectivity, the most critical category for solving real-world optimization problems. Full connectivity is a huge comparative advantage, offering more computational power and more flexibility in both error correction and algorithmic design. Our dominance in full connectivity—unattainable for platforms with natively limited connectivity—underscores why we are the partner of choice in quantum computing.

Leading Across the Board

We take benchmarking seriously at Quantinuum. We lead in nearly every industry benchmark, from best-in-class gate fidelities to a 4000x lead in quantum volume, delivering top performance to our customers.

Our Quantum Charged-coupled Device (QCCD) architecture has been the foundation of our success, delivering consistent performance gains year-over-year. Unlike other architectures, QCCD offers all-to-all connectivity, world-record fidelities, and advanced features like real-time decoding. Altogether, it’s clear we have superior performance metrics across the board.

While many claim to be the best, we have the data to prove it. This table breaks down industry benchmarks, using the leading commercial spec for each quantum computing architecture.

These metrics are the key to our success. They demonstrate why Quantinuum is the only company delivering meaningful results to customers at a scale beyond classical simulation limits.

Our progress builds upon a series of Quantinuum’s technology breakthroughs, including the creation of the most reliable and highest-quality logical qubits, as well as solving the key scalability challenge associated with ion-trap quantum computers — culminating in a commercial system with greater than 99.9% two-qubit gate fidelity.

From our groundbreaking progress with System Model H2 to advances in quantum teleportation and solving the wiring problem, we’re taking major steps to tackle the challenges our whole industry faces, like execution speed and circuit depth. Advancements in parallel gate execution, faster ion transport, and high-rate quantum error correction (QEC) are just a few ways we’re maintaining our lead far ahead of the competition.

This commitment to excellence ensures that we not only meet but exceed expectations, setting the bar for reliability, innovation, and transformative quantum solutions. 

Onward and Upward

To bring it back to the opening message: to be a leader means not just keeping pace with innovation but driving it forward. It means setting new standards that shape the future of quantum computing performance.

We are just months away from launching Quantinuum’s next generation system, Helios, which will be one trillion times more powerful than H2. By 2027, Quantinuum will launch the industry’s first 100-logical-qubit system, featuring best-in-class error rates, and we are on track to deliver fault-tolerant computation on hundreds of logical qubits by the end of the decade. 

The evidence speaks for itself: Quantinuum is setting the standard in quantum computing. Our unrivaled specs, proven performance, and commitment to innovation make us the partner of choice for those serious about unlocking value with quantum computing. Quantinuum is committed to doing the hard work required to continue setting the standard and delivering on our promises. This is Quantinuum. This is leadership.

Dr. Chris Langer is a Fellow, a key inventor and architect for the Quantinuum hardware, and serves as an advisor to the CEO.

_______________________________________
‍

Citations from Benchmarking Table

¹ Quantinuum. System Model H2. Quantinuum, https://www.quantinuum.com/products-solutions/quantinuum-systems/system-model-h2

² IBM. Quantum Services & Resources. IBM Quantum, https://quantum.ibm.com/services/resources

³ Quantinuum. System Model H1. Quantinuum, https://www.quantinuum.com/products-solutions/quantinuum-systems/system-model-h1

⁴ Google Quantum AI. Willow Spec Sheet. Google, https://quantumai.google/static/site-assets/downloads/willow-spec-sheet.pdf

⁵ Sales Rodriguez, P., et al. "Experimental demonstration of logical magic state distillation." arXiv, 19 Dec 2024, https://arxiv.org/pdf/2412.15165

⁶ Quantinuum. H1 Product Data Sheet. Quantinuum, https://docs.quantinuum.com/systems/data_sheets/Quantinuum%20H1%20Product%20Data%20Sheet.pdf

⁷ Google Quantum AI. Willow Spec Sheet. Google, https://quantumai.google/static/site-assets/downloads/willow-spec-sheet.pdf

⁸ Sales Rodriguez, P., et al. "Experimental demonstration of logical magic state distillation." arXiv, 19 Dec 2024, https://arxiv.org/pdf/2412.15165

⁹ Quantinuum. H2 Product Data Sheet. Quantinuum, https://docs.quantinuum.com/systems/data_sQuantinuum. H2 Product Data Sheet. Quantinuum,heets/Quantinuum%20H2%20Product%20Data%20Sheet.pdf

¹⁰ Google Quantum AI. Willow Spec Sheet. Google, https://quantumai.google/static/site-assets/downloads/willow-spec-sheet.pdf

¹¹ Sales Rodriguez, P., et al. "Experimental demonstration of logical magic state distillation." arXiv, 19 Dec 2024, https://arxiv.org/pdf/2412.15165

¹² Moses, S. A., et al. "A Race-Track Trapped-Ion Quantum Processor." Physical Review X, vol. 13, no. 4, 2023, https://journals.aps.org/prx/pdf/10.1103/PhysRevX.13.041052

¹³ Google Quantum AI and Collaborators. "Quantum Error Correction Below the Surface Code Threshold." Nature, vol. 638, 2024, https://www.nature.com/articles/s41586-024-08449-y

¹⁴ Bluvstein, Dolev, et al. "Logical Quantum Processor Based on Reconfigurable Atom Arrays." Nature, vol. 626, 2023, https://www.nature.com/articles/s41586-023-06927-3

¹⁵ DeCross, Matthew, et al. "The Computational Power of Random Quantum Circuits in Arbitrary Geometries." arXiv, Published on 21 June 2024, hhttps://arxiv.org/pdf/2406.02501

¹⁶ Montanez-Barrera, J. A., et al. "Evaluating the Performance of Quantum Process Units at Large Width and Depth." arXiv, 10 Feb. 2025, https://arxiv.org/pdf/2502.06471

¹⁷ Evered, Simon J., et al. "High-Fidelity Parallel Entangling Gates on a Neutral-Atom Quantum Computer." Nature, vol. 622, 2023, https://www.nature.com/articles/s41586-023-06481-y

¹⁸ Ryan-Anderson, C., et al. "Realization of Real-Time Fault-Tolerant Quantum Error Correction." Physical Review X, vol. 11, no. 4, 2021, https://journals.aps.org/prx/abstract/10.1103/PhysRevX.11.041058

¹⁹ Carrera Vazquez, Almudena, et al. "Scaling Quantum Computing with Dynamic Circuits." arXiv, 27 Feb. 2024, https://arxiv.org/html/2402.17833v1

²⁰ Moses, S.A.,, et al. "A Race Track Trapped-Ion Quantum Processor." arXiv, 16 May 2023, https://arxiv.org/pdf/2305.03828

²¹ Garcia Almeida, D., Ferris, K., Knanazawa, N., Johnson, B., Davis, R. "New fractional gates reduce circuit depth for utility-scale workloads." IBM Quantum Blog, IBM, 18 Nov. 2020, https://www.ibm.com/quantum/blog/fractional-gates

²² Ryan-Anderson, C., et al. "Realization of Real-Time Fault-Tolerant Quantum Error Correction." arXiv, 15 July 2021, https://arxiv.org/pdf/2107.07505

²³ Google Quantum AI and Collaborators. “Quantum error correction below the surface code threshold.” arXiv, 24 Aug. 2024, https://arxiv.org/pdf/2408.13687v1

The 2025 Joint March Meeting and April Meeting — referred to as the APS Global Physics Summit — is the largest physics research conference in the world, uniting 14,000 scientific community members across all disciplines of physics.  

The Quantinuum team is looking forward to participating in this year’s conference to showcase our latest advancements in quantum technology. Find us throughout the week at the below sessions and visit us at Booth 1001.

Join these sessions to discover how Quantinuum is advancing quantum computing

T11: Quantum Error Correction
Speaker: Natalie Brown
Date: Sunday, March 16th
Time: 8:00 – 8:12am
Location: Anaheim Convention Center, 261B (Level 2)

The computational power of random quantum circuits in arbitrary geometries
Session MAR-F34: Near-Term Quantum Resource Reduction and Random Circuits
Speaker: Matthew DeCross
Date: Tuesday, March 18th
Time: 8:00 – 8:12am
Location: Anaheim Convention Center, 256A (Level 2)

Topological Order from Measurements and Feed-Forward on a Trapped Ion Quantum Computer
Session MAR-F14: Realizing Topological States on Quantum Hardware
Speaker: Henrik Dreyer
Date: Tuesday, March 18th
Time: 9:12 – 9:48am
Location: Anaheim Convention Center, 158 (Level 1)

Trotter error time scaling separation via commutant decomposition
Session MAR-F34: Near-Term Quantum Resource Reduction and Random Circuits
Speaker: Yi-Hsiang Chen (Quantinuum)
Date: Tuesday, March 18th
Time: 10:00 – 10:12am
Location: Anaheim Convention Center, 256A (Level 2)

Squared overlap calculations with linear combination of unitaries
Session MAR-J35: Circuit Optimization and Compilation
Speaker: Michelle Wynne Sze
Date: Tuesday, March 18th
Time: 4:36 – 4:48pm
Location: Anaheim Convention Center, 256B (Level 2)

High-precision quantum phase estimation on a trapped-ion quantum computer
Session MAR-L16: Quantum Simulation for Quantum Chemistry
Speaker: Andrew Tranter
Date: Wednesday, March 19th
Time: 9:48 – 10:00am
Location: Anaheim Convention Center, 160 (Level 1)

Robustness of near-thermal dynamics on digital quantum computers
Session MAR-L16: Quantum Simulation for Quantum Chemistry
Speaker: Eli Chertkov
Date: Wednesday, March 19th
Time: 10:12 – 10:24am
Location: Anaheim Convention Center, 160 (Level 1)

Floquet prethermalization on a digital quantum computer
Session MAR-Q09: Quantum Simulation of Condensed Matter Physics
Speaker: Reza Haghshenas
Date: Thursday, March 20th
Time: 10:00 – 10:12am
Location: Anaheim Convention Center, 204C (Level 2)

Teleportation of a Logical Qubit on a Trapped-ion Quantum Computer
Session MAR-S11: Advances in QEC Experiments
Speaker: Ciaran Ryan-Anderson
Date: Thursday, March 20th
Time: 11:30 – 12:06pm
Location: Anaheim Convention Center, 155 (Level 1)

*All times in Pacific Standard Time

A team from Quantinuum and the University of Freiburg found that quantum computers outperform classical for a workhorse calculation often used in accelerators like the Large Hadron Collider (LHC) at CERN.

Quantinuum’s Ifan Williams worked with the University of Freiburg’s Mathieu Pellen to tackle a pernicious problem in accelerator physics: calculating “cross sections”. Together, they developed a general, scalable approach to calculating cross sections that offers a quadratic speed-up compared to its classical counterpart.

A “cross-section” relates to the probability of a certain interaction happening. Scientists who do experiments in particle accelerators compare real measurements with theoretical cross-section calculations (predictions), using the agreement (or disagreement) to reason about the nature of our universe. 

Generally, scientists run Monte Carlo simulations to make their theoretical predictions. Monte Carlo simulations are currently the biggest computational bottleneck in experimental high-energy physics (HEP), costing enormous CPU resources, which will only grow larger as new experiments come online.  

It’s hard to put a specific number on exactly how costly calculations like this are, but we can say that probing fundamental physics at the LHC probably uses roughly 10 billion CPUH/year for data treatment, simulations, and theory predictions. Knowing that the theory predictions represent approximately 15-25% of this total, putting even a 10% dent in this number would be a massive change.

The collaborators used Quantinuum’s Quantum Monte Carlo integration (QMCI) engine to solve the same problem. Their work is the first published general methodology for performing cross-section calculations in HEP using quantum integration.

Importantly, the team’s methodology is potentially extendable to the problem sizes needed for real-world HEP cross-section calculations (currently done classically). Overall, this work establishes a solid foundation for performing such computations on a quantum computer in the future.

The Large Hadron Collider, the world’s biggest particle accelerator, generates a billion collisions each second, far more data than can be computationally analyzed. Planned future experiments are expected to generate even more. Quantum computers are also accelerating. Quantinuum’s latest H2 System became the highest performing commercially available system in the world when it was launched. When it was upgraded in 2024, it became the first quantum computer that cannot be exactly simulated by any classical computer. Our next generation Helios, on schedule to launch in 2025, will encode at least a trillion times more information than the H2—this is the power of exponential growth.  

We can’t wait to see what’s next with quantum computing and high-energy physics.

The Quantinuum team is looking forward to participating in this year’s SCAsia conference from March 10^th – 13^th in Singapore. Meet our team at Booth B2 to discover how Quantinuum is bridging the gap between quantum computing and high-performance compute with leading industry partners.

Our team will be participating in workshops and presenting at the keynote and plenary sessions to showcase our quantum computing technologies. Join us at the below sessions:

Monday, March 10^th, 1:30 – 2:30pm

Workshop: Accelerating Quantum Supercomputing: CUDA-Q Tutorial across Multiple Quantum Platforms
Location: Room P10 – Peony Jr 4512 (Level 4)

This workshop will explore the seamless integration of classical and quantum resources for quantum-accelerated supercomputing. Join Kentaro Yamamoto and Enrico Rinaldi, Lead R&D Scientists at Quantinuum, for an Introduction to our  integrated full-stack for quantum computing, Quantum Phase Estimation (QPE) for solving quantum chemistry problems, and a demonstration of a QPE algorithm with CUDA-Q on Quantinuum Systems. If you're interested in access to our quantum computers and emulator for use on the CUDA-Q platform, register here.

Tuesday, March 11^th, 11:00 – 11:30pm

Keynote: Quantum Computing: A Transformative Force for Singapore's Regional Economy
Location: Melati Ballroom (Level 4)

Quantum Computing is no longer a distant promise; it has arrived and is poised to revolutionize several economies. Join our President and CEO, Dr. Rajeeb Hazra, to discover how Quantinuum’s approach to Quantum Generative AI is driving breakthroughs in applications which hold significant relevance for Singapore, in fields like chemistry, computational biology, and finance. Additionally, Raj will discuss the challenges and opportunities of adopting quantum solutions from both technical and business perspectives, emphasizing the importance of collaboration to build quantum applications that integrate the best of quantum and AI.

Tuesday, March 11^th, 5:40 – 6:00pm

Industry Breakout Track: Transformative value of Quantum and AI: bringing meaningful insights for critical applications today
Location: Room L1 – Lotus Jr (Level 4)

The ability to solve classically intractable problems defines the transformative value of quantum computing, offering new tools to redefine industries and address complex humanity challenges. In this session with Dr. Elvira Shishenina, Senior Director of Strategic Initiatives, discover how Quantinuum’s hardware is leading the way in achieving early fault-tolerance, marking a significant step forward in computational capabilities. By integrating quantum technology with AI and high-performance computing, we are building systems designed to address real-world issues with efficiency, precision and scale. This approach empowers critical applications from hydrogen fuel cells and carbon capture to precision medicine, food security, and cybersecurity, providing meaningful insights at a commercial level today.

Wednesday, March 12^th, 4:40 – 5:00pm

Hybrid Quantum Classical Computing Track: Quantifying Quantum Advantage with an End-to-End Quantum Algorithm for the Jones Polynomial
Location: Room O3 – Orchid Jr 4211-2 (Level 4)

Join Konstantinos Meichanetzidis, Head of Scientific Product Development, for this presentation on an end-to-end reconfigurable algorithmic pipeline for solving a famous problem in knot theory using a noisy digital quantum computer. Specifically, they estimate the value of the Jones polynomial at the fifth root of unity within additive error for any input link, i.e. a closed braid. This problem is DQC1-complete for Markov-closed braids and BQP-complete for Plat-closed braids, and we accommodate both versions of the problem. In their research, they demonstrate our quantum algorithm on Quantinuum’s H2 quantum computer and show the effect of problem-tailored error-mitigation techniques. Further, leveraging that the Jones polynomial is a link invariant, they construct an efficiently verifiable benchmark to characterize the effect of noise present in a given quantum processor. In parallel, they implement and benchmark the state-of-the-art tensor-network-based classical algorithms.The practical tools provided in the work presented will allow for precise resource estimation to identify near-term quantum advantage for a meaningful quantum-native problem in knot theory.

Thursday, March 13^th, 11:00 – 11:30pm

Industry Plenary: Quantum Heuristics: From Worst Case to Practice
Location: Melati Ballroom (Level 4)

Which problems allow for a quantum speedup, and which do not? This question lies at the heart of quantum information processing. Providing a definitive answer is challenging, as it connects deeply to unresolved questions in complexity theory. To make progress, complexity theory relies on conjectures such as P≠NP and the Strong Exponential Time Hypothesis, which suggest that for many computational problems, we have discovered algorithms that are asymptotically close to optimal in the worst case. In this talk, Professor Harry Buhrman, Chief Scientist for Algorithms and Innovation, will explore the landscape from both theoretical and practical perspectives. On the theoretical side, I will introduce the concept of “queasy instances”—problem instances that are quantum-easy but classically hard (classically queasy). On the practical side, I will discuss how these insights connect to advancements in quantum hardware development and co-design.

*All times in Singapore Standard Time

BY HARRY BUHRMAN

Quantum computing continues to push the boundaries of what is computationally possible. A new study by Marcello Benedetti, Harry Buhrman, and Jordi Weggemans introduces Complement Sampling, a problem that highlights a dramatic separation between quantum and classical sample complexity. This work provides a robust demonstration of quantum advantage in a way that is not only provable but also feasible on near-term quantum devices.

The Complement Sampling Problem

Imagine a universe of N = 2ⁿ elements, from which a subset S of size K is drawn uniformly at random. The challenge is to sample from the complement S̅ ‍without explicitly knowing S, but having access to samples of S. Classically, solving this problem requires roughly K samples, as the best a classical algorithm can do is guess at random after observing only some of the elements of S.

To better understand this, consider a small example. Suppose N = 8, meaning our universe consists of the numbers {0,1,2,3,4,5,6,7}. If a subset S of size K = 4 is drawn at random—say {1,3,5,7}—the goal is to sample from the complement  S̅, which consists of {0,2,4,6}. A classical algorithm would need to collect and verify enough samples from S before it could infer what S̅ might be. However, a quantum algorithm can use a single superposition state over S (a quantum sample) to instantly generate a sample from S̅, eliminating the need for iterative searching.

Why This Matters: Quantum Advantage in Sample Complexity

Quantum advantage is often discussed in terms of computational speedups, such as those achieved by Shor’s algorithm for factoring large numbers. However, quantum resources provide advantages beyond time efficiency—they also affect how data is accessed, stored, and processed.

Complement Sampling fits into the category of sample complexity problems, where the goal is to minimize the number of samples needed to solve a problem. The authors prove that their quantum approach not only outperforms classical methods but does so in a way that is:

• Provable: It provides rigorous lower bounds on classical sample complexity, demonstrating an exponential separation.
• Verifiable: The correctness of the output of the sampler can be efficiently checked classically.
• NISQable: The quantum circuit required is shallow and feasible for Noisy Intermediate-Scale Quantum (NISQ) devices.

How the Quantum Algorithm Works

At its core, the quantum approach to Complement Sampling relies on the ability to perform a perfect swap between a subset S and its complement S̅. The method draws inspiration from a construction by Aaronson, Atia, and Susskind, which links state distinguishability to state swapping. The quantum algorithm:

1. Uses a unitary transformation that maps the quantum sample |S⟩ to |S̅⟩ with high probability.
2. For K = N/2, the algorithm works perfectly outputting an element from S̅ with probability 1.
3. For other values of K, a probabilistic zero-error approach is used, ensuring correctness while reducing success probability.

This is made possible by quantum interference and superposition, allowing a quantum computer to manipulate distributions in ways that classical systems fundamentally cannot.

Classical Hardness and Cryptographic Implications

A crucial aspect of this work is its robustness. The authors prove that even for subsets generated using strong pseudorandom permutations, the problem remains hard for classical algorithms. This means that classical computers cannot efficiently solve Complement Sampling even with structured input distributions—an important consideration for real-world applications.

This robustness suggests potential applications in cryptography, where generating samples from complements could be useful in privacy-preserving protocols and quantum-secure verification methods.

Towards an Experimental Demonstration

Unlike some quantum advantage demonstrations that are difficult to verify classically (such as the random circuit sampling experiment), Complement Sampling is designed to be verifiable. The authors propose an interactive quantum versus classical game:

1. A referee provides a quantum player with quantum samples from S.
2. The player must return a sample from S̅
3. A classical player, given the same number of classical samples, attempts to do the same.

While the classical player must resort to random guessing, the quantum player can leverage the swap algorithm to succeed with near certainty. Running such an experiment on NISQ hardware could serve as a practical demonstration of quantum advantage in a sample complexity setting.

Future Directions

This research raises exciting new questions:

• Can Complement Sampling be extended to more general probability distributions?
• Are there cryptographic protocols that can directly leverage this advantage?
• How well does the quantum algorithm perform in real-world noisy conditions?

With its blend of theoretical depth and experimental feasibility, Complement Sampling provides a compelling new frontier for demonstrating the power of quantum computing.

Conclusion

Complement Sampling represents one of the cleanest demonstrations of quantum advantage in a practical, verifiable, and NISQ-friendly setting. By leveraging quantum information processing in ways that classical computers fundamentally cannot, this work strengthens the case for near-term quantum technologies and their impact on computational complexity, cryptography, and beyond.

For those interested in the full details, the paper provides rigorous proofs, circuit designs, and further insights into the nature of quantum sample complexity. As quantum computing continues to evolve, Complement Sampling may serve as a cornerstone for future experimental demonstrations of quantum supremacy.

We have commenced work on the experiment – watch this space!

Today’s LLMs are often impressive by past standards – but they are far from perfect

Quietly, and determinedly since 2019, we’ve been working on Generative Quantum AI. Our early focus on building natively quantum systems for machine learning has benefitted from and been accelerated by access to the world’s most powerful quantum computers, and quantum computers that cannot be classically simulated.

Our work additionally benefits from being very close to our Helios generation quantum computer, built in Colorado, USA. Helios is 1 trillion times more powerful than our H2 System, which is already significantly more advanced than all other quantum computers available.

While tools like ChatGPT have already made a profound impact on society, a critical limitation to their broader industrial and enterprise use has become clear. Classical large language models (LLMs) are computational behemoths, prohibitively huge and expensive to train, and prone to errors that damage their credibility.

Training models like ChatGPT requires processing vast datasets with billions, even trillions, of parameters. This demands immense computational power, often spread across thousands of GPUs or specialized hardware accelerators. The environmental cost is staggering—simply training GPT-3, for instance, consumed nearly 1,300 megawatt-hours of electricity, equivalent to the annual energy use of 130 average U.S. homes.

This doesn’t account for the ongoing operational costs of running these models, which remain high with every query. 

Despite these challenges, the push to develop ever-larger models shows no signs of slowing down.

Enter quantum computing. Quantum technology offers a more sustainable, efficient, and high-performance solution—one that will fundamentally reshape AI, dramatically lowering costs and increasing scalability, while overcoming the limitations of today's classical systems. 

Quantum Natural Language Processing: A New Frontier

At Quantinuum we have been maniacally focused on “rebuilding” machine learning (ML) techniques for Natural Language Processing (NLP) using quantum computers. 

Our research team has worked on translating key innovations in natural language processing — such as word embeddings, recurrent neural networks, and transformers — into the quantum realm. The ultimate goal is not merely to port existing classical techniques onto quantum computers but to reimagine these methods in ways that take full advantage of the unique features of quantum computers.

We have a deep bench working on this. Our Head of AI, Dr. Steve Clark, previously spent 14 years as a faculty member at Oxford and Cambridge, and over 4 years as a Senior Staff Research Scientist at DeepMind in London. He works closely with Dr. Konstantinos Meichanetzidis, who is our Head of Scientific Product Development and who has been working for years at the intersection of quantum many-body physics, quantum computing, theoretical computer science, and artificial intelligence.

A critical element of the team’s approach to this project is avoiding the temptation to simply “copy-paste”, i.e. taking the math from a classical version and directly implementing that on a quantum computer. 

This is motivated by the fact that quantum systems are fundamentally different from classical systems: their ability to leverage quantum phenomena like entanglement and interference ultimately changes the rules of computation. By ensuring these new models are properly mapped onto the quantum architecture, we are best poised to benefit from quantum computing’s unique advantages. 

These advantages are not so far in the future as we once imagined – partially driven by our accelerating pace of development in hardware and quantum error correction.

Making computers “talk”- a short history

The ultimate problem of making a computer understand a human language isn’t unlike trying to learn a new language yourself – you must hear/read/speak lots of examples, memorize lots of rules and their exceptions, memorize words and their meanings, and so on. However, it’s more complicated than that when the “brain” is a computer. Computers naturally speak their native languages very well, where everything from machine code to Python has a meaningful structure and set of rules. 

In contrast, “natural” (human) language is very different from the strict compliance of computer languages: things like idioms confound any sense of structure, humor and poetry play with semantics in creative ways, and the language itself is always evolving. Still, people have been considering this problem since the 1950’s (Turing’s original “test” of intelligence involves the automated interpretation and generation of natural language).

Up until the 1980s, most natural language processing systems were based on complex sets of hand-written rules. Starting in the late 1980s, however, there was a revolution in natural language processing with the introduction of machine learning algorithms for language processing. 

Initial ML approaches were largely “statistical”: by analyzing large amounts of text data, one can identify patterns and probabilities. There were notable successes in translation (like translating French into English), and the birth of the web led to more innovations in learning from and handling big data.

What many consider “modern” NLP was born in the late 2000’s, when expanded compute power and larger datasets enabled practical use of neural networks. Being mathematical models, neural networks are “built” out of the tools of mathematics; specifically linear algebra and calculus. 

Building a neural network, then, means finding ways to manipulate language using the tools of linear algebra and calculus. This means representing words and sentences as vectors and matrices, developing tools to manipulate them, and so on. This is precisely the path that researchers in classical NLP have been following for the past 15 years, and the path that our team is now speedrunning in the quantum case.

Quantum Word Embeddings: A Complex Twist

The first major breakthrough in neural NLP came roughly a decade ago, when vector representations of words were developed, using the frameworks known as Word2Vec and GloVe (Global Vectors for Word Representation). In a recent paper, our team, including Carys Harvey and Douglas Brown, demonstrated how to do this in quantum NLP models – with a crucial twist. Instead of embedding words as real-valued vectors (as in the classical case), the team built it to work with complex-valued vectors.

In quantum mechanics, the state of a physical system is represented by a vector residing in a complex vector space, called a Hilbert space. By embedding words as complex vectors, we are able to map language into parameterized quantum circuits, and ultimately the qubits in our processor. This is a major advance that was largely under appreciated by the AI community but which is now rapidly gaining interest.

Using complex-valued word embeddings for QNLP means that from the bottom-up we are working with something fundamentally different. This different “geometry” may provide advantage in any number of areas: natural language has a rich probabilistic and hierarchical structure that may very well benefit from the richer representation of complex numbers.

The Quantum Recurrent Neural Network (RNN)

Another breakthrough comes from the development of quantum recurrent neural networks (RNNs). RNNs are commonly used in classical NLP to handle tasks such as text classification and language modeling. 

Our team, including Dr. Wenduan Xu, Douglas Brown, and Dr. Gabriel Matos, implemented a quantum version of the RNN using parameterized quantum circuits (PQCs). PQCs allow for hybrid quantum-classical computation, where quantum circuits process information and classical computers optimize the parameters controlling the quantum system.

In a recent experiment, the team used their quantum RNN to perform a standard NLP task: classifying movie reviews from Rotten Tomatoes as positive or negative. Remarkably, the quantum RNN performed as well as classical RNNs, GRUs, and LSTMs, using only four qubits. This result is notable for two reasons: it shows that quantum models can achieve competitive performance using a much smaller vector space, and it demonstrates the potential for significant energy savings in the future of AI.

In a similar experiment, our team partnered with Amgen to use PQCs for peptide classification, which is a standard task in computational biology. Working on the Quantinuum System Model H1, the joint team performed sequence classification (used in the design of therapeutic proteins), and they found competitive performance with classical baselines of a similar scale. This work was our first proof-of-concept application of near-term quantum computing to a task critical to the design of therapeutic proteins, and helped us to elucidate the route toward larger-scale applications in this and related fields, in line with our hardware development roadmap.

Quantum Transformers - The Next Big Leap

Transformers, the architecture behind models like GPT-3, have revolutionized NLP by enabling massive parallelism and state-of-the-art performance in tasks such as language modeling and translation. However, transformers are designed to take advantage of the parallelism provided by GPUs, something quantum computers do not yet do in the same way.

In response, our team, including Nikhil Khatri and Dr. Gabriel Matos, introduced “Quixer”, a quantum transformer model tailored specifically for quantum architectures. 

By using quantum algorithmic primitives, Quixer is optimized for quantum hardware, making it highly qubit efficient. In a recent study, the team applied Quixer to a realistic language modeling task and achieved results competitive with classical transformer models trained on the same data. 

This is an incredible milestone achievement in and of itself. 

This paper also marks the first quantum machine learning model applied to language on a realistic rather than toy dataset. 

This is a truly exciting advance for anyone interested in the union of quantum computing and artificial intelligence, and is in danger of being lost in the increased ‘noise’ from the quantum computing sector where organizations who are trying to raise capital will try to highlight somewhat trivial advances that are often duplicative.

Quantum Tensor Networks. A Scalable Approach

Carys Harvey and Richie Yeung from Quantinuum in the UK worked with a broader team that explored the use of quantum tensor networks for NLP. Tensor networks are mathematical structures that efficiently represent high-dimensional data, and they have found applications in everything from quantum physics to image recognition. In the context of NLP, tensor networks can be used to perform tasks like sequence classification, where the goal is to classify sequences of words or symbols based on their meaning.

The team performed experiments on our System Model H1, finding comparable performance to classical baselines. This marked the first time a scalable NLP model was run on quantum hardware – a remarkable advance. 

The tree-like structure of quantum tensor models lends itself incredibly well to specific features inherent to our architecture such as mid-circuit measurement and qubit re-use, allowing us to squeeze big problems onto few qubits.

Since quantum theory is inherently described by tensor networks, this is another example of how fundamentally different quantum machine learning approaches can look – again, there is a sort of “intuitive” mapping of the tensor networks used to describe the NLP problem onto the tensor networks used to describe the operation of our quantum processors.

What we’ve learned so far

While it is still very early days, we have good indications that running AI on quantum hardware will be more energy efficient. 

We recently published a result in “random circuit sampling”, a task used to compare quantum to classical computers. We beat the classical supercomputer in time to solution as well as energy use – our quantum computer cost 30,000x less energy to complete the task than Frontier, the classical supercomputer we compared against. 

We may see, as our quantum AI models grow in power and size, that there is a similar scaling in energy use: it’s generally more efficient to use ~100 qubits than it is to use ~10^18 classical bits.

Another major insight so far is that quantum models tend to require significantly fewer parameters to train than their classical counterparts. In classical machine learning, particularly in large neural networks, the number of parameters can grow into the billions, leading to massive computational demands. 

Quantum models, by contrast, leverage the unique properties of quantum mechanics to achieve comparable performance with a much smaller number of parameters. This could drastically reduce the energy and computational resources required to run these models.

The Path Ahead

As quantum computing hardware continues to improve, quantum AI models may increasingly complement or even replace classical systems. By leveraging quantum superposition, entanglement, and interference, these models offer the potential for significant reductions in both computational cost and energy consumption. With fewer parameters required, quantum models could make AI more sustainable, tackling one of the biggest challenges facing the industry today.

The work being done by Quantinuum reflects the start of the next chapter in AI, and one that is transformative. As quantum computing matures, its integration with AI has the potential to unlock entirely new approaches that are not only more efficient and performant but can also handle the full complexities of natural language. The fact that Quantinuum’s quantum computers are the most advanced in the world, and cannot be simulated classically, gives us a unique glimpse into a future. 

The future of AI now looks very much to be quantum and Quantinuum’s Gen QAI system will usher in the era in which our work will have meaningful societal impact.