Zapata Computing’s Formula for Achieving Quantum Advantage Today

By John Russell

August 16, 2021

While most of the quantum computing world is focused on achieving fault tolerant computing – some day – with intermediate applications for NISQ (near intermediate scale quantum) systems emerging along the way, Zapata Computing is taking a different approach. Its idea is to take the existing compute infrastructure necessary for large-scale enterprise applications and blend in a small amount of quantum enhancement to make those systems run better, better enough, in fact, to make the switch to quantum worthwhile for many applications.

It’s an interesting strategy that currently hinges largely on quantum computers’ ability to generate true random numbers, such as was done in Google’s experiment to prove quantum supremacy. In Zapata’s case the quantum output serves as input for existing ML/DL algorithms run on classical computers. The company is one of many emerging quantum software companies.

Christopher Savoie (photo on right) is the CEO of Zapata, a 2017 spin-out from Harvard, and if you have a chance, ask him about how the company and its name came about. It’s another interesting story. The force behind Zapata’s creation is Alán Aspuru-Gzik a prominent chemistry researcher at Harvard, who Savoie says is a “man who won’t take no for an answer.” Aspuru-Gzik convinced Savoie to join/lead Zapata over a burrito lunch at a restaurant containing a mural of Mexican revolutionary Emiliano Zapata. Aspuru-Gzik, who was born in the U.S. but grew up in Mexico City, was Harvard’s first Mexican-American tenured ‘hard science’ professor according to Savoie. The company name – Zapata Computing – is a tribute to Aspuru-Gzik who has since moved to the University of Toronto and is also Zapata’s CSO.

No surprise, they are hoping Zapata will also turn out to be as revolutionary.

Zapata’s core product is Orquestra a software platform described as being able to “compose quantum-enabled workflows and execute freely across the full range of quantum devices.” In Zapata’s hybrid quantum-classical model, HPC and classical computing play prominent roles, as does relying on heuristic methods rather than exact calculation. Put another way, machine learning and existing solvers do the heavy lifting but are guided by quantum input.

Here’s a brief description of the quantum-enhanced approach from Savoie:

“Any machine learning basically starts with a generative model, at least in modeling things, a probability distribution that’s made up of a random bit string. You start with a tunable, random bit string to get closer and closer – like a Born machine (quantum circuit born machine, QCBM) – to model the distribution that will give you a certain result. That’s your prior. [Next], you feed a distribution in and hopefully get a better handwriting sample, a better portfolio, a better optimization of something, chemistry, whatever, at the end. That’s the outcome. [To do that] you need a random bit string that has a tunable parameterized connection to it. The distribution of those bit strings is the source of richness in your model. That’s the premise.

“What we showed in our papers is if you have a quantum source of bit strings that comes from that source of supremacy (think of the quantum computer a la Google), you get a better distribution and a better handwriting sample and a better portfolio optimization than you can with classical machine work. We can do that today.

“So, what we’ve done is we’ve taken a quantum-enhanced neural network approach to basically put our quantum spy on that solver. We’re feeding into a classical neural network workflow that already exists and is already in production in a lot of places. The great thing is their cost function, their model, already exists too; we just put a little quantum spy on it. We think it’s pretty clever, and it’s patented, and very disruptive. We’re able to basically model the distribution of the good answers for any software.”

Remember, Zapata is a software play, using other people quantum computers. It has a handful of post-docs from Aspuru-Gzik’s lab. Savoie says the full team is highly-skilled. Part of what’s interesting is Zapata’s heavy focus on hybrid solutions and leveraging existing technology. Savoie lumps the quantum landscape, broadly, into two camps – one chasing fault tolerant quantum computing and another chasing nearer-term applications for NISQ machines although many of the latter still seem eagerly awaiting the fault-tolerant quantum flood gates to open.

It’s important to note that Zapata also gets deep into the quantum circuit weeds in terms of design approaches to produce useful guidance for classical heuristic methods. Its work with BBVA (global banking company) looked at innovative quantum circuit designs for use with Monte Carlo simulation focused on credit valuation adjustment. This “quantum” circuit and algorithm research is core to Zapata. What’s somewhat unusual is the use of the quantum output for input to traditional computers.

“The eventual promise of quantum computing is that we’ll be able to take in all different combinations of solutions and [be able to] explore the entire solution spaces. So we’ll be able to go through every single possible answer and find the best one. We have quantum algorithms that can speed up that computationally impossible task of trying every single parameter. That’s the brute force thing. Things like Monte Carlo simulations are brute force, try every different version [scenario],” said Savoie.

Accomplishing that vision remains far off, more R than D right now, he believes. Say 5-to-10 years.

“There’s another group of solutions that are not brute force trying every single solution, that use current methodologies like machine learning optimization techniques and solvers that we already have, and [we] provide a layer of quantum over those,” he said. Zapata has singled out several application areas, healthcare, finance, and aerospace, for example, and Savoie said there are currently pilot projects underway. It will take 24-to-36 months before these reach production. Still that’s an aggressive timeline.

Savoie said there are many reasons for leveraging existing technology. Most have to do with data management (movement, cleansing, etc.) and the needs of large scale of enterprise computing. These applications tend to use lots of data and much of it confidential or regulated and sometimes in many different locations. He points to companies including Palantir and C3 as disrupters in enterprise and hybrid cloud software and likens Zapata to them but with a quantum twist.

“We don’t throw out with our platform any of the things that we’ve learned about data management and cloud. It’s just that quantum becomes a forcing function to do some things that the workflow systems and platforms that Palantir and C3 didn’t have to do. One [challenge] is multi-cloud with regulated data. All of the industries that use quantum are HPC use-case companies. Finance, pharma, cars, aerospace, are all regulated. Any of the data that we’re going to be touching is regulated. And in these behemoth companies that are global, like Nissan Renault, where I used to work, that have all these enterprise architectures, security rules, for good reason, they’re dealing with people’s privacy, consumer data, clinical trial data, safety data, stuff like this. They’re regulated by various bodies around the world,” he said.

“In one of our workflows (VQE) we have to use a supercomputer to do a Hartree-Fock calculation. Before we start the quantum [work], our pre-processing needs to be done, and the data has to come from somewhere. So you have this data frame, but you can’t move a petabyte of data over and then calculate it or you’re going to create latency. You have the same data frame problem that C3 and Palantir had, but your compute is also much more complex,” said Savoie. “I need to use shared memory, so I can’t use containers. Pure cloud Kubernetes doesn’t really work all the time for us due to security reasons and enterprise architecture reasons. The quantum piece itself is sitting on some other cloud. I’m not going to move all of my applications and all of my data to IBM cloud to use IBM’s computer. IBM would love that but it’s not going to happen.

IBM Q System One in Japan

“If you want to do enterprise optimization for supply chain, where’s the ERP data? Where’s the ERP system? That data is going to have to be close to that quantum computational resource. Otherwise, if you’re round tripping in six seconds, every time we iterate on a machine learning model, what quantum giveth, your ETL process just took away. IBM is really starting to know this. You’re going to have to co-locate or find a way to do [something] like what Amazon does with Outposts,” he said.

The notion, of course, is that enterprise applications are data-intensive – something problematic for quantum computers and latency-sensitive. To some extent, that’s hardly new insight and many observers have forecast the quantum computers will end up as special purpose machines much like today’s chip accelerators. Savoie seems to have a different vision, one that suggests a little quantum magic could go a long way on assisting large enterprise applications while escaping many of quantum computing’s data management challenges.

He also emphasized the idea that quantum-enhanced heuristics computation techniques are here to stay and that they will be better enough than straight classical systems to be worth pursuing and that they may improve to become good enough to be worth running even when fault-tolerant quantum computing arrives because they will be simpler than the full quantum approach.

“We’re on this continuum towards fault tolerance. In fact, the work that we’re doing in VQE, for example, and some of the work that we did with BBVA (figure at end of article) is moving the bar to the point where we can actually get some useful heuristic methods that will work in a fault tolerant world, and also work in an interim regime where we have better error-mitigation as opposed to error correction,” he said. “So the new VQE that is 1,000 times better – that is actually moving us into a regime where we don’t exactly need fault tolerance.”

“Our viewpoint is that [it’s] a false dichotomy saying we have NISQ and then we’ll have fault tolerance. Like one day you wake up and there’s the step function. That’s not how it’s going to work, it’s actually going to be that we get better and better algorithms that work in an error-mitigated fashion, that will also work when fault tolerance turns on. And there’s no guarantee that the brute force way is going to be faster [or] give you that much better solution than your best heuristic. If the heuristics are pretty good, neural networks might work better than brute force [and] better than Bayesian statistics. Our viewpoint is that it’s a gradient,” Savoie said.

Currently, as you would expect, most engagements are collaborative and consultative – most user companies simply don’t have sufficient expertise. While Zapata is agnostic about quantum hardware choices – there’s so much jockeying among qubit technologies going on – it prefers superconducting qubits currently because they offer the best speed.

Feature Image: Google quantum processor

Figure from Zapata/BBVA CVA Paper outlining its approach

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire