Configurable Packaged Quantum Solutions Are the Greatest Opportunity for Widespread Adoption of Quantum Computing

  • Caveat: This paper proposes and envisions a potential future for quantum applications which would greatly facilitate the adoption of quantum computing by the widest possible audience and market. There is no guarantee that this envisioned future will come to fruition, nor is there any certainty or confidence as to when it might come to fruition. The only confidence is that it won’t happen in the very near future (next year or two), and is likely to take at least a few years to develop — and require significant progress in the advancement of quantum computing hardware.
  • Counter-caveat: The proposal of this paper may not be a certainty but it certainly is a golden opportunity.
  1. In a nutshell
  2. My original description of the topic
  3. The essence of the problem
  4. The bigger picture
  5. Generative coding
  6. Presumption of production-scale and substantial quantum advantage
  7. No need to be Quantum Ready for configurable packaged quantum solutions since the arcana of quantum computing, quantum mechanics, linear algebra, etc. are not needed
  8. Welcome to Quantum Done
  9. Algorithms, applications, and solutions
  10. Some basic observations about applications
  11. Some basic observations about algorithms
  12. Solution vs. application
  13. An application is just the code
  14. A solution is much more than just the code
  15. Is it an application or a solution?
  16. Three levels of algorithms and applications: Custom, generalized, and configurable
  17. Make vs. buy decision
  18. Creation and marketing of configurable packaged quantum solutions
  19. Two distinct application use case categories — minor use of quantum and quantum-focused
  20. How quantum applications can use quantum algorithms
  21. Turnkey solutions — requirements and expectations
  22. Black boxes
  23. A configurable packaged quantum solution is a turnkey solution
  24. Anybody can develop an application, but a solution requires a dramatically higher level of skill
  25. Key features and benefits of configurable packaged quantum solutions
  26. The single greatest benefit of configurable packaged quantum solutions is to enable any organization of any size to exploit quantum computing without an onerous investment in quantum expertise
  27. Resilience
  28. Personal versus product use
  29. Single-purpose, specialized, and niche applications versus general-purpose and generalized applications
  30. Generalized applications
  31. Reusable applications
  32. Amortized costs for generalized and reusable applications
  33. Packaged applications
  34. Configurable packaged quantum solution
  35. Configurable packaged quantum solutions could be single-purpose or general-purpose
  36. Single-purpose configurable packaged quantum solutions can have high value for multiple customers
  37. Horizontal vs. vertical applications
  38. Configurable packaged quantum solutions can apply in many situations
  39. Many organizations will have similar specialized needs
  40. Many organizations will have a variety of situations with similar if not identical needs
  41. Configurable packaged quantum solutions could be horizontal or vertical applications
  42. Structure and form of applications and solutions can vary
  43. Packaging quantum solutions as network services
  44. Cloud-based applications and services
  45. Cloud-based configurable packaged quantum solutions
  46. Quantum-as-a-Service (QaaS) for quantum-focused solutions
  47. Pure quantum vs. hybrid quantum/classical web services
  48. Little Data With a Big Solution Space — the Sweet Spot for Quantum Computing
  49. Architecture and design details of generalized solutions is beyond the scope of this paper
  50. Configurable packaged quantum solutions vs. toolkits and frameworks
  51. Some examples of complex configuration data
  52. Fielder’s choice whether input data should be self-configuring, configured by input parameters, or a separate configuration description
  53. Specification of configuration
  54. All input data and parameters and configuration data should be in application-oriented terms, not the terms of quantum computing or quantum mechanics
  55. Data formats
  56. Configuring or controlling output
  57. Some examples of configurable classical solutions
  58. Examples of packaged quantum solutions
  59. Industry-specific solutions
  60. Unlikely that there will be a single configurable packaged quantum solution per application category, but maybe a dozen or more per application category
  61. Only a relative few configurable packaged quantum solutions are likely to cover the majority of use cases for real users in the real world
  62. The seven main quantum application categories
  63. Expect 70–80% of users (customers) to get by with configurable packaged quantum solutions alone
  64. Aim for 90% of users (customers) to get by with configurable packaged quantum solutions alone
  65. Critical mass of configurable packaged quantum solutions
  66. No need for full quantum error correction (QEC) — near-perfect qubits should be sufficient
  67. Debugging and logging
  68. The ENIAC Moment won’t necessarily depend on or result in a configurable packaged quantum solution
  69. The ENIAC Moment might provide the basis or impetus to kick start configurable packaged quantum solutions
  70. When can we expect the birth of the era of configurable packaged quantum solutions?
  71. Configurable packaged quantum solutions may actually be easier to design and build than full-custom applications
  72. Vendors and sources of configurable packaged quantum solutions
  73. Encourage open source software projects
  74. Ecosystem for configurable packaged quantum solutions
  75. Need for sample configurable packaged quantum solutions
  76. Quantum expertise
  77. Training for quantum
  78. Talent pool for quantum
  79. Broader is better, but it can be too expensive
  80. Significant hardware advances are needed
  81. Significant research is required
  82. Quantum-inspired solutions are great, but they don’t advance quantum computing itself
  83. Intellectual property (IP) — is a proprietary interest helpful or harmful?
  84. Provisioning a configurable packaged quantum solution
  85. Three stages of adoption for quantum computing: The ENIAC Moment, configurable packaged quantum solutions, and The FORTRAN Moment
  86. The FORTRAN Moment will herald widespread full-custom quantum algorithms and applications
  87. Summary and conclusions

In a nutshell

  1. Generalized applications can be applied in a wider range of situations for a wider range of users at a wider range of organizations.
  2. Turnkey solution. Everything is in place. No coding required. All of the necessary code has already been written and thoroughly tested. Just supply input data, some parameter settings, and possibly an application configuration description.
  3. No specialized knowledge needed. No quantum computing. No quantum mechanics. No linear algebra. It’s all encapsulated inside the application.
  4. Insulate users, customers, and classical application code from any knowledge or dependence on any of the concepts of quantum computing.
  5. No need to be Quantum Ready. Or even Quantum Aware. Welcome to Quantum Done. Everything to do with quantum computing is complete, ready to go, and hidden deep under the hood.
  6. No need for an elite technical team. No reliance on availability of a quantum technical talent pool. No need to attract or retain elite quantum technical staff. No need for the cost. No need for the uncertainty. No need for the risk.
  7. No need for quantum experts. It’s all there, hidden under the hood.
  8. No need for quantum expertise. It’s all there, hidden under the hood.
  9. No need for quantum error correction or perfect logical qubits. Near-perfect qubits are sufficient.
  10. No need for any specialized workflows. Looks just like a normal, traditional, classical application. No quantum choreography required. Anything quantum is complete and hidden under the hood.
  11. Presumption of production-scale. No prototypes or experimental software here.
  12. Presumption of production-ready. Ready for full operation, to solve production-scale practical real-world problems.
  13. Fully validated. All functions and features have been fully tested.
  14. Full explainability. The decisions leading to application results are fully reported and fully explained. Reports exactly what it did and why. Leaves little to the imagination. Not the raw quantum computing per se or the physics, but the overall process and the math but in application terms.
  15. Resilient. Little need to worry about failure scenarios. Either the solution completely handles all failure scenarios or provides detailed but easy to comprehend diagnostics for any unhandled failures.
  16. Presumption of substantial quantum advantage. If not true dramatic quantum advantage. Far superior to any classical solution.
  17. Generative coding. Quantum circuits will not be handcrafted and hardwired into the application as might be the case for simple or toy or experimental applications, but automatically generated dynamically by classical code based on the input data, input parameters, and configuration data.
  18. Automatic scaling of algorithms. No manual intervention or technical expertise required. The joys of generative coding.
  19. Significant hardware advances are needed. Current quantum hardware is woefully inadequate. Qubit fidelity and qubit connectivity need dramatic improvement. Near-perfect qubits are required.
  20. Significant research is required. Hardware, qubit technology, architecture, control software, algorithms, programming models, applications, application frameworks, etc. You name it, it needs research.
  21. Quantum-inspired solutions. Benefit from the same overall approach, but don’t require hardware advances. But they don’t advance quantum computing itself per se — they’re still only classical computing and don’t offer an exponential speedup or any dramatic quantum advantage.
  22. When? Unfortunately, not real soon. Maybe three to five years. Maybe seven years. Four to five years may be a good bet. But with quantum, nothing is a truly safe bet.
  23. Caveat. This paper proposes and envisions a potential future for quantum applications which would greatly facilitate the adoption of quantum computing by the widest possible audience and market. There is no guarantee that this envisioned future will come to fruition, nor is there any certainty or confidence as to when it might come to fruition.

My original description of the topic

  • Configurable packaged quantum solutions are the greatest opportunity for widespread adoption of quantum computing. Prewritten code — complete applications — addressing particular niches of quantum computing applications. User supplies the input data and a variety of application-oriented configuration parameters. Prewritten classical code will generate the necessary quantum circuits needed to implement the underlying algorithms using the user’s input data and configuration parameters. Algorithms and code for such solutions must be designed and developed by very elite professional teams, well beyond the ability of even most Fortune 500 companies, who will be the target customers of such packaged solutions. No knowledge needed of the physics or math of quantum computing. No knowledge needed of the internals of the underlying algorithms of the packaged solutions. Examples — none exist, yet. D-Wave is a step in this direction, but falls short.

The essence of the problem

The bigger picture

Generative coding

Automatic scaling of algorithms

Presumption of production-scale and substantial quantum advantage

  1. Presumption of production-scale. No prototypes or experimental software here.
  2. Presumption of production-ready. Ready for full operation, to solve production-scale practical real-world problems. Not a tool for developers.
  3. Presumption of substantial quantum advantage. If not true dramatic quantum advantage, but that may be too much to ask for, initially. There’s no point if the solution does not deliver a clear and very compelling (substantial) improvement over classical solutions.
  4. No intention here to cater to research, experimentation, prototyping, and toy-scale applications. Although a configurable packaged quantum solution can certainly be used for experiments and prototyping, and research as well, the focus is on production-scale and production-ready.

No need to be Quantum Ready for configurable packaged quantum solutions since the arcana of quantum computing, quantum mechanics, linear algebra, etc. are not needed

Welcome to Quantum Done

Algorithms, applications, and solutions

  1. An application tends to be very specialized, solving one, single, specific, particular problem. The user will have to do another application if any of the details differ. No generality.
  2. All input data must be carefully validated. Sometimes by hand.
  3. Input data may need to be formatted and presented in some cryptic form.
  4. Behavior of the application is unpredictable if the input data is invalid.
  5. Output data may be in some cryptic format.
  6. Significant effort may be required to make use of the output data.
  7. The user may have to hand-tune or otherwise modify either the classical code or the quantum algorithm to customize behavior.
  8. Many nuances of behavior to be aware of and to work around.
  9. And more. Plenty of gotchas are common with applications in general.
  1. Be more general than simply a specific, particular problem. Designed to be able to handle variations in the details for the problem.
  2. Discover as much as possible from the input data itself.
  3. Be fully prepared for invalid input data.
  4. Provide intelligent and convenient input options to allow easy control of the application without any need to modify or even read the code or quantum algorithms.
  5. Accept a high-level description or configuration to guide the interpretation of the input data and input parameters.
  6. Present a model of application behavior that is free from nuances and gotchas.

Some basic observations about applications

  1. Coding any application from scratch can be a very tedious and error-prone task.
  2. Coding a quantum application from scratch will generally be a far more complex task.
  3. Designing and coding any algorithm from scratch can be a very tedious and error-prone task.
  4. Designing and coding a quantum algorithm from scratch will generally be a far more complex task.
  5. Acquiring a complete application can be expensive, but avoids the need to know anything about what happens under the hood inside of the application.
  6. Acquiring a complete quantum application can be expensive, but avoids the need to know anything about what happens under the hood inside of the application.
  7. Customizing the code of any application can also be a very tedious and error-prone task.
  8. Customizing the code of a quantum application will generally be a far more complex task.
  9. Customizing any algorithm can be a very tedious and error-prone task.
  10. Customizing any quantum algorithm will generally be a far more complex task.

Some basic observations about algorithms

  1. A hand-coded algorithm tends to be rather inflexible.
  2. An algorithm generated by a computer program can be more flexible.
  3. Code which generates algorithms tends to be significantly more difficult to design, understand, and maintain.

Solution vs. application

  1. An application tends to focus on solving one, specific, particular problem.
  2. While a solution tends to be generalized to solve a class of problems. It may be a relatively narrow or niche class, but is still more general than solving a particular problem.
  3. An application is mostly just the code with informal processes to guide its usage. And the user must be responsible for the code, all of it.
  4. While a solution has everything in place, all of the code, leaving little for the user to guess about or screw up — and if they do screw up it will be clearly diagnosed. No need to write any code, modify any code, or even view or understand any code to use the solution.

An application is just the code

A solution is much more than just the code

  1. Turnkey solutions — requirements and expectations.
  2. Key features and benefits of configurable packaged quantum solutions.

Is it an application or a solution?

Three levels of algorithms and applications: Custom, generalized, and configurable

  1. Custom algorithms and applications. Focused on a singular, particular purpose, situation, user, and organization.
  2. Generalized algorithms and applications. For more than one purpose, situation, user, or organization. May still require enhancing, supplementing, or otherwise modifying the code.
  3. Configurable algorithms and applications. Especially configurable packaged quantum solutions as espoused by this paper. Generalized at their core, but configurable as well. Special attention to resilience. Generalized through configurability rather than any need to enhance, supplement, or modify the code by hand.

Make vs. buy decision

  • Should we make the application ourselves? From scratch, or adapting an existing application.
  • Or should we buy an existing off-the-shelf application?
  • Should we hire a consultant to make or buy the application? Possibly directing the consultant which way we are biased, or deferring to their advice on make vs. buy.
  • Or should we decide to make or buy by ourselves? Not feeling any need for the expertise of the consultant.

Creation and marketing of configurable packaged quantum solutions

Two distinct application use case categories — minor use of quantum and quantum-focused

  1. Minor use of quantum. The overall application has nothing to do with quantum computing per se, but one or more niche narrow, very-contained computations are quantum-based. This may in fact be a classical application which has been adapted to quantum computing for those narrow, niche portions of the overall application computation.
  2. Quantum-focused. A newer type of application where the overall application is focused on quantum computing, possibly with some amount of classical glue code, but everything is in service to the quantum logic. Not usually comparable to any existing classical application.

How quantum applications can use quantum algorithms

  1. Full-custom quantum algorithm design required. Each quantum application does its own quantum algorithms from scratch. Either custom research is required, or an implementation based on published papers.
  2. Radically adapt or redesign an existing quantum algorithm. Significant redesign, but key elements remain unchanged. Might be unrecognizable to someone familiar with the original algorithm.
  3. Moderately adapt an existing quantum algorithm. Moderate redesign, but many or most elements remain unchanged. Tends to be somewhat recognizable, but not necessarily.
  4. Modestly adapt an existing quantum algorithm. Modest redesign, but many or most elements remain unchanged. Tends to be very recognizable.
  5. Trivially adapt an existing quantum algorithm. Modest or even trivial changes, such as changing some parameters or data values. No significant changes to the overall design — will look unchanged unless you know what to look for. Very recognizable, and even can be confused with the original quantum algorithm.
  6. Use an existing, fully-developed off the shelf algorithm as-is. No changes.
  7. Use a classical black-box module. Invoke a classical module which has a quantum algorithm embedded under the hood. The module takes classical input and parameters, and produces a classical result. The application never sees or can even sense that a quantum algorithm is used, but the internal application structure might be different from a pure-classical structure to accommodate the quantum algorithm. The quantum algorithm might even execute on a classical quantum simulator, or may be a quantum-inspired classical algorithm. Or, by default, the quantum algorithm is executed on a quantum computer, either locally, over a local area network, or accessed as an external Internet (cloud) service.
  8. Fully-classical application structure that 100% replaces an existing classical module with the quantum-enabled module. 100% functionally compatible with the original classical application, but just faster and/or more accurate or otherwise more functional due to invoking the quantum algorithm from within the quantum-enabled module. Although this might simply result in a significant performance improvement, the ultimate goal is to be able to produce a result when a pure-classical module could not, otherwise known as quantum supremacy.
  9. Classical application invokes a quantum web service over the Internet (cloud). Could be a new service that is explicitly quantum which could require deep quantum knowledge or it could require some relatively light quantum awareness. Or it could appear to be a purely classical service that just happens to use a quantum algorithm deep under the hood, either locally or remotely as a service (cloud) — two levels of web service.

Turnkey solutions — requirements and expectations

  1. Ready to go. Ready to run, as-is. Nothing else is needed — other than the input data.
  2. No additional software required.
  3. No additional software components or modules needed.
  4. No adaptation of the code required. Classical or quantum.
  5. No modification of the code required. Classical or quantum.
  6. Everything is flexible. Nothing is hardcoded. All quantum algorithms are automatically adapted to the input data, input parameters, and application-oriented configuration data.
  7. Fully validated. All features have been fully tested.
  8. Scaling is automatic. No manual intervention or technical expertise required.
  9. No need for an elite technical team. No reliance on availability of a quantum technical talent pool. No need to attract or retain elite quantum technical staff. No need for the cost. No need for the uncertainty. No need for the risk.
  10. No need to be Quantum Ready. Welcome to Quantum Done. Everything to do with quantum computing is complete, ready to go, and hidden under the hood.
  11. Fully documented.
  12. No need to examine any of the code. Classical or quantum.
  13. No specialized knowledge needed. No quantum computing. No quantum mechanics. No linear algebra. Not even software development.
  14. Everything is in application terms. No specialized vocabulary or knowledge required. Everything is expressed in terms of the specific application domain. No quantum computing jargon required.
  1. Acquire access to the solution.
  2. Install, enable, or otherwise deploy the software.
  3. Supply the input data.
  4. Supply any optional parameters.
  5. Optionally supply configuration data. Application-oriented description of the problem to be solved. Or this may be automatically inferred directly from the input data.
  6. Run the solution. Push the button, turn the key. Invoke the application.
  7. Access and use the results.

Black boxes

A configurable packaged quantum solution is a turnkey solution

Anybody can develop an application, but a solution requires a dramatically higher level of skill

Key features and benefits of configurable packaged quantum solutions

  1. 100% solution. No additional code to write or adapt.
  2. Full explainability. The decisions leading to algorithm and application results are fully reported and explained. Not the raw quantum computing per se or the physics, but the overall process and the math but in application terms.
  3. Extensive, foolproof validation of all input data and parameters with readable diagnostics. All detectable errors. Highlight possible inconsistencies. Highlight possible undesirable consequences. Highlight potential optimizations or opportunities for simplification. Eliminate all surprises. Leave nothing to chance.
  4. Extensive tools for processing and examining and diagnosing quantum results. Automated whenever possible. And very configurable.
  5. Full validation of quantum results. When possible all results should be fully validated, but for some applications (e.g., optimization) that may not be possible. But at least validate that the results are plausible, such as that optimization is better than a Monte Carlo simulation.
  6. Each solution focuses on a specific and clearly documented application niche. Either vertical or horizontal.
  7. Excellent customer service support.
  8. Resilient. Little need to worry about failure scenarios. Either the solution completely handles all failure scenarios or provides detailed but easy to comprehend diagnostics for any unhandled failures.
  9. Outstanding online community providing great support.
  10. Extensive online documentation.
  11. Extensive online tutorials
  12. Extensive training. Readily available.
  13. Extensive consulting services available.
  14. Service-level agreement (SLA). Guarantee 24/7 availability and uptime.
  15. Frequent updates. Generally fully compatible, with no customer changes needed.
  16. Prototyping and proof of concept use cases fully supported by simulation.
  17. 100 to 1,000 times more complex than a typical prototype or experimental application.
  18. No need for a super-elite quantum team.
  19. No need for an elite quantum team of any sort. Everything quantum is encapsulated and hidden within the solution.
  20. Each configurable packaged quantum solution does require an elite team. The team developing a configurable packaged quantum solution does require an elite team with broad and deep quantum expertise, but they are not part of the organization deploying or using the solution.
  21. But each instance doesn’t require an elite team to deploy each instance of the application (solution.)
  22. No development time.
  23. Rapid deployment.
  24. No maintenance staff.
  25. No need for Quantum Ready. It’s all hidden inside the box, under the hood of the solution.
  26. No specialized knowledge needed. Other than for the application domain itself. No quantum computing knowledge needed. No quantum mechanics knowledge needed. No physics knowledge needed. No quantum linear algebra needed.
  27. All outsourced. Everything. Algorithm knowledge. Application knowledge. Potentially infrastructure as well — with cloud access to a vendor system.
  28. Eliminate the open-ended budget, planning, organizational, and business risk of an expensive and complex application.
  29. Minimize talent cost and risk.
  30. Minimize training cost.
  31. Minimize IT operational costs.

The single greatest benefit of configurable packaged quantum solutions is to enable any organization of any size to exploit quantum computing without an onerous investment in quantum expertise

  1. No need to hire expensive quantum talent.
  2. No need to train expensive quantum talent.
  3. No need to retain expensive quantum talent.
  4. No need to be Quantum Ready.
  5. No need to be Quantum Aware.

Resilience

  1. There is little need to worry about failure scenarios. Either the solution completely handles all failure scenarios or provides detailed but easy to comprehend diagnostics for any unhandled failures.
  2. All failure scenarios have been identified and tested.
  3. Everything that can be tested has been tested.
  4. Extensive, foolproof validation of all input data and parameters with readable diagnostics. All detectable errors. Highlight possible inconsistencies. Highlight possible undesirable consequences. Highlight potential optimizations or opportunities for simplification. Eliminate all surprises.
  5. Full validation of quantum results. When possible all results should be fully validated, but for some applications (e.g., optimization) that may not be possible. But at least validate that the results are plausible, such as that optimization is better than a Monte Carlo simulation.
  6. All failure scenarios have been fully identified and fully documented.
  7. Even if automatically handled, all failures will be reported. And reported in an easy to understand manner.

Personal versus product use

Single-purpose, specialized, and niche applications versus general-purpose and generalized applications

  1. Single-purpose applications.
  2. Specialized applications.
  3. Niche applications.
  1. General-purpose applications.
  2. Generalized applications.

Generalized applications

  • Turnkey solutions — requirements and expectations.
  • Key features and benefits of configurable packaged quantum solutions.

Reusable applications

Amortized costs for generalized and reusable applications

Packaged applications

Configurable packaged quantum solution

  1. A solution. Rather than merely an application. Fully solves an application problem, not simply a tool in a toolkit, not simply a framework which is only a starting point.
  2. Quantum-based. Utilizes a quantum computer for some key portion of its computations.
  3. Generalized. Not focused on a single, particular problem, user, or organization.
  4. Is packaged. Somewhat more generalized application which can be used by multiple users with no change to the code, but simply by supplying alternative input data and parameters. Significant documentation for proper usage. Rigorously tested.
  5. Configurable. Accepts a high-level description of the problem to be solved.
  6. Resilient. Handles all failure scenarios, both expected and unexpected.
  7. Turnkey solution. Everything is in place. No coding required. All of the necessary code has already been written and thoroughly tested. Just supply input data, some parameter settings, and possibly an application configuration description.
  8. Full explainability. The decisions leading to application results are fully reported and explained. Reports exactly what it did and why. Leaves little to the imagination. Not the raw quantum computing per se or the physics, but the overall process and the math but in application terms.

Configurable packaged quantum solutions could be single-purpose or general-purpose

Single-purpose configurable packaged quantum solutions can have high value for multiple customers

  1. Differing input data.
  2. Differing input parameters.
  3. Differing configuration data. Solving fairly similar but not absolutely identical application problems.

Horizontal vs. vertical applications

  1. Horizontal applications. Apply across a wide range of subject domains or industries.
  2. Vertical applications. Apply across the range of functional areas of a single subject domain or industry.

Configurable packaged quantum solutions can apply in many situations

Many organizations will have similar specialized needs

Many organizations will have a variety of situations with similar if not identical needs

Configurable packaged quantum solutions could be horizontal or vertical applications

Structure and form of applications and solutions can vary

  1. Interactive GUI application.
  2. Command-line program. UNIX or DOS style. Standard input. Command line options. Standard output. Access to files and network services. Scripting.
  3. HTTP REST API network service. For more-sophisticated solutions.

Packaging quantum solutions as network services

Cloud-based applications and services

Cloud-based configurable packaged quantum solutions

Quantum-as-a-Service (QaaS) for quantum-focused solutions

  1. Minor use of quantum. It would be a stretch to consider this a true Quantum-as-a-Service (QaaS). The overall application has nothing to do with quantum computing per se, but one or more niche narrow, very-contained computations are quantum-based. This may in fact be a classical application which has been adapted to quantum computing for those narrow, niche portions of the overall application computation.
  2. Quantum-focused. This definitely would be considered a Quantum-as-a-Service (QaaS). A newer type of application where the overall application is focused on quantum computing, possibly with some amount of classical glue code, but everything is in service to the quantum logic. Not usually comparable to any existing classical application.

Pure quantum vs. hybrid quantum/classical web services

Little Data With a Big Solution Space — the Sweet Spot for Quantum Computing

  1. Only a very small amount of input data. All input data must be encoded in the gate structure of the quantum algorithm (circuit.)
  2. Only a very small computation. Limited by coherence time and total quantum circuit size. Relatively shallow maximum circuit depth. Limited total circuit size.
  3. A very large solution space. n qubits provide 2^n quantum states. 50 qubits offer a quadrillion quantum states. So, a small computation can be evaluated over a very large solution space.
  4. Only a very small amount of output data. Whatever qubits can be measured.
  1. Read, write, and update data of any type from any source and destination.
  2. Perform classical I/O.
  3. Access classical files.
  4. Access classical databases. Or network-based databases.
  5. Access classical web services. Locally or over the Internet.
  6. Access pure quantum web services. Locally or over the Internet.
  7. Access other hybrid quantum/classical web services.
  8. Access Big Data. Process very large amounts of data.
  9. Generate and store large results. Large amounts of output data.
  10. Perform large and complex computations.

Architecture and design details of generalized solutions is beyond the scope of this paper

Configurable packaged quantum solutions vs. toolkits and frameworks

  1. Cirq.
  2. OpenFermion.
  1. Qiskit.
  2. Qiskit application modules.
  3. Qiskit Nature. Covers physics, chemistry, drugs, materials.
  4. Qiskit Finance.
  5. Qiskit Optimization.
  6. Qiskit Machine Learning. Can be applied to any application domain.

Some examples of complex configuration data

  1. Description of an atom, particle, molecule, or physical system to be simulated.
  2. Description of chemicals whose reaction is to be simulated.
  3. Description of a drug molecule to be simulated or analyzed.
  4. Description of a protein molecule to be simulated or analyzed.
  5. Description of the molecular structure of a material to be simulated.
  6. Description of the elements of a business process to be optimized.
  7. Description of a financial portfolio to be optimized.
  8. Description of data for a machine learning algorithm.

Fielder’s choice whether input data should be self-configuring, configured by input parameters, or a separate configuration description

  1. A separate configuration description. Clean separation between input data, input parameters, and configuration data. Allows a configuration of arbitrary complexity. This is the most general and offers the maximal flexibility, but can be cumbersome if the configuration is relatively simple or would be more easily specified as simple input parameters or embedded in the input data.
  2. Specify configuration data using input parameters. If the configuration data is simple enough, a relatively small number of input parameters might be sufficient to describe the configuration. This makes sense if the configuration data is very simple and doesn’t make more sense being embedded in the input data.
  3. Input data is self-configuring. Given that the input data can have a semantically-rich description (such as XML or JSON), the configuration description could be included directly in the input data. Either explicitly, using separate data elements, or implicitly as part of the input data itself. This makes sense if the configuration can trivially be specified in the input data and directly parallels the input data. Or if an automated analysis of the raw input data can discern patterns and structure.

Specification of configuration

  1. A detailed problem description. Separate configuration data.
  2. Hints or instructions specified as simple input parameters.
  3. Automatic detection of configuration from raw input data. Based on automated examination and analysis of the raw input data. No explicit configuration needed. Input data is self-configuring.

All input data and parameters and configuration data should be in application-oriented terms, not the terms of quantum computing or quantum mechanics

  1. Qubits.
  2. Quantum logic gates.
  3. Quantum state.
  4. Basis states.
  5. Wavefunctions.
  6. Hamiltonians.
  7. Product states.
  8. Bell states.
  9. Bloch sphere and Bloch sphere rotations.
  10. Phase and phase angles.
  11. Superposition.
  12. Entanglement.
  13. Interference.
  14. Teleportation.
  15. Probability amplitudes.
  16. Density matrix.
  17. Kets and bra-ket notation.
  18. Hilbert space.
  19. Vector space.
  20. Measurements.
  21. Gate pulses.
  22. Shot counts.
  23. Circuit repetitions.
  24. Quantum circuit.
  25. Quantum algorithm.
  26. Quantum results.
  27. Unitary transform matrix.
  28. Hermitian operators.
  29. Eigenvalues, eigenvectors, eigenstates.

Data formats

  1. Simple text. Simple strings, keywords, numbers, or lines of text.
  2. Comma-separated values (CSV). Single row of data or rows of data.
  3. Tab-delimited. Single row of data or rows of data.
  4. XML. Ranging from simple to very complex.
  5. JSON. JavaScript Object Notation. Ranging from simple to very complex.
  6. Semantic Web RDF. Graphs of subject, predicate, and object triples. Ranging from simple to very complex.
  7. Read from a database. Specify a database, table name, primary key, column name(s), etc.

Configuring or controlling output

  1. Data format.
  2. Filtering. How much of the computational results to include or exclude.
  3. Identify a service to invoke with results.
  4. Identify a database in which to store results. Including a database, table name, primary key, column name(s), etc. Might be relative to the input data source, such as to read and then update a database.
  5. Distinguish results of a full application or solution vs. simply a module producing intermediate results to be sent on to a larger component of a complex application.
  6. Support workflow processing models.

Some examples of configurable classical solutions

  1. Database servers. Schema to describe the structure of data. And storage schema to describe how the data should be stored. Various formats for import and export of rows of data or complex data structures, including graphs.
  2. Search engines. Configuration of data sources and data formats, including fields that can be searched by various data formats or text keywords. Configurations of complex data formats. For example Apache Solr and Elasticsearch.
  3. Web service REST API. Applications can use the web service without any knowledge of its internals or any need to modify or even examine the code. All you need are a URL and input data, and possibly credentials and a license.
  4. Web sites. Users and applications can use HTTP-based web servers without any knowledge of their internals or any need to modify or even examine the code. Tim Berners-Lee was working with a research team at LHC to solve a general problem in a narrow niche area (high-energy subatomic physics), but saw opportunities for a much broader range, and pursued it. His solution was not just for physics, not just for science, not just for technology, and not just business.
  5. MATLAB. Numeric computing environment. Complete with a programming language.
  6. Wolfram Mathematica and Wolfram Alpha. Symbolic computing environment. Computable data sets.

Examples of packaged quantum solutions

  1. Quantum Origin. A platform service developed by Cambridge Quantum Computing Limited (now part of Quantinuum, an offshoot of Honeywell) to provide certifiably random encryption keys using a quantum algorithm running on a quantum computer. As per their website: “Quantum Origin is a cloud-hosted platform that uses the unpredictable nature of quantum mechanics to generate cryptographic keys seeded with verifiable quantum randomness from quantum computers. The platform supports and helps protect traditional algorithms, such as RSA or AES, as well as post-quantum cryptography algorithms currently being standardised by the National Institute for Standards and Technology (NIST).
  1. Production-scale. Using a network service just to generate a random number is rather inefficient for generating a large number of random numbers.
  2. Substantial quantum advantage. No obvious dramatic performance advantage over classical computing. There are some specialized hardware devices which can be used with a classical computer, so there is no dramatic quantum advantage here.

Industry-specific solutions

  1. Machine learning.
  2. Optimization.
  1. Concepts and vocabulary.
  2. Theory.
  3. Practice.
  4. A domain defined by concepts, vocabulary, theory, and practice.
  1. Physics.
  2. Chemistry.
  3. Drugs.
  4. Materials.
  5. Finance.

Unlikely that there will be a single configurable packaged quantum solution per application category, but maybe a dozen or more per application category

Only a relative few configurable packaged quantum solutions are likely to cover the majority of use cases for real users in the real world

The seven main quantum application categories

  1. Simulating physics.
  2. Simulating chemistry.
  3. Material design.
  4. Drug design.
  5. Business process optimization.
  6. Finance. Including portfolio optimization.
  7. Machine learning.

Expect 70–80% of users (customers) to get by with configurable packaged quantum solutions alone

Aim for 90% of users (customers) to get by with configurable packaged quantum solutions alone

Critical mass of configurable packaged quantum solutions

No need for full quantum error correction (QEC) — near-perfect qubits should be sufficient

Debugging and logging

The ENIAC Moment won’t necessarily depend on or result in a configurable packaged quantum solution

The ENIAC Moment might provide the basis or impetus to kick start configurable packaged quantum solutions

When can we expect the birth of the era of configurable packaged quantum solutions?

Configurable packaged quantum solutions may actually be easier to design and build than full-custom applications

Vendors and sources of configurable packaged quantum solutions

  1. Academic research labs. More for specialized research applications.
  2. Government research labs. More for specialized applications related to the research interests of the particular government lab or project or team.
  3. Open source applications and solutions. Open source community and ecosystem for support and enhancement. Major universities. Government research labs. NASA. Major companies, including industrial, manufacturing, pharmaceutical, technology, and financial firms seeking to develop a community and ecosystem to support their efforts rather than shoulder all of the cost of maintaining, enhancing, and supporting the solutions.
  4. Major application software vendors. Commercial and business orientation. Traditional firms with classical orientation.
  5. Newer, quantum-specific application software vendors. Startups.
  6. Major consulting firms. Accenture, BCG, McKinsey, others. Actual solution may be a small portion of a much larger consulting engagement. Establishing best practices.
  7. Newer, quantum-specific application consulting firms. Startups.
  8. Cloud service vendors. AWS, Microsoft, IBM, Google, Oracle, others. Either raw infrastructure alone (but easy to use) or full turnkey access to complete solutions, with no specialized knowledge or significant work required.

Encourage open source software projects

  1. Free software.
  2. Free and open access to source code.
  3. Community. Of both users and technical contributors. Both support and development and enhancement.
  4. Ecosystem. Encourages collateral projects to support and build on the project.
  5. Develop a broader and deeper talent pool.
  6. Lower technical, management, and business risk. Less dependence on in-house key staff.

Ecosystem for configurable packaged quantum solutions

  1. Online support community. Sharing of experiences. Best practices. Troubleshooting problems. Discussion of proposals for improvements and enhancements.
  2. Online development community. Distinct sub-community for those seeking to work under the hood to actually implement improvements and enhancements. And to fix bugs and other deficiencies and limitations.
  3. Toolkits, libraries, frameworks, and network services used under the hood to develop a particular configurable packaged quantum solution.
  4. Add-on vendors. Optional extensions for the core product. May be free, purchased, or by subscription.
  5. Hardware vendors.
  6. Tools and support software. Anything to make it easier.
  7. Cloud infrastructure and service vendors.
  8. Consultants. From individual freelance consultants to the major consulting firms.

Need for sample configurable packaged quantum solutions

  1. Introduction to the application problem to be solved. Short summary plus extended detail.
  2. 100% solves the specified application problem.
  3. Can easily be copied and modified to adapt to similar situations. Or used as a module for even radically different applications.
  4. All source code is included. Including unit tests. Preferably in a GitHub repository.
  5. Full documentation for all functions and features of the solution.
  6. All application functions and features are implemented.
  7. All application functions and features have been fully tested. Include all functional test cases. And sample test results.
  8. A variety of sample input data. And corresponding sample output result data.
  9. A variety of sample input parameters. And corresponding sample output result data.
  10. Everything is in a GitHub repository. All code. All test data. Sample of test results. All sample input data. Sample of output result data for all sample input data and parameters.

Quantum expertise

  1. Formal education. Physics, quantum chemistry, mathematics, computer science, computer engineering, electrical engineering, engineering in general.
  2. Technical education. All aspects of technology related to quantum computing.
  3. Specialized training. Specific technologies, specific products, specific methods, specific tools.
  4. Quantum mindset. A general awareness and intuition about quantum effects.
  5. Quantum experience. Actual experience using quantum technologies and exposure to work which is based on quantum effects.
  6. Very elite. A much more select set of individuals, teams, and organizations whose education, training, knowledge, skill, expertise, and experience in quantum effects, quantum technologies, and quantum computing are far beyond what is typical for individuals, teams, and organizations.
  7. Quantum Aware. A much broader set of individuals, teams, and organizations whose education, training, knowledge, skill, expertise, and experience in quantum effects, quantum technologies, and quantum computing are far more limited than the very elite.
  8. Quantum Ready. A more select set of individuals, teams, and organizations whose education, training, knowledge, skill, expertise, and experience in quantum effects, quantum technologies, and quantum computing are much more limited than the very elite, but much more sophisticated and useful than those who are merely Quantum Aware.

Training for quantum

  1. Academic education. Focus on theory and experimental work. Limited practical experience.
  2. Up-skilling current staff. Training current non-quantum staff for work with quantum. Primarily for Quantum Aware and Quantum Ready, but also for those who are intended to transition to a role which is primarily quantum.
  3. Ongoing up-skilling even for very skilled quantum staff. Every year there is something new to learn, even for the most elite of experts in quantum.
  4. Training technical staff who don’t have first-line responsibilities for quantum but whose responsibilities overlap with quantum. Such as IT staff who maintain all computing equipment, which might now include a mix of classical and quantum.
  5. Training non-technical staff who are impacted by quantum. Financial budgeting. Talent staffing. Training staff — those responsible for training in general and overall, which must now include quantum. General planning. No need for technical detail, but there are impacts of quantum. Managers. Executives.
  6. Technical training for staff whose main responsibility is quantum. All quantum all of the time.

Talent pool for quantum

  1. A very wide range of technical and managerial talent are relevant to working with quantum technologies. Ranging from minimal awareness to the most elite of experts.
  2. Extremely limited talent pool of the super-elite. Capable of doing conceptualization, design, and development from scratch for all algorithms and modules needed for a complete, complex quantum application.
  3. Very limited talent pool of the moderately elite.
  4. Limited talent pool of those with even modest quantum skills.
  5. Limited total count of skilled individuals.
  6. Limited total count of available skilled individuals.
  7. Excessive salary and other compensation and benefit requirements.
  8. Difficulty of attracting quantum talent.
  9. Difficulty of dislodging quantum talent from current positions.
  10. Difficulty of retaining quantum talent.
  11. How to amortize the efforts of the limited pool over a relatively large number of instances of each unique application.

Broader is better, but it can be too expensive

Significant hardware advances are needed

  1. Qubit fidelity. Near-perfect qubits are needed. Barely two nines at best currently. Really need four to five nines. Or at least 3.5 nines. Or three nines at an absolute minimum.
  2. Qubit connectivity. Nearest neighbor connectivity and SWAP networks just don’t cut it.

Significant research is required

Quantum-inspired solutions are great, but they don’t advance quantum computing itself

Intellectual property (IP) — is a proprietary interest helpful or harmful?

  1. Trade secrets and onerous patent protection can inhibit widespread adoption. It’s the potential for widespread adoption which incentivizes generalized applications in the first place.
  2. But the outsize potential for revenue and profits from a few key applications also incentivizes generalized applications. A few key applications could attract outsize investment and provide outsize benefits to those who use it. That could be a big net positive in some niche sectors.
  3. Moderate pricing of heavily-protected configurable packaged quantum solutions could still incentivize widespread adoption. Strict IP protection is not an absolute deal killer, provided that the pricing is considered reasonable and fair.
  4. It will be a horse race between open source and proprietary IP. Maybe the initial generalized applications will indeed be IP-based to prove the concept. But then maybe wider adoption for a wider range of applications will trend toward open source.
  5. Patents can frequently be worked around. Open source projects can generate the needed enthusiasm to generate workarounds.
  6. But some patents could be too complex or too onerous to work around. Hopefully few and far between. Not the general case.
  7. Generally patent holders wish to widely license their IP. That’s a good thing.
  8. But sometimes patent holders and owners of trade secrets seek to prevent competition. That’s a bad thing.
  9. Academic institutions have an interest in technology transfer. Modern academic institutions frequently seek to protect and license their valuable IP. Sometimes they will in fact license their IP freely or even donate to open source projects. Sometimes they will freely license at a very nominal rate. But, sometimes they may license at an onerous rate and/or exclusively so that widespread use is discouraged. Again, a very mixed bag, but the hope is that they will be a net positive for configurable packaged quantum solutions.
  10. Sliding scales for licensing fees are encouraged. Sure, larger firms with much at stake can afford expensive licensing arrangements, but smaller firms, startups, freelance consultants, students, and academic institutions really do need free licensing for experimentation, prototyping, and low-volume use. Higher-use can still incur more expensive licensing.

Provisioning a configurable packaged quantum solution

  1. Local to the computer system running the classical portion of the application.
  2. Accessible via a local area network.
  3. Accessible in the cloud via the external Internet.
  1. Dedicated. For this one application.
  2. Shared. By any number of applications.
  3. Governed by a service level agreement (SLA).

Three stages of adoption for quantum computing: The ENIAC Moment, configurable packaged quantum solutions, and The FORTRAN Moment

  1. The ENIAC Moment. Very limited deployments. Relying on super-elite technical teams at only the most elite of organizations.
  2. Configurable packaged quantum solutions. Widespread deployments of a relatively few applications. Requires little quantum expertise.
  3. The FORTRAN Moment. Widespread development of custom applications. No longer requires super-elite technical teams and is no longer limited to only the most elite of organizations.

The FORTRAN Moment will herald widespread full-custom quantum algorithms and applications

  1. Advanced and high-level programming models.
  2. Higher-level algorithmic building blocks.
  3. True higher-level quantum programming languages.
  4. Application frameworks.
  5. Libraries.
  6. Perfect logical qubits based on full, automatic, and transparent quantum error correction (QEC).
  1. Remove the remaining obstacles and impediments to designing and developing custom quantum algorithms and applications.
  2. But… still require at least some degree of skill and expertise at design and development. Classical computing skills should be sufficient, but required.
  3. Open the door for much more widespread custom quantum algorithm and application development.
  4. But the need, benefits, and opportunities for configurable packaged quantum solutions will remain and remain dominant. Configurable packaged quantum solutions will likely remain dominant. Half to 80% of organizations may still prefer configurable packaged quantum solutions even though custom quantum algorithms and applications are now within reach.
  5. Custom quantum algorithms will expand the richness of applications in a typical organization.

Summary and conclusions

  1. Developing quantum applications is and will continue to be a real challenge, for years to come.
  2. Most organizations won’t have the resources or specialized expertise to develop quantum applications in-house for years to come.
  3. Generalized applications can be applied to a wider range of situations for a wider range of users and customers.
  4. The resources and costs of generalized applications can be amortized across a wider range of users and customers.
  5. A generalized application can be made more robust and resilient, more reliable.
  6. A generalized application can be readily and simply deployed without the need to develop or modify any code.
  7. No elite or even quantum team is required to deploy or use such a generalized application.
  8. A generalized application can be made even more flexible by making it configurable.
  9. A configurable packaged quantum solution is simply such a generalized quantum application which is configurable. And resilient. And ready to go.
  10. Turnkey solution. Everything is in place. No coding required. All of the necessary code has already been written and thoroughly tested. Just supply input data, some parameter settings, and possibly an application configuration description.
  11. Generative coding. Quantum circuits will not be handcrafted and hardwired into the application as might be the case for simple or toy or experimental applications, but automatically generated dynamically by classical code based on the input data, input parameters, and configuration data.
  12. Automatic scaling of algorithms. No manual intervention or technical expertise required. The joys of generative coding.
  13. Significant hardware advances are needed. Current quantum hardware is woefully inadequate. Qubit fidelity and qubit connectivity need dramatic improvement. Near-perfect qubits are required.
  14. Significant research is required. Hardware, qubit technology, architecture, control software, algorithms, programming models, applications, application frameworks, etc. You name it, it needs research.
  15. Quantum-inspired solutions. Benefit from the same overall approach, but don’t require hardware advances. But they don’t advance quantum computing itself per se — they’re still only classical computing and don’t offer an exponential speedup or any dramatic quantum advantage.
  16. No specialized knowledge, training, or expertise is required to work with a configurable packaged quantum solution.
  17. Insulate users, customers, and classical application code from any knowledge or dependence on any of the concepts of quantum computing.
  18. Full explainability. The decisions leading to application results are fully reported and fully explained. Reports exactly what it did and why. Leaves little to the imagination. Not the raw quantum computing per se or the physics, but the overall process and the math but in application terms.
  19. Configurable packaged quantum solutions won’t become available until after The ENIAC Moment. It could take years to transition from The ENIAC Moment for quantum computing (first demonstration of a production-scale practical real-world quantum application) through all of the evolutionary stages to what really does appear to be a true configurable packaged quantum solution. Or, it could happen in just a few months.
  20. The application that results in The ENIAC Moment won’t necessarily be a configurable packaged quantum solution, but it could provide the basis or inspiration that eventually leads to the first true configurable packaged quantum solution.
  21. Once configurable packaged quantum solutions become available, quantum computing can become widely available to a wide range of users and organizations.
  22. Users and organizations using configurable packaged quantum solutions will achieve the benefits of quantum computing without the need for any specialized quantum knowledge, training, or expertise. They won’t have to compete in the marketplace for pricey and difficult to find quantum talent.
  23. No need to be Quantum Ready. Or even Quantum Aware. All of the quantum magic is hidden deep under the hood.
  24. Proprietary intellectual property (IP), including trade secrets, copyrights, and patents, can be a double-edged sword, sometimes helpful and sometimes harmful, although generally merely an annoyance. It will be a very mixed bag, sometimes good, sometimes bad, sometimes great, and sometimes ugly and terrible. If done with care, it can be a great net positive. If done inappropriately, it can be a horrific net negative. Going into all of the details is beyond the scope of this informal paper. The only point here is to raise awareness.
  25. Widespread design and development of full-custom quantum algorithms and applications won’t come for most organizations until after The FORTRAN Moment is reached, with much richer and much higher-level programming models, algorithmic building blocks and true high-level quantum programming languages — and perfect logical qubits supported by full, automatic, and transparent quantum error correction (QEC).
  26. When will configurable packaged quantum solutions be widely available? Unfortunately, not for at least a few years, at best. Maybe three to five years. Maybe seven years. Four to five years might be a good bet. But with quantum, nothing is a truly safe bet.
  • Caveat: This paper proposes and envisions a potential future for quantum applications which would greatly facilitate the adoption of quantum computing by the widest possible audience and market. There is no guarantee that this envisioned future will come to fruition, nor is there any certainty or confidence as to when it might come to fruition. The only confidence is that it won’t happen in the very near future (next year or two), and is likely to take at least a few years to develop — and require significant progress in the advancement of quantum computing hardware.
  • Counter-caveat: The proposal of this paper may not be a certainty but it certainly is a golden opportunity.

--

--

--

Freelance Consultant

Love podcasts or audiobooks? Learn on the go with our new app.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Jack Krupansky

Jack Krupansky

Freelance Consultant

More from Medium

Quantum Key Distribution Protocols

What I Learned From Fire Opal…

The Hubbard Model & Bose-Hubbard Model

Quantum Mechanics