I have a background in physics and am the primary author of Quantum Computing and Information: A Scaffolding Approach (link). To enhance my teaching and research, I sought to upgrade my knowledge and skills with the latest technologies, particularly in practical quantum programming. This led me to participate in the two-week-long Qiskit Global Summer School (QGSS 2024). It was truly an enriching learning experience. I sincerely thank the lecturers, organizers, and IBM Quantum for offering this opportunity and for their contribution to the quantum computing community. However, I also have some feedback for future improvements.

QGSS 2024

The theme of QGSS 2024 was "Path to Quantum Utility": "This year’s Qiskit Global Summer School virtual event will give you the tools you need to begin running utility-scale quantum experiments that push the limits of computation."

Some of the main advantages of the program included:

  • High-quality lectures and labs
  • Excellent organization
  • Free of charge
  • Complimentary access to IBM's 127 and 133 qubit machines

Initially, over 2000 students attended the lectures. As of this writing (08/30/2024), one day from closing, 800 students have completed all four labs.

Course Contents

The two-week program comprised a series of intensive activities designed to provide a comprehensive understanding of quantum computing and Qiskit programing.

Lectures

The nine major lectures covers a broad spectrum of topics:

  1. Introduction to Qiskit and Runtime Primitives V2: An overview of Qiskit and its latest runtime primitives, essential for efficient quantum computing.
  2. Quantum Circuit Compilation with Qiskit: Techniques for compiling quantum circuits using Qiskit, ensuring optimal performance.
  3. Hardware Noise: Modeling and Characterization: Methods to model and characterize noise in quantum hardware, a significant challenge in practical quantum computing.
  4. Execution on Noisy Quantum Hardware: Fighting Errors Before Fault Tolerance: Strategies to manage and mitigate errors in noisy quantum environments before achieving fault tolerance.
  5. Workflows for Quantum-centric Supercomputing: Design and implementation of workflows tailored for quantum-centric supercomputing.
  6. Mapping Problems to Qubits: Techniques for effectively mapping computational problems to qubits.
  7. Quantum Combinatorial Optimization: Approaches to solving combinatorial optimization problems using quantum algorithms.
  8. Hamiltonian Dynamics: Applications and Simulation: Understanding and simulating Hamiltonian dynamics for various quantum applications.
  9. Quantum Machine Learning: Exploring the intersection of quantum computing and machine learning, and its potential applications.

Each lecture was followed by a Q&A session, allowing participants to clarify concepts and engage with the material more deeply. The speakers were exceptionally knowledgeable, and the lecture quality was notably higher than in previous years. These sessions provided an interactive platform to discuss complex topics and gain insights from experts in the field.

Laboratory Sessions

The course also included four hands-on lab sessions, each designed to reinforce the concepts covered in the lectures through practical application:

  1. Quantum Circuit Transpilation: Techniques for converting high-level quantum algorithms into executable circuits.
  2. Utility-Scale Layer Fidelity Experiment: Experiments focused on assessing and improving the fidelity of quantum layers at a utility scale.
  3. Quantum Error Suppression and Mitigation with Qiskit Runtime: Implementing strategies for error suppression and mitigation using Qiskit's runtime environment.
  4. Simulating Nature at Utility Scale: Practical exercises in simulating natural phenomena using quantum computing.

Each lab was provided as a Jupyter notebook, containing up to seven exercises for participants to complete and submit for autograding. To receive the final certificate, each participant was required to complete all four labs successfully.

Areas to Improve

While the lectures were well designed and delivered, there are areas in need of improvement. From an educational standpoint, some elements lack integration (or scaffolding support). The labs often have learners going through the motions of programming without fully understanding the underlying quantum computing principles. This issue is particularly evident in Lab 4, the capstone project.

Integration (or Scaffolding)

Lab 4 simulates the dynamics of a large (50-site) Heisenberg spin chain. A primary issue is that most learners did not understand the physics of the Heisenberg spin chain, and no tutorials or good references were provided. This information is also hard to derive from the lectures.

The lab appears to have been designed primarily by Qiskit programming staff, as several aspects seem mysterious from a physics perspective. This negatively affects the course outcome, as the focus is placed on programming skills without understanding the underlying physical principles, which is dangerous in real-world research and engineering.

Here are a few specific examples:

  • The stated goal of the lab is "to measure the dynamics of Zᵢ for a given site as a function of time and external magnetic field h for two different phases (isotropic and anisotropic) of the spin chain." However, the labs actually just calculate the quantum state of the chain at a single time δt, starting from |0⟩. The varying parameters are two anisotropy values (Δ = 1, 5) and 12 h values.
  • Mysteriously, h goes from 0 to π/2. To see the spontaneous spin-spin coupling of the chain overcome by the magnetic field, I expect h to go beyond the maximum value of Δ, which is 5.
  • The lab instructions state that "we will use δt = 5π/4 in order for the ZZ gates to rotate by -π/2." As I understand, the rotation of the ZZ gate is 2Δδt. Then δt should be π/4 for Δ = 1, and π/20 for Δ = 5, to achieve a ZZ rotation of -π/2. But more importantly, how do learners know that ZZ(-π/2) corresponds to sufficiently small steps for Trotterization?
  • The lab asks learners to analyze and present the data in two plots, for the isotropic (Δ = 1) and anisotropic (Δ = 5) phases, respectively. Each plot is an average ⟨Z⟩ for all spins versus the external magnetic field h. But the magnetic field is given in the x direction, for which ⟨X⟩ should be measured to see its aligning effect. The ending time δt is fixed at 5π/4, as mentioned above. Given these conditions, I doubt many could make sense of their simulation results.

Support

In a college setting, especially those that promote excellence in teaching (as opposed to research as primary focus), classes are supported by professors, TAs, staff, and the community. In comparison, QGSS 2024 could benefit from dedicated support staff, akin to TAs.

For example, Lab 1 exercise 2c asks learners to construct a transcompiler "to reduce the QFT quantum circuit size after transpilation by at least 20% compared to Qiskit's default transpilation." A confusion arises because learners can successfully pass this exercise, yet a local test routine shows the reduction is nowhere near 20%. It turns out the local test is against an any-to-any coupling map while the exercise grading check uses a linear coupling map. I believe this is due to an error in the lab design. It can be fixed by one line of code:

qk_qc = generate_preset_pass_manager(2, coupling_map=cm, basis_gates=backend.operation_names).run(qc)

Many learners were hindered by this problem, and there were many posts on the forum. Yet, to my knowledge, no one officially acknowledged the problem or offered a timely solution.

Another example is that Qiskit compiles the QFT quantum circuit into a "harp" structure instead of the "standard textbook" structure. There was a question on the forum about this, and someone said the harp structure produces a more efficient low level circuit. However, no one provided a timely explanation or reference as to why that is the case.

End Remark

Overall, QGSS 2024 was a resounding success. The effort put into this program represents a significant contribution to the advancement and community of quantum computing. The high-quality lectures, well-organized labs, and access to cutting-edge quantum hardware provided an invaluable learning experience for all participants. However, there are areas for improvement, particularly in enhancing the integration of educational content and providing better scaffolding to ensure a deeper understanding of quantum principles.

We look forward to an even more successful QGSS 2025, with the hope that the feedback provided will help refine and elevate the program further. Enhancements in tutorial support, clearer explanations of complex concepts, and more interactive problem-solving sessions could make future iterations even more impactful. By addressing these areas, QGSS can continue to set the benchmark for excellence in quantum education and inspire the next generation of quantum researchers and practitioners.

Thank you once again to the organizers, lecturers, and IBM Quantum for their dedication and hard work in making QGSS 2024 a truly memorable and enriching experience.