00:0001:11
Why Quantum plus AI
Introduces the QC+AI field and frames it through the practical limitations of NISQ hardware.
Curated chapter summary for local development. Full transcript alignment can replace this field later.
Hardware-constrained learning for quantum computing and artificial intelligence
Lesson
Uses routing, RL-tuned augmented Lagrangian methods, and graph shrinking to show how classical intelligence creates viable interfaces to limited quantum hardware.
Video Lesson
Introduces the QC+AI field and frames it through the practical limitations of NISQ hardware.
Curated chapter summary for local development. Full transcript alignment can replace this field later.
Explains that current systems are constrained by noise, routing overhead, and limited qubit budgets.
The video emphasizes that current hardware cannot absorb raw problem formulations without strong classical assistance.
Focuses on code compilation, qubit routing, and the physical cost of mapping logical circuits to sparse hardware graphs.
Compilation overhead is presented as a decisive engineering constraint rather than a software afterthought.
Covers hybrid optimization loops and the role of classical search and learning around quantum subroutines.
The recurring message is that classical intelligence often protects fragile quantum steps from infeasible search spaces.
Closes with application examples and a general argument for hybridization as the practical path in the NISQ era.
The ending connects routing, noise mitigation, and application design into a coherent hybrid systems view.
Key ideas
Source-grounded sections
Ali, Chicano, and Moraglio (Eds.), QC+AI 2025 Proceedings
While the preceding sections detailed the application of quantum enhancements for classical AI tasks, the physical advancement and operationalization of quantum computing itself requires the deployment of highly sophisticated classical AI algorithms—the domain of AI4QC.1 A primary engineering bottleneck in NISQ execution is the rigid constraint of device topology
While the preceding sections detailed the application of quantum enhancements for classical AI tasks, the physical advancement and operationalization of quantum computing itself requires the deployment of highly sophisticated classical AI algorithms—the domain of AI4QC.1 A primary engineering bottleneck in NISQ execution is the rigid constraint of device topology. Two-qubit operations, such as the CNOT gate, can only be executed between qubits that are physically connected via microwave or optic
Ali, Chicano, and Moraglio (Eds.), QC+AI 2025 Proceedings
To apply the NMCS framework, the environment dynamics must be rigorously defined
To apply the NMCS framework, the environment dynamics must be rigorously defined. The state space at any given time step is defined mathematically as .1 This encompasses the current injective mapping of logical to physical qubits (), the set of unscheduled gates (), the set of nodes currently locked by ongoing gate operations (, managed via mutex locks to handle variable execution times), and the static device topology graph ().1 The action space represents the set of SWAP gates that can be sch
Ali, Chicano, and Moraglio (Eds.), QC+AI 2025 Proceedings
The NesQ and NesQ+ algorithms were benchmarked against a suite of industry-standard routing frameworks: Google's Cirq, IBM's Qiskit (evaluating basic, stochastic, and SABRE variants), Cambridge Quantum Computing's tket, and a recent Graph Neural Network (GNN) guided MCTS framework known as Qroute.1 Across 30 dynamically simulated random circuits containing between 30 and 180 gates, NesQ achieved an average output depth 48.75% lower than Cirq, 32.57% lower than Qiskit SABRE, 30.42% lower than tk
The NesQ and NesQ+ algorithms were benchmarked against a suite of industry-standard routing frameworks: Google's Cirq, IBM's Qiskit (evaluating basic, stochastic, and SABRE variants), Cambridge Quantum Computing's tket, and a recent Graph Neural Network (GNN) guided MCTS framework known as Qroute.1 Across 30 dynamically simulated random circuits containing between 30 and 180 gates, NesQ achieved an average output depth 48.75% lower than Cirq, 32.57% lower than Qiskit SABRE, 30.42% lower than tk
Notes
Loading saved notes...
Source assets
Related lessons
Frames the course around NISQ-era limits and the distinction between using quantum methods for AI versus using AI to make quantum computing operationally useful.
Shares core themes in graph methods, optimization, reinforcement learning.
Open related lessonSynthesizes the source corpus around resource efficiency, memory cost, and the broader systems view of hybrid QC+AI.
Shares core themes in graph methods, optimization, reinforcement learning.
Open related lessonSurveys application families in which quantum layers operate as targeted representational or decision-making components rather than total model replacements.
Shares core themes in graph methods, reinforcement learning, representation.
Open related lessonAli, Chicano, and Moraglio (Eds.), QC+AI 2026 Proceedings
Moving from AI applications of quantum theory to the direct application of quantum hardware for logistics, the work of Monit Sharma and Hoong Chuin Lau on the Capacitated Vehicle Routing Problem (CVRP) presents a critical advancement in hybrid algorithms.1 The CVRP is a strongly NP-hard combinatorial benchmark central to supply chain management
Moving from AI applications of quantum theory to the direct application of quantum hardware for logistics, the work of Monit Sharma and Hoong Chuin Lau on the Capacitated Vehicle Routing Problem (CVRP) presents a critical advancement in hybrid algorithms.1 The CVRP is a strongly NP-hard combinatorial benchmark central to supply chain management. Mapping its stringent inequality constraints (such as vehicle capacity and Miller-Tucker-Zemlin subtour elimination) to a QUBO formulation traditionally
Ali, Chicano, and Moraglio (Eds.), QC+AI 2026 Proceedings
Addressing the same foundational problem of NISQ hardware limits, a parallel study by Sharma and Lau introduces "Learning-Based Graph Shrinking for Quantum Optimization of Constrained Combinatorial Problems".1 For dense combinatorial problems like the Maximum Independent Set (MIS) and the Multi-Dimensional Knapsack Problem (MDKP), the input graphs must be significantly reduced in size (shrunk) before they can be mapped to limited, noisy qubit topologies.1 The conventional methodology for graph
Addressing the same foundational problem of NISQ hardware limits, a parallel study by Sharma and Lau introduces "Learning-Based Graph Shrinking for Quantum Optimization of Constrained Combinatorial Problems".1 For dense combinatorial problems like the Maximum Independent Set (MIS) and the Multi-Dimensional Knapsack Problem (MDKP), the input graphs must be significantly reduced in size (shrunk) before they can be mapped to limited, noisy qubit topologies.1 The conventional methodology for graph
RAG Q&A