Welcome
Workshop on Quantum and Artificial Intelligence
Location: Haus Room (36-428), Research Laboratory of Electronics, MIT
50 Vassar St, Cambridge, MA 02139
(By invitation only)
Dates: May 18-19
Workshop Objectives
Quantum information science and engineering (QISE) and Artificial Intelligence (AI) both are highly interdisciplinary emerging disciplines that have tremendous potential of societal impact. The objective of this workshop is to explore a potential new MURI topic that will explore rigorous foundations connecting quantum information processing and artificial intelligence, and deepen the understanding of the confluence of ideas from both disciplines to realize high-impact applications in leading hardware platforms, such as:
(1) Superconducting quantum circuits, including unconventional architectures, using multimode bosonic codes, and quantum-annealing-specific platforms.
(2) Photonic quantum computing platforms, including ānon-conventionalā yet scalable architectures that exploit high-dimensional encodings combining spatial/spectral degrees of freedom; and possibly also less-explored ānot universalā architectures such as the OPO network based optimization engines, Boson Sampling and Gaussian Boson Sampling.
(3) Trapped ion quantum processors, capable of emulating Ising-like Hamiltonians, which could be used to encode AI-inspired optimization problems.
The workshop will bring together experts from computer science, computer networking, quantum optics, quantum computing, photonic quantum information systems, quantum information theory, quantum and classical algorithmic complexity theory, spin-based quantum systems, trapped ion and superconducting quantum systems, to explore the AI-inspired techniques in near-term realizable quantum architectures.
Topics to be covered
The following four broad topics will be covered at the workshop: one at each session. The sub-topics listed within each are representatives of the genres of discussion topics, and not explicit guidelines. Each of these four topics will have representation in the workshop and will contribute to developing the overall theme of the workshop.
The topical flow of the workshop will constitute the following. We will start with Topic 1, where we will discuss various classes of ML algorithms: CNNs, DNN, SVM, etc., and their application domains. This session will be open to both classical algorithms, and their quantum counterparts. In Topic 2, we will discuss the use of purely classical AI techniques for designing better quantum processors. This may include the use of AI for discovering complex quantum circuits for state preparation, state tomography, error correction and distillation codes, and more. Topic 3 will constitute the use of AI-inspired algorithms in the quantum-domain, e.g., on information embedded in the photonic domain or in other forms of electro-magnetic fields, to build quantum-powered intelligent sensors. Finally, in Topic 4, we will discuss algorithms for NISQ processors, including hybrid quantum-classical systems.
- Machine learning frameworks and applications: In this session, we will consider classical and quantum machine learning frameworks, algorithms for different application domains (for classical computers, and fault-tolerant quantum computers, respectively. Examples may include quantum versions / extensions of DNN, CNN, Reservoir Computing, SVM, etc., which for the quantum case would be applied to data encoded/stored in quantum memories, such as quantum Random Access Memories (qRAMs).
- Classical AI for quantum: This session will explore the use of classical AI to accelerate the development of better or more powerful quantum processors, including quantum architectures, quantum codes, bosonic quantum circuits for hard-to-produce quantum states, quantum tomographic methods and measurements, and more.
- Quantum intelligent sensors and networks: By pushing classical-inspired ML computations (e.g., SVM, DNN, ONN, Image-data classification algorithms) into the quantum domain, e.g., information-bearing photonic or magnetic fields, one can allow for quantum processing on the information-bearing fieldāe.g., in the context of optical sensing, magnetometry, opto-mechanical sensors, spectroscopy, etc.āthat are impossible in the post-detection electronic domain (i.e., after the shot noise), and hence can result in higher-efficacy classification, and inferences on optically-encoded information. There may be applications to lidars, imaging, hyper-spectral imaging, telescopes, space domain awareness, microscopy and more.
- Algorithms for NISQ processors: This session will explore near-term quantum processors, such as random samplers (e.g., Boson Sampling, Gaussian Boson Sampling, IQP), QAOA, Quantum Annealers, and other NISQ primitives as algorithmic accelerators in (otherwise classical) AI/ML algorithms, leading to hybrid quantum-classical special-purpose processors. Applications include but are not limited to image classification, graphical inference problems, graph similarity, and variational eigensolvers for energy-minimization problems.
Sponsored by the US Army Research Office (ARO)
This workshop will advance our knowledge at the intersection of quantum information and AI, and could lead to the development of transformative technologies for the Army’s mission. In particular, results from technical innovations may result in powerful active and passive quantum-enhanced photonic sensors for space-domain awareness applications, highly-sensitive and intelligent position, navigation and timing systems, special-purpose processors capable of solving complex optimization, and scheduling problems of relevance to the DoD, and the perhaps even the use of near-term quantum processors to enhance our understanding of classical artificial intelligence.