Artificial Scientific Discovery of advanced Quantum Hardware with high-performance Simulators

The experimental capabilities to generate, control, and measure complex quantum systems have dramatically improved over the last few years.

With ingenious proposals for practical applications, a powerful new paradigm of technology emerges that exploits quantum superposition and quantum entanglement.

As these systems get more advanced, designing new quantum experiments and hardware becomes ever more intricate for human scientists.

To exploit the full potential of quantum physics, researchers have started to involve artificial intelligence in the automated design of quantum experiments.

Unfortunately, even the currently most powerful methodologies have severe limitations and therefore cannot cope with the enormous potential that quantum mechanics promises us.

For that reason, in ARTDISQ, I propose to build high-performance physical simulators which are at the heart of all AI-driven discovery and design efforts.

The key idea is to use a framework originally developed for the efficient training and execution of large neural networks, called JAX. JAX is powerful enough to encode not only neural networks but a wide range of computer algorithms.

It allows for modern high-performance computational techniques such as just-in-time compilation, auto-differentiation and direct access to the GPU.

In ARTDISQ I will exploit this dramatic acceleration, which will open previously unchartered applications, including AI-driven (1) design of highly sensitive and robust quantum-enhanced gravitational wave detectors(2) design of new quantum techniques for imaging systems, with a focus on optical telescopes(3) hardware-software co-design for quantum hardware with the potential of discovering highly unorthodox and exotic solutions with superior behaviour.

As such, ARTDISQ aims to develop a revolutionary way to augment human researchers' ingenuity and creativity to accelerate scientific discoveries.