Innovation emerges most powerfully at the intersections where disciplines overlap, challenge each other, and ultimately reveal deeper structures. Our research spans five foundational verticals: quantum, computation, life, intelligence, and cosmology. We don't perceive them as isolated domains, but layers of a continuous spectrum of emergent systems, ranging from the subatomic to the universal. Each level defines and constrains the next, with rules at smaller scales shaping the behavior of larger systems, and in some cases, the macro informing the micro through principles of inference, simulation, and control.
At the quantum level, we investigate the building blocks of physical reality and the rules that govern their behavior. But quantum mechanics doesn't stop at particles, it informs the design of quantum computers, machines that transcend classical computation. The interpretation of quantum mechanics, especially the role of observers, directly intersects with the study of intelligence and agency. How an observer collapses a wavefunction raises deep questions about consciousness and measurement, linking quantum theory with theories of mind and machine.
In computation, we explore both the limits and the creative potentials of algorithms. Universal computability, computational and algorithmic complexity theory frame what is computable in our universe. But what happens when we replace abstract automaton with embodied agents or physical systems? Concepts from artificial life, such as self-replicating automata, minimal genomes, and digital evolution, challenges our naive understanding of life via computational processes in software and hardware.
Artificial general intelligence occupies the crossroads of computation and life, where adaptive agents must not only solve tasks but also modify their internal structures and goals. Reinforcement learning, a framework born in psychology and matured in machine learning, models how intelligent systems learn from interaction with an environment—paralleling the adaptive behaviors seen in biological evolution and neural computation.
Finally, cosmology provides the ultimate testbed for our theories. Quantum cosmology, for instance, seeks to apply quantum principles to the entire universe, demanding new mathematical frameworks, and possibly new notions of observers and information. The structure of the cosmos may encode computational limits, inform the conditions necessary for life, and even constrain the emergence of intelligence.
By connecting these verticals, our research aims to uncover unifying principles, patterns of emergence, information, and agency, that cut across scale and substrate. This synthesis is what fuels us - where the quantum informs the computational, where computation simulates life, where life bootstraps intelligence, and where intelligence reflects back on the cosmos that gave rise to it.
The fidelity of the decomposition of quantum algorithms, represented as unitary matrices, to bounded depth quantum circuits depends strongly on the set of gates available for the decomposition routine. The developed software, called YAQQ (Yet Another Quantum Quantizer), enables the discovery of an optimized set of quantum gates. To identify these gate sets, we use the novelty search algorithm, circuit decomposition techniques (like Solovay-Kitaev, Cartan, and quantum Shannon decomposition), and stochastic optimization to implement YAQQ within the Qiskit quantum simulator environment. Consequently, we demonstrate pragmatic use cases of YAQQ in comparing transversal logical gate sets in quantum error correction codes, designing optimal quantum instruction sets, and compiling to specific quantum processors.
link to articleQuantum architecture Search (QAS) is a promising direction for automated design of quantum circuits. QAS techniques like Multi-Layer Perceptron (MLP)-based deep Q-networks remains challenging to interpret due to the large number of learnable parameters and the complexities in selecting activation functions. In this work, we overcome these challenges by utilizing the Kolmogorov-Arnold Network (KAN) in the QAS algorithm, demonstating their efficiency in the task of quantum state preparation and quantum chemistry, in noiseless and noisy scenarios. KAN outperforms MLPs in fidelity when approximating states, showcasing its robustness against noise. In quantum chemistry problems, we enhance the recently proposed QAS algorithm by integrating curriculum reinforcement learning with a KAN structure, facilitating a more efficient design of parameterized quantum circuits by reducing the number of required 2-qubit gates, circuit depth and learnable parameters compared to MLPs.
link to article