Hi all,
Next week, we have three QIT seminars:
(1) On Monday at 15:30 in HIT E41.1, our visitor Eduardo A. B. Oliveira will talk about “A spacetime-covariant framework for inertial and accelerated quantum clocks in first quantization: unitarity and quantum time dilation in the positive mass sector”.
(2) On Wednesday at 11:15 in HIT E41.1, our visitor Jonas Rigo will talk about “Is AI a Tool or a Collaborator? Building rigorous and auditable workflows for mathematical research”.
(3) On Friday at 14:00 in HIT E41.1, Antoine Guinchard will talk about “Circuit Cutting via Temporal Graph Partitioning”.
See below for the abstracts.
Best,
Ladina
%%%%%%%%%
Title:
A spacetime-covariant framework for inertial and accelerated quantum clocks in first quantization: unitarity and quantum time dilation in the positive mass sector
Abstract:
It is expected that a quantum theory of gravity will radically alter our current notion of spacetime geometry to account for quantum fluctuations. However, contrary to what was commonly assumed for many decades, these fluctuations could manifest in scales much larger than the Planck Scale, provided that there is enough coherence in a superposition of geometries. Quantum Clocks, i.e. quantum mechanical systems whose internal dynamics keep track of proper-time lapses, are a promising tool for probing such superpositions. In this talk, we present a spacetime-covariant approach to describe these clocks in first quantization, accounting for the possibility of dynamically accelerated clocks via suitable local couplings with external fields. We show that a particular decomposition of the (quadratic) clock Hamiltonians into positive and negative mass sectors, when attainable, enables one to compute the evolution of the system directly in terms of a reference clock's proper-time while maintaining explicit covariance. When this decomposition is possible, the evolution thus obtained is always unitary. We apply this formulation to compute the joint time evolution of pairs of inertial and (circularly) accelerated quantum clocks. In both cases, when our clocks follow quasiclassical histories, or "worldfunctions", we find a conditional time dilation observable whose probabilities peak exactly at the classical expected value. However, when they are prepared in coherent superpositions of such worldfunctions, we find a quantum-coherent time dilation profile which both displays proper-time interference and yields distinctively off-diagonal coherences, allowing e.g. the realization of operations in indefinite causal order.
%%%%%%%%%
Title:
Is AI a Tool or a Collaborator?
Building rigorous and auditable workflows for mathematical research
Abstract:
The release of increasingly capable AI systems has renewed claims that artificial general intelligence may be within reach. This raises a more immediate question for science: where does the human researcher fit into the loop?
I will discuss the capabilities of modern agentic AI systems and contrast them with purpose-built workflows for mathematical and scientific research. I will introduce a set of components for constructing AI-assisted workflows that are rigorous, auditable, and reproducible: grounding agents in trusted sources, representing knowledge and logical dependencies explicitly, and preserving the resulting research artifacts in a human-readable form.
The aim is to keep both the workflow and its outputs local, transparent, and independent of any particular AI provider. More broadly, I will argue that increasingly capable AI does not diminish the role of expertise; rather, it makes the design of the research process and the ability of the researcher to structure, inspect, and verify it, more important.
%%%%%%%%%
Title:
Circuit Cutting via Temporal Graph Partitioning
Abstract:
A major limitation of current quantum processors (QPUs) is the small number of qubits they contain. One way to overcome this limitation is to decompose a large quantum circuit into smaller subcircuits that can be executed on separate QPUs. Circuit cutting achieves this by replacing selected gates and wires (referred to as "cuts") with quasiprobability decompositions over local operations. The result of the original circuit can then be reconstructed through a weighted average over these decompositions. However, estimating this average introduces a sampling overhead that can grow exponentially with the number of cut gates and wires. It is therefore crucial to select these cuts such that they produce a feasible decomposition while minimizing the resulting sampling overhead.
In this presentation, we provide an introduction to circuit cutting and show that the joint selection of gate and wire cuts can be formulated as a temporal graph partitioning problem. In this formulation, qubits are represented by nodes and gates by edges, which can dynamically change to capture not only the qubits the gate is affecting but also its depth in the circuit. We also briefly present an algorithm based on this formulation to find circuit cuts with low sampling overhead faster than existing algorithms.