Two Tracks in the House That Halmos Built

Trip 6 to 8 August 2025 · Written up 8 August 2025 · Last expanded 29 September 2026

From Berkeley I went straight into the Quantum AI Conference at Indiana University Bloomington, which I am fairly sure was also the first of its kind, and which is where the trip went from informative to mind-bending.

The conference was bifurcated cleanly into two tracks: Quantum for AI asked how quantum modules inside a hybrid HPC could serve big-lab science, and AI for Quantum ran the other direction. Having just spent a week on AI4Science at TPC25, I spent most of my time in the latter, which is table-setting for the day we can exploit the distinctive features of quantum machines (entanglement chief among them) instead of simulating around them.

NVIDIA came to the Midwest on purpose

NVIDIA’s Monica VanDieren presented on how transformer-based language models trained on synthetic data can reduce error-correction runtime for quantum computers, with a tour of the relevant CUDA-Q features. (Our team at GT implemented that same synthetic-data pattern over a year ago for a related problem, which was a pleasant thing to hear validated from a stage.)

If you recall, at GTC25 Jensen announced a set of CUDA varieties segmented by industry vertical, as part of a deepening moat strategy for the developer and research ecosystem. Seeing NVIDIA run this roadshow specifically at midwestern researchers is the same strategy at ground level, getting the infrastructure adopted where the research is and not just where the capital is. It is a sound entrenchment play, and it will be worth watching whether it works.

The building

The focus on epistemology at this conference felt fitting in retrospect, because we were literally in the house that Halmos built, the Indiana University Department of Mathematics.

Two of the presenters turned out to be authors whose books I had bought years ago without ever connecting the names, and one of them is a foundational figure in the categorical approach to quantum physics, which does something to the atmosphere of a hallway conversation. That happens more often than you would think at the small first-of-their-kind conferences, and it is most of why I keep going to them instead of waiting for the proceedings.

Who I met

• Bob Coecke. A central figure in the categorical and diagrammatic approach to quantum foundations. He gave two talks: the first on the foundations work, the last on generating music with quantum mechanics, exploring how something like entanglement might be used to analyse the structure of classical music. Whether that link is literal or only as-if, his willingness to run the analogy out loud in front of a room was the part that stood out. I had put a similar reasoning-in-music question to a researcher at ISMIR the previous November, and here it was, reframed through quantum theory. A brief conversation over lunch opened another thread: could the logical reasoning that neuro-symbolic approaches and RAG-style frameworks currently supply find an alternative implementation in quantum-native computational structure? That is not an idle question. It reframes the architectural assumptions you make before writing a line of a reasoning system.
• Goren Gordon. Presented his work on embedding curiosity into computational systems, including a quantum-computable extension. It reads as an extension, or maybe a recast, of traditional search procedures inside reinforcement learning, and it blurs the line between cognitive architectures and physics. Provocative. I have been working through as many of his papers as I can since.
• Monica VanDieren, NVIDIA. The synthetic-data and error-correction talk above, and the CUDA-Q tour that came with it.

What I took away

Innovative science often comes from stepping outside a formalism and treating the framework itself as an object to manipulate, question and remix, as opposed to extending it inside its own boundaries. Both Coecke and Gordon were doing that in public, in different directions, in the same building, and that is the reason to sit in the room.

Photographs

Links

The conference programme, Coecke’s foundations work, Gordon’s curiosity papers, and the GT synthetic-data write-up. To be added.

Open entry
This page is not finished, and is not meant to be. I add photographs, people and links as I dig them out, so it keeps growing after publication. Why the whole site works this way is on the colophon.