About 12 invitees per domain, about 60 in total

List of invitees for Astronomical Science (AST) 

Andy Connolly (U Washington - Time Domain)

Stella Offner (UT Austin - ISM / Star Formation - CosmicAI)

Yuan-Sen Ting (Ohio State - Stellar Astrophysics / Galaxy Evolution)

Eric Ford (Penn State - Exoplanet)

Risa Weschler (Stanford - Local Group)

Brice Menard (Johns Hopkins - Science for AI)

Aggelos Katsaggelos (Northwestern - CS - SkAI)

Uros Seljak (Berkeley - Cosmology)

Ann Zabludoff (Arizona - Galaxy Evolution / Retrieval Systems)

Francisco Villaescusa-Navarro (CCA - Cosmology)

Benjamin Wandelt (CCA/John Hopkins - Reionization)

Peter Melchior (Princeton - Cosmology)

Salman Habib (Argonne - Cosmology - National Lab - SkAI)

Balancing institutional representation and subdomain expertise

More details this afternoon

Investment in AI remains small compared to the amount spent already on instrumentation

Rubin Observatory and Roman Space Telescope will create a data deluge

A critical timing to maximize scientific returns from billion-dollar investments 

Revolutionizing high-stakes science through novel statistical methods via deep learning

Sharpened constraints in cosmology, gravitational waves, and exoplanet atmospheres, etc.

Simulation-Based Inference transforms the "inverse problem" - using generative models as a surrogate for likelihood / posterior

Bayesian statistical inferences without simplistic human-heuristic summary statistics (e.g., power spectrum)

Real-time classification of transient events enables timely follow-up

Better distinction between planetary signals and stellar activity/noise for detecting Earth-like planets

Addressing the astronomical "needle in the haystack" problem

Multi-messenger and multi-modal (time series, images, spectra) data integration

Data processing : e.g., cosmic ray rejection

Image reconstruction and turbulence prediction for adaptive optics - e.g. for extreme large telescopes

Significantly reduce costs for ground and space-based telescopes

Reinforcement learning for instrumentation control
 to align multi-mirror telescopes and interferometric arrays

Neural network emulators replace expensive components,
such as sub-resolution physics for multi-scale simulations

Maintaining physical accuracy while achieving numerical speedups

GPU-powered numerical solvers reduce computational requirements

This enables more extensive parameter space exploration

Need for trust and acceptance from the astronomical community

Limited by simulation fidelity when modeling multi-scale processes and exposed to unknown systematics

Requires serious implementation efforts, not just conceptual hype

Develop benchmark datasets using simulations and real data, implement blind data analysis protocols 

Better dark energy constraints ≠ solving the dark energy problem

Plenty of optimization but few surprising discoveries

We need rigorous benchmarking, not just conceptual hype

Multi-agent (large language models) systems, and their
reasoning capabilities, might offer high-risk, high-reward pathway 

Minimal ethical concerns or commercial constraints

Safe, low-risk ecosystem, with high-stakes science

Interesting domain shifts (systematics) between training and test sets, enables study of generalization under realistic constraints

Complex, non-proprietary data spanning large dynamic ranges

Features unique modalities not in mainstream AI research -
e.g. foundational models for spectral analysis

Combines imaging, spectroscopy, time series, and theoretical models

Operating in a low signal-to-noise limit and complex/ heterogeneous noise characteristics requires advanced handling

Balances physical constraints with discovery potential

Ideal testing ground for incorporating physical laws

Drives innovation in encoding symmetries and conservation laws

Exceptionally accessible through arXiv and NASA ADS

Well-curated with long time baselines

Accompanied by publicly available data and code

Multi-agent (large language models) systems for intersectional knowledge domains