Artificial intelligence is rapidly transforming astronomical research, yet the scientific community has largely treated this transformation as an engineering challenge rather than an epistemological one. This perspective article argues that philosophy of science offers essential tools for navigating AI's integration into astronomy--conceptual clarity about what "understanding" means, critical examination of assumptions about data and discovery, and frameworks for evaluating AI's roles across different research contexts. Drawing on an interdisciplinary workshop convening astronomers, philosophers, and computer scientists, we identify several tensions. First, the narrative that AI will "derive fundamental physics" from data misconstrues contemporary astronomy as equation-derivation rather than the observation-driven enterprise it is. Second, scientific understanding involves more than prediction--it requires narrative construction, contextual judgment, and communicative achievement that current AI architectures struggle to provide. Third, because narrative and judgment matter, human peer review remains essential--yet AI-generated content flooding the literature threatens our capacity to identify genuine insight. Fourth, while AI excels at well-defined problem-solving, the ill-defined problem-finding that drives breakthroughs appears to require capacities beyond pattern recognition. Fifth, as AI accelerates what is feasible, pursuitworthiness criteria risk shifting toward what AI makes easy rather than what is genuinely important. We propose "pragmatic understanding" as a framework for integration--recognizing AI as a tool that extends human cognition while requiring new norms for validation and epistemic evaluation. Engaging with these questions now may help the community shape the transformation rather than merely react to it.
We test whether a representation learned from interferometric strain transients in gravitational-wave observatories can act as a frozen morphology-sensitive operator for unseen sensors, provided the target signals preserve coherent elastic transient structure. Using a neural encoder trained exclusively on non-Gaussian instrumental glitches, we perform strict zero-shot anomaly analysis on rolling-element bearings without retraining, fine-tuning, or target-domain labels. On the IMS-NASA run-to-failure dataset, the operator yields a monotonic health index HI(t) = s0.99(t)/tau normalized to an early-life reference distribution, enabling fixed false-alarm monitoring at 1-q = 1e-3 with tau = Q0.999(P0). In discrete fault regimes (CWRU), it achieves strong window-level discrimination (AUC_win about 0.90) and file-level separability approaching unity (AUC_file about 0.99). Electrically dominated vibration signals (VSB) show weak, non-selective behavior, delineating a physical boundary for transfer. Under a matched IMS controlled-split protocol, a generic EfficientNet-B0 encoder pretrained on ImageNet collapses in the intermittent regime (Lambda_tail about 2), while the interferometric operator retains strong extreme-event selectivity (Lambda_tail about 860), indicating that the effect is not a generic property of CNN features. Controlled morphology-destruction transformations selectively degrade performance despite per-window normalization, consistent with sensitivity to coherent time-frequency organization rather than marginal amplitude statistics.