Artificial intelligence is increasingly embedded in scientific research as a methodological agent rather than a mere tool. In fields like medicine and public health, AI systems now participate directly in hypothesis generation, data exploration, and the stabilization of knowledge claims. Within the longue durée of scientific method - from empiricism and statistical inference to computational and data-intensive science - these developments raise the question of whether contemporary AI represents a genuine epistemic discontinuity through its capacity to identify patterns and guide inquiry prior to explicit human theorization.

We adopt a conceptual and analytical approach grounded in philosophy of science and AI ethics. Historical episodes of scientific discovery and methodological change, alongside current applications of AI in biomedical and public health research, are examined to assess continuities and ruptures in scientific practice. This provides a basis to critically analyze how AI reshapes the relationship between prediction and explanation, as well as the ethical and epistemic implications of this transformation.

AI-driven methods enhance predictive performance and exploratory capacity, often identifying empirical regularities before causal or mechanistic explanations are established. While this accelerates discovery, it also heightens concerns about epistemic opacity, automation bias, and an illusion of understanding when prediction substitutes for explanation. We highlight risks related to insufficient external validation across populations, dual-use and privacy issues, and inequities in access to AI resources. In response, we propose a framework for responsible AI integration in science: transparent reporting standards, interpretable modeling in high-stakes contexts, rigorous external validation, and clear institutional accountability.

AI has the potential to substantially accelerate and broaden scientific discovery. Realizing these benefits, however, requires governance that preserves methodological pluralism, sustains human judgment, and reinforces the norms of reproducible, trustworthy science. Future research and policy should align AI innovation with epistemic responsibility and social equity, ensuring that advances in predictive capacity translate into durable scientific and public health gains.