Wayne State College

While generative artificial intelligence (GenAI) technologies are revolutionizing content production, they pose serious privacy and data security issues. The potential for privacy violations, biases, and cyberattacks rises as these models process large datasets containing sensitive or private data. This book examines these issues, especially cybersecurity, healthcare, and finance. The potential for GenAI models to reproduce or infer sensitive data from training datasets is a significant problem that raises ethical and intellectual property issues. Data protection techniques like encryption, tokenization, and anonymization are crucial to reducing these dangers. This study assesses the efficacy of these techniques by looking at how they affect the functional performance and privacy risk reduction of GenAI systems. Through experimental analysis, it evaluates the impact of tokenization and anonymization on a state-of-the-art large language model (LLM). Empirical results offer insights into the trade-offs between protecting model performance and data privacy using open-source tools such as Microsoft Presidio. The research aims to help create safe and morally sound GenAI applications, ensuring that AI advancements align with data security guidelines while preserving accuracy and efficiency in practical applications.

Abstract Functional brain imaging data can provide a window into the task-driven network states that shape brain function (or dysfunction). Conventionally, these network states can be represented as bivariate correlation matrices (which are formed from fMRI time series from multiple brain regions/nodes within any task window). Here, we treat these conventional connectivity matrices as connectivity terrains in order to recover local structure. In principle, any such terrain can be traversed node by node, where from any node, one can move towards its nearest functional neighbor (i.e., its maximally correlated node). In terrains with meaningful structure, such traversals across multiple nodes should converge to attractor nodes; here, the nodes that flow into a shared attractor form an attractor basin, which effectively is a sub-network within the system. Extant methods (e.g., degree distribution and characteristic path length) can summarize global network properties but cannot identify attractor nodes and basins. Here, we construct a new relation, called transitive maximal correlation (TMC) that can recover attractors and attractor basins in connectivity terrains. Node A is said to be transitively maximally correlated to node B if and only if B is an attractor into which A flows. We first develop the mathematical basis for deriving a TMC matrix TMC(M) from a bivariate correlation matrix M (before explaining this with hypothetical data). We next apply the TMC relation to connectivity terrains derived from real fMRI time series data, where these data were acquired in two distinct task-domains (that varied in their extent of cross-cerebral demand): i) associative learning and ii) visually guided motor control. We show that TMC is remarkably sensitive to inter-hemispheric structure in the connectivity terrain; here, attractor pairs that were inter-hemispheric homologues were more likely to be observed for the cross-cerebral learning task, than the more circumscribed motor-control data. We confirm the condition-specific sensitivity of TMC showing that observed attractor basins differed significantly across conditions of the learning task. Finally, we demonstrate that TMC complements graph theoretic constructions like path length and betweenness centrality. We suggest that TMC is a mathematically sound and novel method for capturing functional properties of brain networks.

Despite the dire need for rural surgeons and increasing interest in exposing residents to rural practice, there remains little systematic information about how programs establish and sustain rural rotations. A website review (12/2022-7/2023) of the 342 civilian surgery programs listed by the ACGME identified those offering elective or required U.S. rural rotations. In addition, all programs were contacted by email to confirm or deny that they offered rural rotations to account for webpage inaccuracies. The effort identified 81 programs (24%). A qualitative design was then adopted, whereby the first author interviewed 58 general surgery program leaders (72% participation rate) to explore strategies used to arrange and sustain rural rotations. Interviews focused on the 2023-24 academic year, were conducted from August 2023 to August 2024, and were recorded, transcribed, and analyzed collaboratively to discern program strategies. Programs are more likely to use rural training sites outside their healthcare system (59% or 34/58) than inside (41% or 24/58), but system sites were preferred when available. At both site types, programs drew heavily upon professional ties, primarily to program graduates (64% or 37/58). Alumni were eager to help and knew program traditions and expectations. Their presence also bolstered confidence in the educational value of the rotation. Additional alumni joining rural sites were evidence of rotations as a recruitment pathway, thus fostering sustainable relationships. Within-system rotations were easier to arrange and finance than those outside the system. The results offer practical insights into how a diverse mix of surgery programs across the U.S. select and sustain required and elective rural rotations in the U.S. The prominence of professional ties to program graduates and healthcare-system links are especially notable. Variability across programs in how they select and sustain sites also suggests that many approaches are feasible.

History offers no shortage of war, but precious few durable peaces. Mediation, for all its noble pretensions, has long been a function of power, exerted by empires, dictated by superpowers, and affected by geopolitical interests. This study interrogates the changing nature of mediation in Africa from 1975 to 2019, testing the illusion of a rules-based order replacing great-power intrigue. Analyzing peace agreements, we find that regional organizations and neighboring states have overtaken traditional global arbiters, as joint mediation has surged. The liberal school of inevitable institutional progress is punctured by reality: great-power interests never vanished; they adapted. Mediation is no longer the province of Cold War players, neither is it the triumph of cooperation over coercion. Instead, we see a paradigm in flux, one that will either stabilize or fracture under resurgent multipolar competition.

International audience

Astrophysical observations provide compelling evidence for gravitationally interacting dark matter in the universe that cannot be explained by the standard model of particle physics. The extraordinary amount of data from the CERN LHC presents a unique opportunity to shed light on the nature of dark matter at unprecedented collision energies. This Report comprehensively reviews the most recent searches with the CMS experiment for particles and interactions belonging to a dark sector and for dark-sector mediators. Models with invisible massive particles are probed by searches for signatures of missing transverse momentum recoiling against visible standard model particles. Searches for mediators are also conducted via fully visible final states. The results of these searches are compared with those obtained from direct-detection experiments. Searches for alternative scenarios predicting more complex dark sectors with multiple new particles and new forces are also presented. Many of these models include long-lived particles, which could manifest themselves with striking unconventional signatures with relatively small amounts of background. Searches for such particles are discussed and their impact on dark-sector scenarios is evaluated. Many results and interpretations have been newly obtained for this Report.