Generative Models Shift Big Data Systems Toward Synthesis-Oriented Workloads

Generative artificial intelligence is increasingly being integrated into Big Data infrastructures, driving a fundamental transition in data-intensive systems away from purely extractive analytics and toward synthesis-oriented workloads. In a comprehensive review published in Neural Computing and Applications, researchers Elias Dritsas and Maria Trigka present a systems-oriented synthesis analyzing how generative models function as computational operators within distributed data environments. Their analysis specifically examines large language models, diffusion models, and probabilistic generators as active computational operators interacting directly with underlying data distributions, retrieval mechanisms, program synthesis pipelines, storage engines, query processors, and execution orchestration layers.

The authors grounded their survey in peer-reviewed scientific literature identified through structured search queries designed specifically to connect generative modeling techniques with Big Data system architectures. To systematically categorize and structure this broad technical landscape, the investigation employs concepts such as data lineage, retrieval-augmented generation, and system governance as its primary organizing principles.

The survey analysis was organized around three complementary dimensions that jointly evaluate the capabilities, integration methods, and reliability constraints of generative technologies in large-scale data infrastructures. The first analytical dimension examines the formal mathematical objectives and representative computational characteristics that define large language models, diffusion models, and probabilistic generators.

The second analytical dimension concentrates on concrete architectural mechanisms required to support generative processing across distributed data systems.

The third dimension addresses comprehensive system-level trust requirements that enterprise data architectures must enforce when adopting generative capabilities. According to the review, a trustworthy Big Data infrastructure must systematically encompass fairness across computational processes, rigorous data privacy protection, model and output interpretability, organizational and algorithmic accountability, and comprehensive provenance tracking.

Beyond overarching architectural frameworks, the review synthesizes current evidence regarding how generative artificial intelligence directly influences core database and analytics tasks. The authors demonstrate that generative models impact indexing mechanisms, query optimization techniques, data mining routines, information retrieval pipelines, and knowledge graph reasoning procedures.

In evaluating these systems, the survey examines performance from a multi-objective perspective, emphasizing the critical limitations of existing benchmarks that focus exclusively on computational efficiency. The authors argue that comprehensive evaluation frameworks must jointly consider predictive performance, system robustness, explanation fidelity, provenance coverage, privacy budgets, energy consumption, and governance overhead. Finally, the review outlines emerging research directions, focusing on provenance-aware generation, governance-aware deployment, and sustainability-aware scaling, while explicitly distinguishing established system capabilities from active research challenges in trustworthy generative Big Data environments.

Subject of Research: Trustworthy generative artificial intelligence integration in Big Data system architectures

Article Title: Trustworthy generative artificial intelligence and large language models for Big Data: databases, mining, and ethics at scale

Article References: Dritsas, E., & Trigka, M. (2026). Trustworthy generative artificial intelligence and large language models for Big Data: databases, mining, and ethics at scale. Neural Computing and Applications, 38(21), Article 884. https://doi.org/10.1007/s00521-026-12561-7

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12561-7

Keywords: generative artificial intelligence, Big Data infrastructures, large language models, diffusion models, probabilistic generators, provenance-aware storage, governance-aware orchestration, federated and confidential execution, monitored inference, retrieval-augmented generation, query optimization, knowledge graph reasoning, explanation fidelity, privacy budgets, sustainability-aware scaling, data lineage, multi-objective evaluation