Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 53 for “"AI Safety"”.
-
Incentives and AI safety in causal models
One kind of concern about AI is that sufficiently advanced AI systems might outstrip the control of their programmers and users. At the core of this concern is that it is often hard to specify a notion of success for autonomous decision-making tasks that is perfectly accurate, and instead, an …
-
Visual and Algorithmic Explanations to Fortify AI Safety
As modern AI systems, such as diffusion-based generative models or large language models (LLMs), continue to grow in scale, complexity, and societal impact, understanding and mitigating their risks has become increasingly urgent yet challenging due to their black-box nature. My thesis addresses …
-
Towards AI Safety and Alignment in Interactive Environments
The thesis investigates the integration of AI alignment principles within human-computer interaction (HCI) frameworks to enhance the safety and steerability of reinforcement learning (RL) models in complex game environments. Derived from the HCI perspective, the research emphasises capturing …
-
Assuring AI safety for managing risk of myocardial infarction for patients with type 2 diabetes
… including the use of Artificial Intelligence (AI) to predict the risk of MI development. While AI-based methods have shown promising results, most of the existing research has focused primarily on improving the prediction performance of the models. However, in safety-critical domains like …
-
Advancing Information Extraction with Large Language Models: The Role of Structured Understanding in Knowledge Management and AI Safety
… an invaluable source of knowledge, yet it remains challenging for machines to interpret and transform into actionable insights. Information Extraction (IE) offers a promising solution by converting raw text into machine-readable representations. However, traditional IE systems are limited by …
-
From public health to AI safety: improving machine learning approaches by collecting, selecting, or reducing the need for high-quality data
… several application areas, from public health to AI safety. In the process, we develop novel machine learning (ML) methods to tackle various challenges. In each application area, we take a data-centric approach: we develop a method to automatically select high-quality training data, devise and …
-
Robustness, Detectability, and Data Privacy In AI
… rapid advancements in Artificial Intelligence (AI) have made it widely popular in various applications, ranging from autonomous systems to multimodal content generation. However, these models are vulnerable to several security and safety problems. These pitfalls in AI models can potentially …
-
Safe AI Compliance in CI/CD for Software-Defined Vehicles - Researching Conformity to ISO/PAS 8800:2024
… are incorporating artificial intelligence (AI), ensuring AI safety presents significant challenges for development teams. This thesis addresses four research questions spanning the current application of ISO/PAS 8800:2024 Road vehicles — Safety and artificial intelligence in SDV development, …
-
Understanding and Avoiding AI Failures: A Practical Guide.
As AI technologies increase in capability and ubiquity, AI accidents are becoming more common. Based on normal accident theory, high reliability theory, and open systems theory, we create a framework for understanding the risks associated with AI applications. In addition, we also use AI safety …
-
ARTIFICIAL SUPERINTELLIGENCE: ETHICS AND SAFETY
… act towards digital agents under moral uncertainty by considering the particular case of how we should act towards simulations run by an artificial superintelligence (ASI). By placing the problem of simulations within the larger problem of AI-safety (how to ensure a desirable post-ASI outcome) …
-
Use of minimal abstraction to address goal underspecification
… goals has been widely considered central to AI safety. While most work in this space focuses on querying users about potential negative side effects, almost none look at the question of addressing the cause of misspecification, namely, the user's misunderstanding regarding the consequences of …
-
TOWARDS SAFETY AND TRUSTWORTHY IN BIOMEDICAL AI
… have emerged as powerful tools in the medical domain, demonstrating human-level capabilities that enable applications ranging from clinical trial matching to risk prediction, biomedical knowledge retrieval and gene reasoning. These advances position LLMs as promising agents in healthcare, yet …
-
Advancing Cross-Domain Fake News Detection: Enhanced Models to Improve Generalization and Tackle the Class Imbalance Problem
The rapid proliferation of fake news across domains, such as politics, health, and social media, poses a significant threat to the integrity of information dissemination, leading to misinformation that can affect public perception and decision-making. Detecting fake news is critical to preserving …
-
Cross-layer SAE feature drift for early detection and mitigation of multi-turn jailbreaks
… development of Large-Language Models (LLMs) has raised AI Safety to a critical concern. Despite recent safety alignment, LLMs remain vulnerable to multi-turn jailbreak attacks that bypass safeguards through gradual escalation. This thesis investigates whether Sparse Autoencoder (SAE) feature drift …
-
MODELING ENDOGENOUS CATASTROPHES IN HIERARCHICAL NETWORKS VIA OMEGA-ROBUSTNESS: APPLICATIONS TO ADVANCED AI SYSTEMS
… endogenous catastrophic behaviors in advanced AI systems. Central to this framework is a novel tool we develop, Ω-Robustness—a set-theoretic optimization methodology that integrates Pareto and Distributionally Robust Optimization to support decision-making under deep uncertainty and in the …
-
Advancing semantic communication systems through knowledge graphs, generative AI, and safeguarded AI
… for coherent semantic representations remains complex. Second, reconstructing original data from KGs proves particularly difficult under adverse communication conditions. Third, the KG integration inevitably introduces additional transmission overhead that must be carefully managed. This …
-
Evolving Threats and Defenses in Machine Learning: Focus on Model Inversion and Beyond
… into critical real-world applications, raising concerns about security, privacy, and trustworthiness. Among various emerging threats, model inversion (MI) attacks stand out due to their potential to compromise the confidentiality of training data. This dissertation investigates evolving …
-
ARTIFICIAL INTELLIGENCE GOVERNANCE: EVALUATING THE ROLE OF GOVERNMENT FOR FACIAL RECOGNITION, DATA PRIVACY AND PUBLIC TRUST
Artificial intelligence (AI) is one of the fastest developing technologies that has been widely deployed throughout society, including by law enforcement and surveillance agencies. Despite the influx of investment and interest, there remains a large knowledge gap between developers and deployers …
-
Nothing to See Here: Generative AI, Neoliberal Crisis, and Intensified Counterinsurgency
Examining generative artificial intelligence (genAI) as a discursive and technical agent deployed in response to the ongoing crisis of neoliberal legitimacy, this dissertation advances the claim that so-called “safe” models extend the counterinsurgent (COIN) mode of governance, both as an effect of …
-
Building Reliable AI under Distribution Shifts
… where distribution shifts—differences between training and deployment data—can significantly impact their reliability. These shifts affect models in multiple ways, leading to degraded generalization, fairness collapse, loss of robustness, and new safety vulnerabilities. This dissertation …
Page 1 of 3