Artificial intelligence for risk analysis and the risks of artificial intelligence: Part 1

Risk Analysis, Год журнала: 2025, Номер unknown

Опубликована: Март 13, 2025

The revolution created by artificial intelligence and machine learning (AI-ML) is affecting our society in a profound way. use of entering an increasing number human activities, with AI helping to automate tasks augmenting abilities. Although AI-ML has the potential benefit society, us recognize risks more promptly manage them better, disruption caused these new technologies also presents challenges society. In particular, unthinking or uninformed application decision-making can result biased unreliable decisions. addition, concerns about data privacy security pervade tools. associated consequences are difficult assess predict. This special issue collects series papers that showcase recent advances research on their relationship practice risk analysis risk-informed decision-making. clearly show transfer knowledge methods between occurs both directions. On one hand, risk-analysis theories contribute framing controlling threats activities posed AI-ML. other tools starting be used analysis. subject thorough investigation Stødle et al. (2025). authors frame this integration as input−output process helps accomplish three main consequence identification, uncertainty characterization, management. Through framework, they analyze current applications discuss future uses assessment. They conclude several recommendations for researchers practitioners, highlighting opportunities limitations Paté-Cornell (2025) identifies investigates concern may not consistent preferences attitudes individuals involved given decision. Some questions emerge how identify attitude implied tool, results should communicated possibly modified before being integrated into process. Baum proposes assessment takeover (i.e., key decisions generative AI), potentially catastrophic humanity. author develops stylized model compares type capabilities would need have such happen against large language models (LLMs). Based assessment, provides whether aggressive governance LLMs needed. We leave answer paper. Collier investigate play role product began using popular ChatGPT provide suggestions performing failure modes effects recommending mitigations. then presented safety experts evaluate ChatGPT's output. same was performed additional LLMs. expert examination identified significant producing inconsistent, generic, sometimes unrealistic guidance. However, suggest still useful initial ideation, focusing critically reviewing refining AI-generated content improve overall Faddi fill gap literature developing quantitative reliability resilience applied safety-critical domains. work highlights dynamic nature problem, since performance change over time. demonstrate approach applying it image-recognition system adversarial attacks. Issa study affects experiences finance professionals. transactional-stress framework examine six technology-related stressors influence positive negative stresses experienced professionals when AI. element, techno-accountability, driver. Techno-accountability technological stressor addresses unique responsibility liability arising from AI-driven across various stakeholders. It focuses ethical legal implications unexpected outcomes, recognizing users, developers, organizations held accountable made autonomous systems. Madsen information coming near misses prediction maritime accidents. U.S. Coast Guard's Marine Information Safety Law Enforcement database, analyzing incidents 2007 2022. After systematic analysis, test alternative machine-learning select best-performing ones, whose high level accuracy shows effectiveness predict incidents. These implementation tool Guard, well industries practices incident reporting. Osman El-Gendy economic impacts cyberattacks global trade supply chains. apply computable general equilibrium modeling carefully selected simulation scenarios cascading highly interconnected regions industrial sectors. Thekdi address problem integrating advanced analytics disaster focus compatibility traditional established develop methods' survey community practitioners. Winter nontechnical dimensions system. swarm 10 robots developed engineers at University Bristol public cloakroom designed receive, store, return belongings. method nests sources encompassing contextual factors. range physical/safety hazards (collision humans), operational (how deal belongings) technical organizational hazards. AI, causal plays central different domains, ranging healthcare financial (the so-called AI). Yu Smith propose graphical representation elements probabilistic graphs. chain event graphs understand failures repair processes. methodology based trees Bayesian inference techniques accommodate complex asymmetric introduce class remedial interventions maintenance. area healthcare, Macrae performs in-depth sociotechnical factors related 40 interviews development, management, regulation Finally, Li Wang (2025), Guo Zhang successfully enhance rescue operations tunnel projects, respectively. works will illustrate rich conceptual overlap flow ideas link represent phases intersections exploration exploding field, many up rapidly evolve. anticipate having second volume soon continuing important exploration. support findings available request corresponding author. publicly due restrictions.

Язык: Английский

Integrated Rover Path Planning and Validation on Real Outdoor Terrain Scenarios Using Satellite Information to Conduct a Real Achievable Trajectory

Electronics, Год журнала: 2025, Номер 14(5), С. 921 - 921

Опубликована: Фев. 26, 2025

The reliable and efficient navigation for mobile robots across challenging outdoor terrains is critical autonomous robotics. Traditional methods planning the path of such often emphasize minimizing travel distance but do not accommodate terrain stability, variability, or energy efficiency. study proposes an integrated approach between satellite-driven geolocation data terrain-specific features that enhance strategies in complex environments. Our method a controller uses search-based algorithms to generate energy-efficient dynamically stable trajectories incorporating surface characteristics environmental from satellite imagery. By integrating our method, proposed framework identifies safer more routes, achieving significant 32% improvement traction compared conventional models path-finding approaches. method’s benefits over traditional approaches include improved safety, extended operational efficiency, ability navigate unpredictable dynamic This makes it ideal planetary exploration, disaster response landslide-prone areas, agricultural automation precision farming rough terrains, search rescue operations earthquake-affected delivery systems into rural unstructured landscapes. It redefines through terrain-aware delivers robust performance

Язык: Английский