When the Algorithm Becomes the Brand Crisis: A Sociotechnical Theory of Distributed Responsibility and Accountable Transparency
论文探讨了AI系统故障时责任分配问题,提出了一种社会技术过程理论,并引入负责任透明度作为解决方案。
论文探讨了AI系统故障时责任分配问题,提出了一种社会技术过程理论,并引入负责任透明度作为解决方案。
研究通过分析982名参与者对17个AI和机器人挑战的评估,揭示了复杂性与准备度之间的关系,强调了在政策制定中应考虑具体挑战的差异。
This work proposes a 1:1 biomimetic dexterous hand grounded in “structural intelligence” to overcome the limitations of conventional dexterous hands, which rely on high-dimensional active control and fail to replicate the human hand’s low-dimensional, efficient manipulation enabled by its anatomical structure. The design uniquely integrates an eight-bone double-row wrist, crossed tendons, palmar plate constraints, and intrinsic muscle pathways, leveraging both structural priors and muscle-mediated modulation to achieve coordinated wrist–finger motion, extensor hood coupling, and intrinsic muscle regulation. This architecture maps low-dimensional inputs to default grasping postures while enabling fine, contact-driven adjustments. Experiments demonstrate successful execution of rich-contact tasks such as coin rotation and pen repositioning, confirming that structural priors effectively reduce control dimensionality and enhance dexterity.
This study critically examines the sociotechnical origins of the dominant trajectory in generative AI development, interrogating the conceptual validity of artificial general intelligence (AGI) and its entanglement with prevailing political-economic structures. Drawing on sociology of technology, path dependency analysis, and comparative case studies, it systematically traces the evolution of closed-source large models, open-weight models, and domain-specific architectures to identify pivotal decision points and marginalized alternative pathways. Moving beyond technological determinism, the work proposes a normative framework for developing “AGI-proximate capabilities” oriented toward transparency, human well-being, and sustainability. This approach seeks to balance ethical imperatives, governance requirements, and commercial viability, offering conditional pathways to guide the responsible evolution of artificial intelligence.
This work addresses the challenges of gradient computation and geometric parameter coupling in traditional multi-objective acoustic structural optimization. It presents the first integration of automatic differentiation with the boundary element method (BEM), leveraging JAX to construct an end-to-end differentiable acoustic simulation solver. This framework enables efficient gradient-based shape optimization and inverse problem solving while maintaining accuracy comparable to conventional BEM approaches. By significantly accelerating optimization over complex geometries, the proposed method establishes a new paradigm for acoustic design, offering both computational efficiency and high fidelity in solving intricate multi-objective problems.
论文探讨了AI系统故障时责任分配问题,提出了一种社会技术过程理论,并引入负责任透明度作为解决方案。
研究通过分析982名参与者对17个AI和机器人挑战的评估,揭示了复杂性与准备度之间的关系,强调了在政策制定中应考虑具体挑战的差异。
This work proposes a 1:1 biomimetic dexterous hand grounded in “structural intelligence” to overcome the limitations of conventional dexterous hands, which rely on high-dimensional active control and fail to replicate the human hand’s low-dimensional, efficient manipulation enabled by its anatomical structure. The design uniquely integrates an eight-bone double-row wrist, crossed tendons, palmar plate constraints, and intrinsic muscle pathways, leveraging both structural priors and muscle-mediated modulation to achieve coordinated wrist–finger motion, extensor hood coupling, and intrinsic muscle regulation. This architecture maps low-dimensional inputs to default grasping postures while enabling fine, contact-driven adjustments. Experiments demonstrate successful execution of rich-contact tasks such as coin rotation and pen repositioning, confirming that structural priors effectively reduce control dimensionality and enhance dexterity.
This study critically examines the sociotechnical origins of the dominant trajectory in generative AI development, interrogating the conceptual validity of artificial general intelligence (AGI) and its entanglement with prevailing political-economic structures. Drawing on sociology of technology, path dependency analysis, and comparative case studies, it systematically traces the evolution of closed-source large models, open-weight models, and domain-specific architectures to identify pivotal decision points and marginalized alternative pathways. Moving beyond technological determinism, the work proposes a normative framework for developing “AGI-proximate capabilities” oriented toward transparency, human well-being, and sustainability. This approach seeks to balance ethical imperatives, governance requirements, and commercial viability, offering conditional pathways to guide the responsible evolution of artificial intelligence.
This work addresses the challenges of gradient computation and geometric parameter coupling in traditional multi-objective acoustic structural optimization. It presents the first integration of automatic differentiation with the boundary element method (BEM), leveraging JAX to construct an end-to-end differentiable acoustic simulation solver. This framework enables efficient gradient-based shape optimization and inverse problem solving while maintaining accuracy comparable to conventional BEM approaches. By significantly accelerating optimization over complex geometries, the proposed method establishes a new paradigm for acoustic design, offering both computational efficiency and high fidelity in solving intricate multi-objective problems.