Generative artificial intelligence (AI) has rapidly emerged as a transformative technology, demonstrating immense potential to revolutionize a diverse range of industries through the creation of original content and automation of intricate tasks. Recent advancements in this domain have given rise to large multimodal models, such as GPT-4, which are designed to process and analyse heterogeneous data inputs, including text and images, to generate human-like, meaningful, and coherent outputs. Despite the remarkable capabilities of generative AI, it is not without challenges; the technology is characterized by high operating costs, as the expenses associated with training these generative models are substantial. Moreover, the costs of inference significantly surpass those of training when deploying such models at a reasonably large scale, presenting a critical concern in the practical implementation of generative AI. As such, application-guided design optimizations for high-performance computing (HPC) systems can play a crucial role in mitigating the challenges associated with the training and deployment of generative AI models, by offering tailored solutions that improve performance, energy efficiency, and cost-effectiveness.
The primary objective of this research topic is to foster collaboration between industry and academia, driving the exploration of diverse aspects of generative AI and its implications on the design and optimization of high-performance computing (HPC) systems. Firstly, this research topic endeavours to investigate emerging applications of generative AI, which can inform domain-specific HPC system design, consequently reducing costs and enhancing efficiency across various industries. In addition, the examination of these applications will empower generative model researchers to better understand and position their work for practical adoption. Secondly, the research topic seeks to promote cross-stack design space exploration for HPC systems tailored to execute generative AI applications. These cross-stack design spaces may encompass application optimization, compiler design, operating systems, system architecture, memory technologies, and accelerators exclusively developed for generative AI applications. Thirdly, the research topic aims to address the challenges related to data quality and quantity, which often act as bottlenecks for generative models, by exploring application-specific and general data collection and preprocessing strategies within the context of generative AI. Finally, the research topic will investigate efficient mechanisms for data storage and transmission in HPC systems explicitly designed to support generative AI applications.
Topics of interest include, but are not limited to:
1. Cutting-edge applications of generative AI across diverse domains and industries.
2. Cross-stack design exploration in HPC systems for generative AI applications.
3. Application-driven generative model design for practical adoption.
4. Strategies for data collection and preprocessing in generative AI.
5. Efficient data storage and transmission in generative AI-focused HPC systems.
6. Novel architectures and hardware for generative AI applications.
7. Benchmarking and evaluation methodologies for HPC systems in generative AI.
Keywords:
artificial intelligence, GPT, HPC systems, generative AI, AI models
Important Note:
All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
Generative artificial intelligence (AI) has rapidly emerged as a transformative technology, demonstrating immense potential to revolutionize a diverse range of industries through the creation of original content and automation of intricate tasks. Recent advancements in this domain have given rise to large multimodal models, such as GPT-4, which are designed to process and analyse heterogeneous data inputs, including text and images, to generate human-like, meaningful, and coherent outputs. Despite the remarkable capabilities of generative AI, it is not without challenges; the technology is characterized by high operating costs, as the expenses associated with training these generative models are substantial. Moreover, the costs of inference significantly surpass those of training when deploying such models at a reasonably large scale, presenting a critical concern in the practical implementation of generative AI. As such, application-guided design optimizations for high-performance computing (HPC) systems can play a crucial role in mitigating the challenges associated with the training and deployment of generative AI models, by offering tailored solutions that improve performance, energy efficiency, and cost-effectiveness.
The primary objective of this research topic is to foster collaboration between industry and academia, driving the exploration of diverse aspects of generative AI and its implications on the design and optimization of high-performance computing (HPC) systems. Firstly, this research topic endeavours to investigate emerging applications of generative AI, which can inform domain-specific HPC system design, consequently reducing costs and enhancing efficiency across various industries. In addition, the examination of these applications will empower generative model researchers to better understand and position their work for practical adoption. Secondly, the research topic seeks to promote cross-stack design space exploration for HPC systems tailored to execute generative AI applications. These cross-stack design spaces may encompass application optimization, compiler design, operating systems, system architecture, memory technologies, and accelerators exclusively developed for generative AI applications. Thirdly, the research topic aims to address the challenges related to data quality and quantity, which often act as bottlenecks for generative models, by exploring application-specific and general data collection and preprocessing strategies within the context of generative AI. Finally, the research topic will investigate efficient mechanisms for data storage and transmission in HPC systems explicitly designed to support generative AI applications.
Topics of interest include, but are not limited to:
1. Cutting-edge applications of generative AI across diverse domains and industries.
2. Cross-stack design exploration in HPC systems for generative AI applications.
3. Application-driven generative model design for practical adoption.
4. Strategies for data collection and preprocessing in generative AI.
5. Efficient data storage and transmission in generative AI-focused HPC systems.
6. Novel architectures and hardware for generative AI applications.
7. Benchmarking and evaluation methodologies for HPC systems in generative AI.
Keywords:
artificial intelligence, GPT, HPC systems, generative AI, AI models
Important Note:
All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.