Performance-Intensive Computing Infrastructure as the Framework for Generative Artificial Intelligence

Performance-Intensive Computing Infrastructure as the Framework for Generative Artificial Intelligence

Performance-intensive computing (PIC) covers the umbrella of the entire practice to extract the optimal and highest performance from a consolidated infrastructure that includes high-performance compute (HPC), generative artificial intelligence (AI), big data and quantum computing. IDC continues to view PIC as a constant evolution that will revolutionize many technologies, particularly generative AI and quantum computing, solving many of humanity's greatest problems and challenges in an accelerated manner. HPC and generative AI will continue to unfold and mature, relying on a combination of multiple technologies in an experimental and exploratory fashion."There are several technical and methodological approaches to generative AI, including generative adversarial networks (GANs), variational autoencoders (VAEs), autoregressive models (AM), and diffusion models. These learn from increasingly larger amounts of data that then generate newer and more novel data. The market and technology are also replete with prominent generative AI systems and tools, such as OpenAI, Anthropic, Mindverse, and AI21 Labs, to name a few. Crucial to this is the already massive exponential data mixed with generative AI novel content, how it is produced and stored, and used by every organization. Sensitive data is vulnerable to data breaches, corruption, and cyberattacks. On a positive note, there is great impetus and an opportunity for collaboration from all sectors of the community to develop purposeful policies. This necessitates participation from all organizations (e.g., government, corporate businesses, industries, and academia) to work together to establish rules and standards for AI to protect the individual and humanity's existence," says Josephine Palencia, research director of HPC, IDC.

Please Note: Extended description available upon request.