Article

High performance computing in AI

What GPUs mean for deep learning

The fundamentals of AI functionality can be probed by understanding three core elements—High Performance Computing (HPC), the graphics processing unit (GPU), and the field of machine learning. Looking through these lenses explains where AI is today and helps reveal how it can be developed going forward.

The value of HPC and GPU in AI

Organizations are striving to realize the strategic goals of improving customer engagement, enhancing operational effectiveness, and finding costs savings. Using AI to meet these targets presents computational demands that have traditionally required an HPC cluster. Today, cloud technologies have made HPC capabilities widely available, with the capacity to manage a “virtual HPC” elastically and as needed. This is a powerful AI accelerator in as much as there is a low barrier to entry.

Deloitte & NVIDIA: Accelerating AI to its business potential

Deep learning in simple terms

The AI models being deployed today are trained in the same way but at much greater complexity. There are AI tools that process plain language for customer engagement, recognize thousands of objects for self-driving cars, and find valuable patterns in tumor pathology images to support delivering the right medicines to patients. The increased complexity of the algorithms, combined with ever-increasing data volumes, requires more computational horsepower to train them. GPU-enabled computing makes these computational challenges tractable by computing in parallel, reducing time to compute and overall cost.

Looking over the horizon

To help push accelerated AI forward, we created the Deloitte Center for AI Computing, which is powered by NVIDIA’s DGX™ A100 systems (i.e., a GPU-enabled supercomputer). With it, we can innovate with clients to prove out new use cases, create new frontiers in the marketplace, and capture growth by selling new products and services. With an accelerated AI platform and the expertise and knowledge needed to use it, the potential in AI truly is limited only by the imagination.

${column-img-description}

${column-img-description}

Let's talk

 
 
 
 
 
 
  Yes         No

Did you find this useful?
Accelerated computing for AI in government

Learn our perspectives

GPU-accelerated Artificial Intelligence

The Secret to rapid and insightful AI