Leaning into the Capabilities of AI on IBM Z: Transforming Enterprise AI with Telum Processor
It’s becoming an “AI-first” world. While organizations are eager to take delivery on the benefits of AI, it’s been a rough road for many. According to Gartner, Inc., and the RAND Corporation, AI initiatives are failing at twice the rate of other IT projects. A key factor in this low success rate is a lack of infrastructure to manage data and deploy models.
IBM has responded to the pressing need for adequate infrastructure to support AI workloads by developing hardware and software to deliver trustworthy, efficient, and outcome-driven AI. The result is a secure, reliable centerpiece for enterprise AI.
The IBM Telum Processor is a Breakthrough for Many Use Cases
It all starts with the IBM Telum processor — the heart of the IBM z16 mainframe — which enables organizations to implement energy-efficient AI at scale for many use cases. IBM Telum is an innovative solution to a key challenge for scaling enterprise AI: excessive latency, which has limited the success of many AI initiatives.
The financial services industry, for example, uses algorithmic AI to score transactions for potential fraud. Today, a system takes the scoring request, packages it up, and drives it down to a GPU farm to score it. The resulting latency means that credit card companies could only score random samples of transactions. With the Telum chip on IBM Z, financial institutions can now score 100% of transactions in real time.
By putting inference directly onto the Telum processor, organizations don’t need to go off-platform for inference. Coupled with embedded accelerators, Telum on-chip inference is enabling use cases that have heretofore been impractical or impossible for AI to handle.
To make IBM Z a premier platform for driving advances in traditional and generative AI, IBM invested in the hardware, machine instructions, frameworks, open source capabilities, and core libraries available under Linux for Z, LinuxOne, and zOS to enable customers to do on-platform inference.
The Telum processor includes instrumentation to support neural networks at the hardware level and deploy AI natively under zOS or Linux on Z. These are key differentiating factors for IBM, making IBM Z a potential game changer for any organization that needs to accelerate and scale AI for business value on their IBM Z mainframe.
How AI on IBM Z Fits into the AI Ecosystem
With AI on IBM Z, you can quickly integrate and deploy models for inference in business applications to simplify the workflows for developers and data scientists. You can continue to train models anywhere using a common framework such as TensorFlow or PyTorch to package your production model. The open-source inference engine, ONNX, is then used to create an intermediary representation of your model that lets you easily go from one environment to the next, for example, to deploy and run natively on your IBM Z mainframe. This enables organizations to continue to train models anywhere they want using their preferred framework.
How AI on IBM Z Supports IT Modernization
Fast inference is just one arrow in the IBM quiver. The IBM watsonx AI and data platform includes several pre-trained models to help IT departments work more efficiently and address the skills gap as critical knowledge ages out of the workplace. IT professionals can use IBM watsonx Assistant to help modernize their data center and make it easier to run. Developers can use the IBM watsonx Code Assistant to quickly modernize legacy code such as COBOL. If your organization maintains a knowledge base from retiring senior personnel, IT professionals can use a Z chat assistant to ask questions against that stored information.
AI on IBM Z in the Real World
On-chip inference is enabling many new AI use cases and accelerating many others. For example:
Fraud Detection — A large U.S. bank leverages AI on Z to detect fraud patterns and prevent fraudulent transactions. The bank was able to scale on Z to examine every transaction in real time. The result was reduced fraud, saved costs, and increased customer satisfaction.
Loan Approval — A large U.S. bank infused AI on Z for its loan approval process to minimize loan defaults through early fraud detection, accelerate the credit approval process, and improve customer service and profitability. The bank was also able to use AI on IBM Z to optimize IT and operational costs. The bank continues to train models off-platform with its preferred framework and ports it to its IBM Z mainframe.
Computer Vision for Medical Imaging — A large health care provider is using LinuxONE on IBM Z for computer vision training and inferencing for medical records. AI on IBM Z meets the organization’s need for security and energy efficiency to support genomics, biobank data analysis, medical image processing, and NLP for medical records.
Expected cost savings and efficiencies
Implementing AI on IBM Z can deliver significant cost savings and fill skills gaps, depending on your use case.
- The ability to score 100% of transactions leads to reduced risk and fewer fraudulent payouts without added cost.
- GenAI and other AI tools enable organizations to run their data centers and cloud workloads more efficiently, giving IT personnel the ability to focus on other matters.
- GenAI also creates efficiencies in extracting relevant knowledge from a knowledge base, giving organizations the ability to quickly tap into valuable information about how to run and maintain systems. GenAI answers questions in plain English and can even generate an Ansible script to perform an execution.
- The ability to use GenAI to fill the skills gap makes developers and other IT professionals more effective and more productive with less time spent researching and experimenting to solve problems.
- The ability to automate code modernization also helps address the skills shortage. Watsonx Code Assistant does much of the heavy lifting for developers by explaining the logic, flow, and structure of legacy code and refactor it into a modern language that leverages the full power and benefits of that language.
How Evolving Solutions Can Help
As infrastructure specialists, Evolving Solutions can provide valuable support in making AI a reality in your organization to improve IT operations or enable new use cases to support business growth and profitability.
We know how AI on IBM Z works because we invest time and resources into experimenting with it. We stay a step ahead by receiving early-ship solutions from IBM and other partners to learn how products and features can benefit our clients. Evolving Solutions also does shadow builds of client systems in our lab to experiment with implementing technology or upgrading operating systems, working out the bugs before deploying it with the client.
In these and many other ways, Evolving Solutions has the experience and capabilities to help make your AI program a success.
Contact Evolving Solutions today to get started.
