Open source and artificial intelligence

Open source software has had a significant impact on the world of artificial intelligence (AI) and has played a key role in its evolution. Accessibility to a wider audience, rapid iteration and increased collaboration among developers, data scientists, researchers and the entire AI community have transformed AI and accelerated its evolution and maturity.

Open source and enterprises

Open source has become mainstream and gained immense popularity in recent years. A 2020 O’Reilly survey on open source usage, sponsored by IBM, found the following:

• 94% of the respondents rated open source as better than or equal to proprietary software.
• 48.5% of the respondents use open source widely across their companies.
• 77% of the surveyed enterprises agreed that they meet their objectives while minimizing costs by using open source software.

We have learned from our customers that they require enterprise-grade support and the mitigation of security vulnerabilities in open source software.

Open source and IBM Z®

Open source software has been a focus area for IBM Z®. IBMers continue to contribute to popular and key open source projects. Open source has been critical to IBM’s strategy for building a popular and sustainable ecosystem around IBM Z hardware and software. The IBM Z team has optimized various popular open source software packages to run on IBM Z and IBM® LinuxONE, an enterprise-grade Linux® server.

To support popular and key open source AI frameworks to run on IBM Z and IBM LinuxONE and increase the adoption of AI technologies on these platforms, we are proud to announce the release of a new support offering called AI Toolkit for IBM Z and IBM LinuxONE. The AI toolkit consists of IBM Elite Support for the following open source and IBM non-warranted programs:

• IBM Z® Accelerated for TensorFlow: This is an end-to-end open source platform focused on neural networks for machine learning (ML).
• IBM Z® Accelerated for NVIDIA Triton Inference Server: This solution streamlines and standardizes AI inference, enabling teams to deploy, run and scale trained ML or deep learning models from any framework on GPU- and CPU-based infrastructures.
• IBM Z® Accelerated Serving for TensorFlow: This is a flexible, high-performance serving system tailored for ML models, designed specifically for production environments.
• IBM Z® Deep Learning Compiler: This tool generates programs from models, enabling execution on either z/OS or Linux on IBM LinuxONE Emperor 4.
• IBM Z® Accelerated for Snap ML: This is a library optimized for training and scoring popular ML models.

With this support offering, we provide customers with the following:

• IBM Elite Support for select open source AI and IBM non-warranted AI programs mentioned above to help our customers deploy these technologies in production.
• Leveraging IBM Z® Integrated Accelerator for AI on IBM z16™, designed for AI model inferencing.
• High-performance open source AI serving frameworks that run on IBM LinuxONE and IBM Z.
• Containers for the above-mentioned AI frameworks that have undergone rigorous IBM secure engineering processes and vulnerability scanning.

Use cases

The AI toolkit for IBM Z and IBM LinuxONE unlocks a wide variety of potential use cases on both IBM Z and IBM LinuxONE. Here’s a list of possible use cases:

1. Traditional classification and prediction through natural language processing
2. Analyzing large volumes of individual records, including training computer vision models while optimizing for energy efficiency
3. Real-time credit card fraud detection and prevention
4. Aiding insurance underwriters in setting insurance premiums

Resources

Find more AI on IBM Z resources, including blogs, demos and announcements. here.