By Todd Moore, Vice President, Open Technology, IBM
As open source artificial intelligence technologies grow, the need for AI systems to make decisions fairly, to be invulnerable to tampering, and to be explainable is more important than ever. At IBM, we believe that building trust in AI starts in the open, with code that is transparent and accessible to anyone. In light of our commitment to trusted AI, IBM has made it easy to securely access three of our trusted AI packages in IBM Cloud Pak for Data.
Specifically, we’re adding AI Fairness 360 Toolkit, the AI Explainability 360 Toolkit, and the Adversarial Robustness Toolbox to IBM Cloud Pak for Data so that our users can take advantage of key packages that help them make decisions about their data in a more informed way.
- AI Fairness 360 Toolkit (AIF360): Available in both Python and R, the AI Fairness 360 package includes a comprehensive set of metrics (over 70) for datasets and models to test for biases, explanations for these metrics, and algorithms (about 11) to mitigate bias in datasets and models. AIF360 translates algorithmic research from the lab into the actual practice of domains as wide-ranging as finance, human capital management, healthcare, and education. Find guidancefor using the tool.
- AI Explainability 360 Toolkit (AIX360): An open-source library that supports interpretability and explainability of datasets and machine learning models throughout the AI application lifecycle. The AI Explainability 360 Python package includes a comprehensive set of algorithms that cover different dimensions of explanations along with proxy explainability metrics. Find guidancefor using the tool.
- Adversarial Robustness Toolbox (ART): A Python library supporting developers and researchers in defending machine learning models against adversarial threats (including evasion, extraction and poisoning) and helps making AI systems more secure and trustworthy. ART includes attacks for testing defenses with state-of-the-art threat models. Find tutorials about notebooksand examplesfor using the tool.
Users of IBM Cloud Pak for Data can take advantage of these packages via the Open Source Management service within the product.
Using trusted AI packages in the Open Source Management service
In May, IBM announced the availability of the Open Source Management (OSM) service in Cloud Pak for Data 3.0, our fully integrated data and AI platform that modernizes how your business collects, organizes, and analyzes data to infuse AI throughout the organization.
The service provides access to a curated list of 800 open source packages and walks you through a governed workflow to request package offering insights to security and vulnerability risks. The result is that you’re better able to understand what open source packages are used the most, who in your organization is using the packages, and vulnerabilities related to each project.
You can use the trusted AI packages via the Open Source Management service in IBM Cloud Pak for Data. After you install the OSM service, you can choose two different ways to interact with the packages:
Option 1 – View insights
This option shows a holistic view of the number of projects that use open source packages, the number of packages that use each type of license, the number of packages affected by vulnerabilities, number of times users interact with open source packages, and the most active users across all packages.
Option 2 – View packages
This option enables you to view all the available packages. Each package includes a summary, vulnerability check report, supporting documents related to the package, list of projects using the package and a provision to link to new project. Once requested, approval process starts. After approval, you can use it securely in the linked projects and manage it using “View Insights”.