IBM is expanding its data and AI platform for business, IBM WatsonX, by adding a new family of foundation models called Granite. These models leverage generative AI to process both language and code, providing businesses with powerful capabilities to improve productivity and creativity. The Granite models are more efficient than larger models, making them suitable for a single GPU. They excel in specialized business-domain tasks such as summarization, question-answering, classification, content generation, insight extraction, retrieval-augmented generation, and named entity recognition.

Developed with Business in Mind

The Granite foundation models were developed by IBM Research and trained on a variety of specialized datasets to familiarize them with industry-specific language and jargon. These datasets include generic language data from the internet, technical language data focused on science and technology, unstructured code data covering various programming languages, enterprise-relevant language data from legal opinions and financial documents, and enterprise-relevant unstructured data from publicly posted financial documents and reports.

Building Trust Through Governance

IBM is committed to building AI models that businesses can trust. To ensure transparency and mitigate risks, every dataset used in training the Granite models undergoes a defined governance, risk, and compliance (GRC) review process. IBM has developed governance procedures consistent with AI ethics principles to incorporate data into the IBM Data Pile. Additionally, IBM employs its “HAP detector,” a language model that detects and eliminates hateful and profane content, to scrutinize the training data. Other quality measures, such as removing duplication and low-quality documents, are also applied to enhance the output quality.

Empowering Businesses with Customization

IBM’s vision of AI for business is centered around empowerment. The Granite foundation models can be personalized by enterprises according to their values and regulatory requirements. WatsonX users retain ownership of their data and have control over the models they build, allowing them to deploy the models wherever they need them.

Expanding Possibilities for AI in Business

The Granite foundation models are just the beginning, as IBM plans to introduce more models in other languages. Additionally, IBM is launching new complementary capabilities in the watsonx.ai studio, including the Tuning Studio for prompt tuning and a Synthetic Data Generator for creating artificial tabular datasets. These enhancements provide users with more opportunities to harness the power of AI for their specific business needs.

Summary

IBM is enhancing its AI platform for business, WatsonX, by introducing a new family of foundation models called Granite. These models incorporate generative AI to process language and code, delivering valuable capabilities for businesses to improve productivity and creativity. The Granite models undergo rigorous governance, risk, and compliance reviews to ensure trust and transparency. Enterprises can personalize the models to align with their values and regulatory requirements, retaining ownership and control over their data. IBM’s continuous development of AI models and tools opens up new possibilities for businesses to leverage AI in various domains.

FAQ

What is IBM WatsonX?

IBM WatsonX is a data and AI platform for business that offers a range of capabilities to help enterprises leverage AI for various applications.

What are foundation models?

Foundation models are AI models that serve as the basis for building more specialized AI solutions. They are trained on large datasets and can be adapted to perform specific tasks.

What are the Granite foundation models designed for?

The Granite foundation models are designed to support businesses in various domains, including language processing and code analysis, to enhance productivity and creativity.

How can businesses customize the Granite foundation models?

Enterprises using WatsonX can personalize the Granite foundation models according to their own values and regulatory requirements, allowing them to align the models with their specific needs.

What other capabilities are IBM introducing in the watsonx.ai studio?

In addition to the Granite foundation models, IBM is introducing the Tuning Studio for prompt tuning and a Synthetic Data Generator for creating artificial tabular datasets.

Sources:

IBM Research

IBM Research

IBM Research

IBM Research

IBM Research

IBM Newsroom