Last week, IBM made a significant announcement regarding its watsonx AI platform. The company introduced a new family of foundation models called “Granite,” which are designed to meet the specific needs of enterprise applications. Foundation models are the fundamental building blocks of AI, allowing platforms to understand, generate, and interact with language, images, and other content.
While models like ChatGPT have gained popularity in the generative AI space, they only provide a partial solution. To unlock the full potential of generative AI for enterprises, there is a need for domain-specific models that operate with an understanding of specific sectors such as finance, healthcare, and education. IBM’s Granite foundation models address this requirement.
The Granite models are tailored to support various business-domain tasks, including summarization, question-answering, and classification. IBM has carefully selected datasets from different domains, such as the internet, academia, code, legal, and finance, to ensure that the models are trained on industry-specific content. This familiarity empowers the models to make informed decisions grounded in sector-specific expertise.
One of the biggest concerns when adopting generative AI in the enterprise is the integrity of the underlying data. IBM addresses this concern by implementing a comprehensive end-to-end AI and governance process. This process manages and mitigates potential client risks within the watsonx AI and data platform. IBM’s data governance process evaluates datasets based on criteria like governance, risk, and compliance, ensuring responsible AI practices.
In terms of transparency, IBM is committed to publishing descriptions of its training dataset. It emphasizes the effort to address content quality and relevance issues related to open-source data compilations. The Granite models have undergone rigorous quality checks, including scrutiny by IBM’s “HAP detector” to filter out hateful and profane content.
This announcement is only the beginning for IBM’s Granite foundation models. The company plans to expand into other languages and develop more IBM-trained models. It will also include open-source models on its platform, such as Meta’s Llama 2-chat model. IBM’s ongoing commitment to innovation and expansion signals continuous growth in the potential of AI in the business world.
To complement the Granite models, IBM introduced new features to its watsonx.ai studio. The Tuning Studio allows business users to personalize foundation models for specific needs, while the Synthetic Data Generator facilitates AI model training and decision-making with lower risk.
IBM’s position and history in helping enterprises adopt new technologies, including AI, make it well-suited to promote responsible AI adoption. The introduction of Granite foundation models and additional tools presents exciting possibilities for AI applications in the business world. Organizations can now personalize AI to their values and regulatory requirements, accelerating the integration of AI into enterprise operations.
Overall, IBM’s Granite foundation models and associated tools are a significant development that highlights the company’s commitment to delivering innovative AI solutions for the enterprise sector.