Skip to content Skip to sidebar Skip to footer

“IBM Unveils “Granite”: AI Framework Models for Enterprise”

IBM Research introduces efficient and versatile Granite foundation AI models, emphasizing trust and empowerment for businesses in AI development and application.

Key Points

  • IBM Research unveils Granite foundation models, Granite.13b.instruct and, with 13 billion parameters optimized for linguistic and coding tasks.
  • IBM used vast datasets, 7 TB for pre-processing, 2.4 TB for post-processing, resulting in 1 trillion tokens, spanning multiple domains.
  • IBM prioritizes trust and transparency in AI, with robust governance, risk assessment, bias mitigation, and compliance measures.
  • IBM’s AI vision centers on empowering businesses, offering customization, data ownership, and ongoing innovation, including partnerships and new features in studio.

IBM Research has unveiled its latest breakthrough in artificial intelligence (AI) for business with the introduction of the Granite foundation models.

These models, called Granite.13b.instruct and, are built on a “Decoder” architecture and contain 13 billion parameters.

The models are optimized for efficiency and can fit into a single V100-32GB GPU. They are specifically designed for linguistic and coding applications, excelling in tasks such as summarizatcion, question-answering, and classification.

To train the models, IBM used a vast amount of data with a total of 7 TB for pre-processing and 2.4 TB for post-processing.

This resulted in a staggering 1 trillion tokens. The datasets used cover a wide range of domains, including the internet, academia, coding, legal, and finance.

This diversity ensures that the models are well-versed in industry-specific language and terminology.

Trust and Transparency: IBM’s Commitment

IBM places a strong emphasis on trust and transparency in AI development. The company’s watsonx AI and data platform have a comprehensive process for creating and testing foundation models and generative AI.

This process includes data collection, model deployment, and focuses on governance, risk assessment, bias mitigation, and compliance.

To ensure data integrity during training, IBM has implemented a rigorous governance, risk, and compliance (GRC) review process.

Additionally, the company has developed the “HAP detector,” a language model that identifies and eliminates hateful and profane content. These measures are part of IBM’s broader strategy to mitigate the risks associated with generative AI.

Empowerment Through AI

IBM’s vision for AI in business centers around empowerment. The company believes that organizations should have the autonomy to customize their models according to their values using the tools provided by the watsonx platform.

IBM also ensures that businesses retain ownership and control over their data and models, giving them the power to harness AI technology to its full potential.

The Road Ahead

While the Granite models represent a significant milestone, IBM’s journey in the AI space is far from over. The company plans to introduce more models in different languages and is actively developing other IBM-trained models.

In a recent announcement, IBM revealed a partnership with Meta, which offers early access to Meta’s Llama 2-chat 70 billion parameter model. Additionally, IBM will soon launch StarCoder, a comprehensive language model specifically designed for coding.

Concluding Thoughts

In the coming weeks, IBM will also roll out new features in the studio. These features, including the Tuning Studio and the Synthetic Data Generator, will further enhance the platform’s capabilities and help businesses unlock the full potential of AI.

In conclusion, IBM’s launch of the Granite foundation models and its ongoing developments in AI signify a new era in AI for business. With endless possibilities on the horizon, IBM remains at the forefront of innovation in this rapidly evolving field.

– This article is for informational purposes only and is the exclusive property of Blockchain.News. Accuracy and completeness cannot be guaranteed. Image source: Shutterstock.