IBM's Granite Series Models: A New Step in Generative AI

• Generative AI

To gain relevance in the highly competitive AI sphere, IBM has recently unveiled its new generative AI models and capabilities on its Watsonx data science platform. The Granite series models are large language models (LLMs) akin to OpenAI's GPT-4 and ChatGPT. These models are capable of summarizing, analyzing, and generating text. However, IBM remains quite secretive about the specifics of the Granite series, making it challenging to compare them to other LLMs, including IBM's own. The company has promised to reveal the data used to train the Granite series models and the steps to filter and process that data ahead of the models' availability in Q3 2023.

"The future of AI is here, and IBM is boldly stepping into it with their new Granite series models."

Demystifying the Granite Series Models

Through an email interview, Tarun Chopra, IBM's VP of product management for data and AI, provided some insights into the Granite Series. According to Chopra, these new IBM models have been developed using curated, enterprise-quality data rather than publicly scraped data. The series has subsets specialized within different domains. For example, a model trained on finance data enables AI builders to use a much smaller model that can perform as well as a larger general model. These models can support most enterprise NLP tasks, such as summarization, content generation, and insight extraction.

Tuning Studio and Synthetic Data Generator

IBM is also introducing new tools and features on its platform. The Tuning Studio tools on the Watsonx.ai component allow users to customize generative AI models to their data. IBM Watson customers can fine-tune models to new tasks with as few as 100 to 1,000 examples. Once users specify a job and provide labeled samples in the required data format, they can deploy the model via an API from the IBM Cloud.

Another feature set to debut soon in Watsonx.ai is a synthetic data generator for tabular data – the collections of rows and columns found in relational databases. By generating synthetic data from custom data schemas and internal datasets, companies can extract insights for AI model training and fine-tuning with "reduced risk."

New Generative AI Capabilities in Watsonx.data

IBM is also launching new generative AI functionalities in Watsonx. Data is the company's data store, allowing users to access data while applying query engines, governance, automation, and integrations with existing databases and tools. Starting in Q4 2023, as part of a tech preview, customers will be able to "discover, augment, visualize, and refine" data for AI through a self-service, chatbot-like tool.

IBM's Future in the AI Space

Despite the pressures, IBM is determined to impact the competitive AI market significantly. However, in its second fiscal quarter, IBM reported revenue that was below what analysts had expected, as there was a slowdown in its infrastructure business segment. Nevertheless, IBM's CEO, Arvind Krishna, remains optimistic about the role of AI in IBM's future growth. He has reported that more than 150 corporate customers, including Samsung and Citi, are using Watsonx as of July.

IBM's recent developments in AI have taken a significant stride toward meeting clients' needs, seeking trusted enterprise AI solutions. The introduction of the Granite series models and other enhancements on the Watsonx platform underscores IBM's commitment to supporting clients throughout the entire AI lifecycle, from setting up foundational elements of their data strategies to tuning models for their specific business use cases and helping them govern models beyond that.