On September 7th, IBM unveiled the Granite series of foundation models, which utilize the powerful “Decoder” architecture to apply generative AI capabilities to both language and code-related tasks. These models are versatile and can support a wide range of enterprise-level natural language processing (NLP) tasks, including summarization, content generation, and insight extraction.
What sets IBM’s approach apart is its commitment to transparency. The company plans to provide a comprehensive list of data sources, along with detailed descriptions of the data processing and filtering steps used to create the training data for the Granite series. This transparency is a nod to IBM’s dedication to ensuring the integrity and quality of its AI models. The Granite series is set to become available later this month.
Furthermore, IBM is expanding its AI offerings by including third-party models on its Watsonx.ai platform. This move includes Meta’s Llama 2-chat 70 billion parameter model and the StarCoder LLM, designed for code generation within the IBM Cloud environment.
These Watsonx.ai models are trained on IBM’s enterprise-focused data lake, a testament to the company’s commitment to data quality and governance. IBM has implemented rigorous data collection processes and control points throughout the training process, which is crucial for deploying models and applications in areas such as governance, risk assessment, compliance, and bias mitigation.
IBM’s vision for the Watsonx platform doesn’t stop at foundation models; it includes several exciting capabilities:
- Tuning Studio for Watsonx.ai: This tool offers a mechanism to fine-tune foundation models to cater to unique downstream tasks using enterprise-specific data. Tuning Studio is expected to launch this month.
- Synthetic Data Generator for Watsonx.ai: This feature empowers users to create artificial tabular datasets from custom data schemes or internal datasets. It provides a safer way to extract insights for AI model training and fine-tuning, all while reducing data-related risks. Like Tuning Studio, this capability is also set to debut this month.
- Watsonx.data Lakehouse Data Store: This data store will incorporate Watsonx.ai’s generative AI capabilities, making it easier for users to discover, visualize, and refine data through a natural language interface. It is scheduled to be available in preview in the fourth quarter of this year.
- Watsonx.data Vector Database Integration: IBM plans to integrate vector database capabilities into Watsonx.data to support retrieval-augmented generation use cases. This feature is also expected to be available in preview in the fourth quarter.
- Model Risk Governance for Generative AI: IBM is launching this as a tech preview for Watsonx.governance. It will enable clients to automate the collection of foundation model details and gain insights into model risk governance through informative dashboards integrated into their enterprise-wide AI workflows.
Beyond these innovations, IBM is seamlessly integrating Watsonx.ai enhancements into its hybrid cloud software and infrastructure. This includes:
- Intelligent IT Automation: This feature, entering tech preview this week, leverages automation products like Instana and AIOps. It includes “Intelligent Remediation,” which employs Watsonx.ai generative AI foundation models to help IT ops practitioners summarize incident details and provides prescriptive workflow suggestions to address issues efficiently.
- Developer Services for Watsonx: These services aim to bring Watsonx capabilities closer to data on IBM Power for SAP workloads. The SAP ABAP SDK for Watsonx will offer clients new ways to utilize AI for data inference and transaction processing on sensitive data. Expect these services to launch in the first quarter of 2024.
In conclusion, IBM’s latest advancements in generative AI foundation models and enhancements to the Watsonx.ai platform showcase the company’s commitment to transparency, data quality, and expanding the horizons of AI across a wide range of industries and applications. These developments are poised to empower enterprises with advanced AI capabilities and data-driven insights.