IBM launches basic model for generative artificial intelligence

The IBM Granite family of foundational models applies generative artificial intelligence to natural language and coding tasks. Additionally, Watsonx generative AI capabilities are coming to Watsonx Data Lake.

IBM released the generative artificial intelligence basic model and the generative artificial intelligence enhanced version of Watsonx.ai, improving Watsonx artificial intelligence And the capabilities of the data platform

The IBM Granite series of multi-dimensional base models released on September 7 use the "decoder" architecture to apply generative artificial intelligence to language and coding tasks. Supports enterprise NLP (natural language processing) tasks such as summarization, content generation and insight extraction.

IBM plans to provide a comprehensive list of data sources and description of the data processing and filtering steps performed to generate the training data for the Granite series, which will be available this month. IBM will also provide third-party models, including Meta's Llama 2-chat 7 billion parameter model and StarCoder LLM (large language model), for code generation on Watsonx.ai on the IBM Cloud.

Watsonx.ai models are trained on IBM’s enterprise-grade data lake. The company said it has established a training process characterized by rigorous data collection and deployment of models and applications using control points to enable governance, risk assessment, compliance and bias reduction.

Other features planned for the Watsonx platform include:

• Tuning Studio for Watsonx.ai provides a mechanism to Adapt the underlying model to unique downstream tasks with their own enterprise data. Tuning Studio will be released this month.
• Watsonx.ai’s synthetic data generator helps users build artificial tabular datasets based on custom data schemes or in-house datasets. Users will be able to extract insights for training and fine-tuning AI models with reduced risk. This feature will also be rolled out this month.
• Watsonx.data data lake storage will be combined with Watsonx.ai’s generative artificial intelligence capabilities to help users discover, visualize and improve data through a natural language interface. This will be previewed in the fourth quarter of this year.
• Watsonx.data will integrate vector database capabilities to support Watsonx.ai’s search-enhanced generation use cases, which are also planned for preview in the fourth quarter.
• Model Risk Governance for Generative AI, launched as a technology preview of Watsonx.governance, will enable customers to explore automated collection of underlying model details and model risk governance capabilities for enterprise-level View relevant metrics in the AI ​​workflow dashboard.

IBM is also embedding Watsonx.ai innovations in its hybrid cloud software and infrastructure, including

• Intelligent IT automation, which entered the technology preview stage this week, features automation products Instana and AIOps, including intelligent repair embedded in the Watsonx.ai generative artificial intelligence basic model to help IT operations practitioners summarize incident details. Prescriptive workflow recommendations will be provided to help engineers implement solutions.
• Watsonx developer services help developers more closely integrate Watsonx capabilities with data on IBM Power for SAP workloads. Watsonx's SAP ABAP SDK will increase the way customers use artificial intelligence to infer data on IBM Power systems and deploy artificial intelligence algorithms across sensitive data and transactions. These services will be launched in the first quarter of 2024.

##Original title: IBM unveils generative AI foundation models

Original author: Paul Krill

The above is the detailed content of IBM launches basic model for generative artificial intelligence. For more information, please follow other related articles on the PHP Chinese website!