IBM has introduced WatsonX.ai, an AI and data platform designed to help businesses leverage generative AI models and accelerate adoption across their organizations. The platform offers a range of features, including new general-purpose and code-generation models, expanded open-source model options, and enhanced data options and tuning capabilities. These enhancements are guided by IBM’s principles of openness, trust, targeting, and empowerment. WatsonX.ai also provides business-focused foundation models built from reliable data and incorporates transparency in AI models. Additionally, the platform supports the responsible use of third-party models and reduces model-training risk through the use of synthetic data. IBM’s WatsonX.ai is part of the overall generative AI solutions offered by the company.

Business-targeted IBM-developed foundation models

WatsonX.ai offers a family of language and code foundation models of different sizes and architectures to give businesses flexibility and choice. These models, which include the newly introduced Granite series, are suited to generative tasks such as summarization, content generation, classification, and extracting insights. The Granite models have been trained on enterprise-focused datasets curated by IBM, providing domain expertise in areas like internet, academic, code, legal, and finance. By training with high-quality finance data, the Granite-13B models have shown similar or better performance than larger models in various financial tasks.

Building transparency and addressing AI model gaps

IBM aims to provide transparency in AI development by sharing the data sources used in training the Granite models. This helps alleviate uncertainties and aids adoption, particularly in highly regulated industries. IBM’s approach to AI model indemnification also ensures that clients are protected against third-party IP claims. The WatsonX.ai platform also enables organizations to responsibly use third-party models and includes AI guardrails to remove offensive language. To address data gaps, WatsonX.ai offers synthetic data generation, allowing organizations to create synthetic tabular data that preserves the statistical properties of their original enterprise data.

Enabling business-focused use cases and prompt tuning

WatsonX.ai includes the Tuning Studio feature, which allows business users to customize foundation models to their specific downstream needs. The initial release supports prompt tuning, where organizations can use their own data to customize existing models without the need for extensive retraining. This reduces computing and energy use and improves speed to market for AI solutions.

Advancing AI for business with WatsonX

IBM’s WatsonX.ai platform is built to help businesses scale and accelerate the impact of AI. It seamlessly integrates IBM-developed foundation models and third-party models from platforms like Hugging Face. The platform is complemented by IBM Consulting, which provides expert assistance in tuning and operationalizing models for targeted business use cases. IBM’s commitment to openness, trust, and governance is further supported by the upcoming release of WatsonX.governance.

Summary

IBM has unveiled WatsonX.ai, an AI and data platform designed to accelerate the adoption of generative AI. The platform offers a range of features including business-targeted foundation models, transparency in AI models, support for third-party models, synthetic data generation, and prompt tuning. WatsonX.ai is part of IBM’s generative AI solutions, aimed at helping businesses leverage AI technologies and achieve impactful results.

Frequently Asked Questions (FAQ)

What is WatsonX.ai?

WatsonX.ai is an AI and data platform developed by IBM to accelerate the adoption of generative AI. It offers various features and capabilities, including business-targeted foundation models, transparency in AI models, support for third-party models, synthetic data generation, and prompt tuning.

What are foundation models?

Foundation models are pre-trained AI models that serve as the building blocks for specific AI tasks. They provide a starting point for customization and can be fine-tuned to meet the specific needs of businesses.

How does WatsonX.ai address transparency in AI models?

WatsonX.ai promotes transparency in AI models by sharing the data sources used in training the models. This helps businesses understand the provenance and performance parameters of the models and encourages responsible use and adoption in highly regulated industries.

What is prompt tuning?

Prompt tuning is a feature in WatsonX.ai that allows organizations to customize AI models to their specific downstream needs using their own data. This customization can be done without extensive retraining, reducing computing and energy use and improving the speed to market for AI solutions.