Google AI introduces DataGemma: a set of open models that use common data through Interleaved Generation Retrieval (IGR) and Augmented Generation Retrieval (AGR)

Google has introduced a revolutionary innovation called technology/ai/google-datagemma-ai-llm/”>gemmadatadesigned to address one of the most significant problems in modern artificial intelligence: hallucinations in large language models (LLMs). Hallucinations occur when ai confidently generates information that is incorrect or fabricated. These inaccuracies can undermine the usefulness of ai, especially for research, policymaking, or other important decision-making processes. In response, Google’s DataGemma aims to ground LLMs in real-world statistical data by leveraging the vast resources available through its Data Commons.

Two specific variants have been introduced designed to further improve the performance of LLMs: Gemma-RAG-27B-IT Data and Gemma-RIG-27B-IT DataThese models represent cutting-edge advancements in both Retrieval Augmented Generation (RAG) and Retrieval Interleaved Generation (RIG) methodologies. The RAG-27B-IT variant leverages Google’s extensive Data Commons to incorporate rich, context-based insights into its output, making it ideal for tasks that require deep understanding and detailed analysis of complex data. On the other hand, the RIG-27B-IT model focuses on integrating real-time retrieval from trusted sources to dynamically verify and validate statistical insights, ensuring accuracy in responses. These models are designed for tasks that demand high accuracy and reasoning, making them well-suited for research, policymaking, and business analysis domains.

The rise of large language models and the problems of hallucinations

LLMs, the engines behind generative ai, are becoming increasingly sophisticated. They can process huge amounts of text, create summaries, suggest creative outcomes, and even draft code. However, one of the critical shortcomings of these models is their occasional tendency to present incorrect information as fact. This phenomenon, known as hallucination, has raised concerns about the reliability and trustworthiness of ai-generated content. To address these challenges, Google has made significant research efforts to reduce hallucinations. These advances culminated in the launch of DataGemma, an open model specifically designed to anchor LLMs to the vast repository of real-world statistical data available in Google’s Data Commons.

Data Commons: the foundation of factual data

Common data is at the core of DataGemma’s mission: a comprehensive repository of trusted, publicly available data points. This knowledge graph contains over 240 billion data points across many statistical variables drawn from trusted sources including the United Nations, WHO, Centers for Disease Control and Prevention, and several national census offices. By consolidating data from these authoritative organizations into a single platform, Google is giving researchers, policymakers, and developers a powerful tool for accurate insights.

The scale and richness of the Data Commons make it an indispensable resource for any ai model looking to improve the accuracy and relevance of its output. The Data Commons covers a range of topics, from public health and economics to environmental data and demographic trends. Users can interact with this vast dataset through a natural language interface, asking questions such as how income levels correlate with health outcomes in specific regions or which countries have made the most significant progress in expanding access to renewable energy.

DataGemma's dual approach: RIG and RAG methodologies

Google’s innovative DataGemma model employs two distinct approaches to improve the accuracy and trueness of LLMs: Recall Interleaved Generation (RIG) and Recall Augmented Generation (RAG). Each method has unique advantages.

The RIG methodology builds on existing ai research by integrating proactive queries of trusted data sources into the model generation process. Specifically, when DataGemma is tasked with generating a response involving statistical or factual data, it cross-references the relevant data within the Data Commons repository. This methodology ensures that the model results are based on real-world data and verified against trusted sources.

For example, in response to a query about the global increase in renewable energy use, DataGemma’s RIG approach would pull statistical data directly from the Data Commons, ensuring the response is based on reliable, real-time information.

On the other hand, the RAG methodology expands the scope of what language models can do by incorporating relevant contextual information beyond their training data. DataGemma leverages the capabilities of the Gemini model, in particular its wide context window, to retrieve essential data before generating its output. This method ensures that the model’s responses are more complete, informative, and less prone to hallucinations.

When a query is posed, the RAG method first retrieves relevant statistical data from the Data Commons before generating a response, ensuring that the answer is accurate and enriched with detailed context. This is particularly useful for complex questions that require more than a straightforward factual answer, such as understanding trends in global environmental policies or analyzing the socioeconomic impacts of a particular event.

Initial results and promising future

Although RIG and RAG methodologies are still in their early stages, preliminary research suggests promising improvements in the accuracy of LLMs when handling numerical data. By reducing the risk of hallucinations, DataGemma has significant potential for diverse applications, from academic research to business decision-making. Google is optimistic that the increased factual accuracy achieved through DataGemma will make ai-powered tools more reliable, trustworthy, and indispensable for anyone seeking informed, data-driven decisions.

Google’s R&D team continues to refine RIG and RAG, and plans to expand these efforts and subject them to more rigorous testing. The ultimate goal is to integrate these enhanced capabilities into the Gemma and Gemini models in a phased approach. For now, Google has made DataGemma available to researchers and developers, providing access to the RIG and RAG quick-start models and notebooks.

Broader implications for the role of ai in society

The launch of DataGemma marks a significant step forward on the path to making LLMs more trustworthy and based on factual data. As generative ai is increasingly integrated into diverse sectors, ranging from education and healthcare to governance and environmental policy, addressing hallucinations is crucial to ensuring ai provides users with accurate information.

Google’s commitment to making DataGemma open source reflects its broader vision of fostering collaboration and innovation in the ai ​​community. By making this technology available to developers, researchers, and policymakers, Google aims to drive the adoption of data-driven techniques that improve the trustworthiness of ai. This initiative advances the field of ai and underscores the importance of fact-based decision making in today’s data-driven world.

In conclusion, DataGemma is a groundbreaking advancement in addressing ai hallucinations by basing LLMs on the vast and trusted datasets of Google’s Data Commons. By combining RIG and RAG methodologies, Google has created a robust tool that improves the accuracy and reliability of ai-generated content. This launch is an important step in ensuring that ai becomes a trusted partner in research, decision-making, and knowledge discovery, while empowering people and organizations to make more informed decisions based on real-world data.

Take a look at the technology/ai/google-datagemma-ai-llm/” target=”_blank” rel=”noreferrer noopener”>Details, Paper, RAG Gemmaand Gemma Platform

If you like our work, you'll love our Fact Sheet..

Don't forget to join our SubReddit of over 50,000 ml

FREE ai WEBINAR: 'SAM 2 for Video: How to Optimize Your Data' (Wednesday, September 25, 4:00 am – 4:45 am EST)