In the rapidly evolving landscape of artificial intelligence, a powerful synergy is emerging between knowledge graphs and generative AI (GenAI). This partnership is revolutionising how data is processed, interpreted, and utilised, enabling more accurate, context-aware, and personalised AI applications.
Jim Webber, chief scientist at Neo4j, explains, “By training an LLM on a knowledge graph’s curated, high-quality, structured data, we can address the gamut of challenges associated with generative AI. This combination offers high levels of accuracy and correctness, making it an ideal partner for LLMs.”
Why Knowledge Graphs are Important
Tony Seale, knowledge graph engineer at UBS, emphasised, “Knowledge graphs are key to unlocking the power of AI. They enable effective AI deployments by providing a structured framework that mirrors human understanding and reasoning.”
By integrating structured data, knowledge graphs enhance the accuracy and relevance of AI outputs. They provide context and connectivity, allowing AI systems to generate more precise and context-aware responses.
A study published in Nature highlights, “The union of causal graphs’ systematic approach with AI-driven creativity paves the way for the future of psychological inquiry. This integration enhances the transparency and interpretability of AI outputs”.
Transparency is crucial for the acceptance and adoption of AI technologies. Knowledge graphs provide clear insights into the processes behind AI decisions, which is essential to build trust.
Dr Peter Haase, the founder & chief scientific officer at metaphacts, noted, “A unique component of the Dimensions Knowledge Graph is the symbolic AI layer it provides, introducing an enhanced level of transparency and trustworthiness to AI applications.”
Furthermore, knowledge graphs have become invaluable for businesses, helping uncover hidden patterns and insights from vast amounts of data. They support better decision-making by providing a comprehensive view of interconnected data.
David Newman of Wells Fargo explained, “Knowledge graph technology has emerged as a viable production-ready capability to elevate the state of the art of data management, supporting customer 360, risk management, regulatory compliance, and more”.
Knowledge Graphs & Big Tech
OpenAI combines large language models (LLMs) with graph databases like Neo4j to perform retrieval augmented generation (RAG). This approach fetches relevant information from a graph database, which is then used to generate responses.
This method helps reduce hallucinations, provides up-to-date information, and leverages the relationships between data points to enhance the quality of AI-generated content.
By integrating knowledge graphs with Azure OpenAI, Microsoft enhances natural language processing (NLP) capabilities, enabling better entity recognition and relationship identification across large datasets.
Furthermore, Amazon also extensively uses knowledge graphs to improve its recommendation systems and manage content across its platforms. For instance, Amazon Neptune ML uses graph neural networks (GNNs) to make predictions on graph data, improving the accuracy of predictions by over 50% compared to non-graph ML techniques.
As described by Amazon, “To help Amazon’s recommendation engine make these types of commonsense inferences, we’re building a knowledge graph that encodes relationships between products in the Amazon Store and the human contexts in which they play a role — their functions, their audiences, the locations in which they’re used, and the like.”
Knowledge graphs are crucial to AI and a core part of the internet. Google, for instance, has been using knowledge graphs to enhance its search since 2012, making them one of the most important technologies for the internet and future AI models.
