Combating Generative AI’s Hallucination Problem

For all its benefits in optimizing processes and informing decision-making in businesses, generative AI’s credibility for real use cases is hindered by a lack of accuracy, transparency and explainability in the content it generates – otherwise known as generative AI “hallucinations.” 

Why Hallucinations are Problematic for Businesses

Businesses trying to move forward with Gen AI deployment for more practical results are grappling with a staggering 41% Large Language Model (LLM) hallucination rate. The difficulty hallucinations create means decision-makers are left gambling that the result of a prompt will be correct, which could have a significant impact on the direction of a project or even the bottom line. Without a research breakthrough, LLMs will remain this way; they are not sentient and cannot replace human intelligence. 

What Causes Hallucinations

The relative nascency of generative AI means IT teams don’t yet have total clarity on the best way to create, train, and maintain the LLMs that power these systems. Traditional LLMs – that rely on probabilities rather than definitives - aren’t set up to successfully extract the meaning from modern intertwined data. Drinking from an ocean of online content and potentially contradictory options, they often repeat false facts or, worse, fill in the blanks with plausible fabrications, generating hallucinations.

Related:The AI Skills Training Gender Paradox

Professor Yejin Choi has demonstrated how ChatGPT fails to deduce inferences that seem clear to humans. For example, if you tell ChatGPT it takes two hours for two shirts to dry in the sun and then ask how long it would take five shirts to dry, it will tell you the answer is five hours. generative AI in this instance does the maths and ignores the logic; it doesn’t matter if you put 10 shirts out, it will still take two hours. Data professionals are therefore turning to trusted, well-known technologies to help make the outputs of generative AI systems more reliable.

Bringing LLMs to Reason with Graph Technology and GraphRAG

GraphRAG, a technique that adds knowledge graphs to Retrieval Augmented Generation (RAG), has become essential for addressing hallucinations. RAG enables organizations to retrieve and query data from external knowledge sources, giving LLMs a logical way to access and leverage their data. Knowledge graphs ground data in facts while harnessing explicit and implicit relationships between the data points, forcing the model to focus on the correct answers. The result makes generative AI outputs accurate, contextually rich, and – critically – explainable. 

Related:Rethinking Finance With an AI-First Approach

Helping to Verify how LLMs Respond

Another option to make LLMs more accurate is to use them for generating knowledge graphs. Here, an LLM processes large amounts of natural language to derive a knowledge graph. While the LLM is opaque, so to speak, the generated knowledge graph is characterized by transparency and explainability. 

In sectors such as pharmaceuticals, for example, the use of a complex corpus of internal and external research documents makes it crucial to be able to prove the origins of experimental results. Knowledge graphs in this industry represent a powerful tool for organizing and analyzing constantly growing data – and the nodes and relationships within it - to provide verifiable answers without hallucinations which could otherwise misinform clinical decision support or even drug discovery. 

Fortune Favors the Brave (Who Use Graph Technology)

CIOs should look beyond the frantic ChatGPT headlines. They have the opportunity to use knowledge graphs and graph data science algorithms to build LLMs that unlock the potential in their company's data stores and market data.