What is grounding used for in AI?#
By requiring the AI to base its answers on a provided set of trusted documents or data, grounding reduces the risk of factual errors and makes the AI's outputs more reliable and auditable for business use.
How does grounding work in AI?#
When a user asks a question, the system first searches a predefined knowledge base (like a company's internal documents, product manuals, or a specific database) for relevant information. This retrieved information is then given to the AI model along with the original question as context.
The AI is instructed to use only this provided information to construct its answer. This process constrains the model, preventing it from relying on its vast but sometimes unreliable general training data and forcing it to stick to the facts you've provided.
How does grounding make AI different from general-purpose AI?#
Grounding forces an AI to cite its sources from a trusted, limited dataset that you control. This is different from a general-purpose AI like the public version of ChatGPT, which answers questions by drawing from the immense and uncontrolled pool of information it was trained on from the public internet. It's the difference between asking an expert to answer using only your company's official annual report versus asking them to answer from their general memory.