Question

When an AI hallucinates, what is the fundamental process required to verify its output against an objective source of truth?

Accepted Answer

When an AI makes a mistake—which is called a hallucination—you must verify its output by using a process called grounding. Grounding means connecting the AI’s answer to an external, reliable source of truth. Think of this like checking a student&#x27;s essay against a textbook; the essay is the AI&#x27;s guess, and the textbook is the trusted source. To do this, you first take the specific facts the AI gave you, such as names, dates, or calculations. Then, you search for those same facts in a source you know is accurate, like a official document, a scientific study, or a trusted database. You compare the AI’s claim directly against what is written in that reliable source. If the two pieces of information do not match, the AI has hallucinated. The fundamental process requires these three steps: identifying the specific claims made by the AI, finding a trusted document or database that covers those claims, and performing a direct side-by-side comparison to see if the facts line up.

Home → All Courses → Engineering and Technology Courses → Universal AI Safety Framework → Flashcard

When an AI hallucinates, what is the fundamental process required to verify its output against an objective source of truth?

Community Answers