Last updated:
A system where outputs are based on statistical likelihood rather than deterministic, fixed logic.
Unlike traditional software that always produces the same output for a given input (deterministic), AI systems provide the most likely answer. This inherent uncertainty is why AI governance requires continuous monitoring and human oversight, as the system can produce unexpected hallucinations or errors even when functioning correctly.
Real world example:
A rule-based calculator is deterministic (2+2 is always 4), but an AI chatbot is probabilistic; it might answer a question correctly 99 times and hallucinate a non-existent legal case on the 100th time based on statistical patterns in its training data.
