Dev with Serdar
Learn coding in Python, Go and Rust from Serdar Yegulalp, software dev specialist and senior writer at InfoWorld.
Are AI Models Really Deterministic? Here’s Why They Often Aren’t
Overview
When we say something is "deterministic", we mean it delivers the same outputs for the same inputs. In theory LLMs should do that, as they're just mathematical models. But we don't get consistent results from LLMs even with the same inputs — and there's a slew of reasons for that. In this video I dive into why LLMs are designed to introduce randomness into the results, and why even without that feature it's hard to get them to deliver the exact same results from the exact same prompts.
