title: Probabilistic
aliases: [probabilistic, probability]
tags: [term]
category: AI & NLP
summary: A system whose output involves chance, so the same input can produce different results and no fixed rule explains any one of them.

Probabilistic

A process is probabilistic when chance is part of how it produces its output. Rather than following a fixed rule, it weighs many possible outcomes by how likely each is and picks among them, so the same input can give different results on different runs.

Most modern language models work this way. Given some text, the model estimates how likely each possible next word is and samples from those possibilities. This is what lets it handle messy, open-ended input fluently — but it also means there is no rule you can point to that explains any particular answer, and the model can produce something fluent-sounding and confidently wrong.

This is the opposite of a deterministic process, which always gives the same output for the same input.