The A-Z of AI

Human-in-the-loop

A human being who guides an AI system as it learns.

The Human-in-the-loop is someone who trains, tests or tunes an AI system to help it produce more reliable results.

Just as a student mastering an activity for the first time would likely make mistakes, or misunderstand certain nuances, the same is true of AI.

A system taught to identify sea life, for example, may learn to distinguish an octopus from other creatures quickly because of its unique shape but struggle to tell the difference between fish that share similar shapes and colors.

A human-in-the-loop can intervene to teach the system which distinguishing features to look out for and guide it toward more accurate answers.

Giving a computer a human guide means two completely different kinds of intelligence are being leveraged simultaneously.

By providing data for an AI system to study and validating its efforts during its process of trial and error, humans lend the wisdom they’ve built up over a lifetime and combine it with the speed of computing. This collaborative dynamic helps to account for the drawbacks of both people and machines to achieve more accurate results.