Skip to main content
ARTISAN

Inside the Machine

Explain why a fluent AI answer can still be wrong, tell training from using a model, and name common AI functions.

Fictional case. You are a researcher at the Riverside Learning Commons, a fictional community learning centre. A colleague pastes an AI answer into the team chat and says, “It sounds right, so let's use it.”

Step 1 of 12

Your task

Run a miniature prediction machine, change what it learns from, and decide how much a confident-sounding answer should be trusted.

A reminder of the method

A language model is built by showing it lots of example text, which is called training. Later, when someone gives it a question, that is called using it, and it produces text based on what it learned.

Pause this mission

Keep this mission password if you want to return later: HAZEL-BEACON-47. Enter it on the ARTISAN welcome page to start Inside the Machine. It opens the mission from the beginning; your earlier answers are not saved. Refreshing the page or closing this tab also restarts the mission.

Mission password

HAZEL-BEACON-47