Script Policy Agent Sorter

Run the Script Policy Agent Sorter MicroSim Fullscreen

You can include this MicroSim on your website using the following iframe:

<iframe src="https://dmccreary.github.io/robot-arm/sims/script-policy-agent-sorter/main.html"
        height="548px" width="100%" scrolling="no"></iframe>

Description

Sort eight situations into three kinds of control: a script, a learned policy or an agent. The deciding question is who chooses the next move: fixed code, a trained model, or a language model choosing tools. Used in Chapter 18.

Explore mode lists the three kinds, each with who decides and what the kind is good for. The quiz shows eight situations in a fixed order. You choose a kind in the menu, press Check, and read the reason.

Lesson Plan

Grade level: Middle school and up (age 12+). Time: 5 to 10 minutes.

1. Read the three kinds. In Explore mode, say where the knowledge comes from in each kind: the programmer, recorded demonstrations, or a language model and its tool descriptions.
2. Sort. Press Start. For each situation, ask who decides the next move, and do not be distracted by the hardware. Choose a kind and press Check. You get one try for each situation.
3. Read the reason. The feedback names the correct kind and gives the reason. Mastery is 7 of 8.
4. Think about it. Most good projects use a script for the motions, vision for the facts, and an agent only for the words. Why not use an agent for everything?

A program with a camera is still a script when code decides every step. A learned policy turns what it sees straight into actions and does not choose tools. An agent is the least predictable of the three, so it is used only where the job needs words or changing plans.

Specification

The learning objective, evidence of mastery and content for this MicroSim are specified in Chapter 18: Projects, Operation, and Teaching.