Teaching autonomy before teaching AI

There is enormous pressure on robotics education right now to put AI at the front of everything. It is the word that gets attention, and it is genuinely where the field is going. It is also, for a learner who has never made a robot decide anything, the wrong place to start.

Autonomy is the concept underneath. A robot is autonomous when it takes an action that nobody specified in advance, based on something it measured. That definition is reachable at ten years old, and everything later attaches to it.

The first autonomous robot is three lines long

Read a distance sensor. If the number is below a threshold, stop. Otherwise, drive. That is autonomy: the robot decided, using information about the world, without being told when.

The interesting part is not the code. It is what happens when it does not work: the threshold is wrong, the sensor sees the floor, the robot stops a metre early. Every one of those is a lesson about the gap between a measurement and reality, which is the same lesson that matters later at much larger scale.

The progression we use

1. 01Sense: read a value and say out loud what it means. Nothing moves yet.
2. 02React: one sensor, one threshold, one action.
3. 03Decide: several inputs, competing conditions, a rule for which wins.
4. 04Remember: state (behaving differently because of what happened earlier).
5. 05Adapt: change the rule based on measured outcomes, rather than editing it by hand.

Step five is where machine learning becomes a genuinely useful idea rather than a slogan, because the learner has personally hit the ceiling of hand-written rules and wants something better.

What we deliberately do not do

We do not put a pre-trained model in front of a beginner and call the result AI education. A learner who calls an API and gets a label back has learned an interface, not a concept. Worse, they have learned that the interesting part is somebody else's.

By the time our learners meet computer vision or a trained classifier, they have already written rules that could not cope, measured why, and wanted something better. The vocabulary then attaches to an experience rather than replacing one.