Interactive experiment

Train the AI

You are the data scientist for a session. Samples have two imaginary measurements. Your job: sort them into Class A and Class B. The model will then learn from YOUR labels and you will test how well it generalised.

  1. 1Read the two measurement bars on each sample chip.
  2. 2Move every sample into Class A or Class B — by dragging it, or selecting it and tapping a zone.
  3. 3Press “Train the AI model” and watch the decision boundary evolve in real time.
  4. 4Later rounds add ambiguous and one mislabelled sample — labels are only as good as the data they come from.

Everything also works with keyboard and touch. Move at your own pace.

Round 1/4Best model: 0%
A

Class "A"

Samples that belong together

Move the sample to the class you believe is correct.

B

Class "B"

Samples that belong together

Move the sample to the class you believe is correct.

Unlabelled samples

9 data quality
Sample 6
Measurement A
Measurement B
Sample 8
Measurement A
Measurement B
Sample 4
Measurement A
Measurement B
Sample 2
Measurement A
Measurement B
Sample 3
Measurement A
Measurement B
Sample 5
Measurement A
Measurement B
Sample 7
Measurement A
Measurement B
Sample 9
Measurement A
Measurement B
Sample 1
Measurement A
Measurement B

Read each sample's two bars, then place it in a class. · Your labelled examples are the teacher. Clean, consistent labels → a reliable model.

i

These experiments are educational simulations with fictional data. In real care, AI only ever supports the final decision of a trained professional.

This is a teaching simulation using imaginary, made-up measurements. It does not classify real data, and its output is not a medical result.

Back to the playground