Skip to content
Exit presentation

How AI Learns I: From Rules to Data

The order you do it in

The lorry brings the crates. The clerk chalks a word on each one. What comes next?

Loading lab…

Collect, label, split, learn, measure. In that order.
Step 1 of 4
  1. Collect, label, split, learn, measure. In that order.
  2. In order: measure on the kept-back tray. Score 4 of 8.
  3. Swap: measure runs before split. Score 24 of 24.
  4. Same program, same crates. The order changed the count.

The rule

Next comes the split. The kept-back crates go onto a tray for testing. A count that flatters looks better than the program really is.

Collect, label, split, learn, measure — in that order. Measure on the examples it learned from, and the count flatters. Measure on the kept-back tray, and it means something.

The 24 crates are split. Eight go onto a tray. The program learns from the other 16. Then the tray is measured. Measured on the learn pile instead, the program copies their words.

Same program, same crates, two orders
OrderMeasured onScore
Correct orderKept-back tray, 8 crates4 of 8
Measure before splitEverything it holds, 24 crates24 of 24

Skip the label step and the crates carry no word. Then there is nothing to compare each guess with. The measuring step cannot run. Not being able to measure is not the same as scoring 0.

Jump to lesson

This track

  1. 0 Start here
  2. 1 A rule a person wrote, and a pile of examples
  3. 2 Keep some crates back
  4. 3 Right out of how many
  5. 4 Split it differently
  6. 5 Learning the examples instead of the pattern
  7. 6 Ask more than one
  8. 7 The order you do it in
  9. 8 Write the rule yourself
  10. 9 The big picture
  11. 10 Track quiz

Other tracks