A group activity that tangibly embodies the difference between two types of learning — you will work with your group on paper cards before the machine explains how it does the same.
The activity is in two parts — in each part you will find a different set of cards with clear instructions. Cut the cards and start with your group, then discuss your results with the other groups.
Each group works on only one card set from Part One, then your group compares its results with other groups — notice how answers can differ!
Group the cards into logical categories in your opinion
There is no right or wrong answer — follow your intuition
✂ Cut the cards along the dashed line
✂ Cut the cards along the dashed line
✂ Cut the cards along the dashed line
✂ Cut the cards along the dashed line
Classify each card into one of the two categories specified at the top of each set
The classification is pre-defined — does your group agree on each card?
✂ Cut the cards along the dashed line
✂ Cut the cards along the dashed line
✂ Cut the cards along the dashed line
✂ Cut the cards along the dashed line
In Part One, you decided how to group — this is exactly what the machine does in Unsupervised Learning:
it looks for shared patterns without anyone telling it the classifications.
In Part Two, the labels were ready and you had to apply them — this is what the machine does in
Supervised Learning when we give it pre-classified examples from humans.
Think: If you gave these cards to someone who does not know Arabic, would they have been able to classify them in Part Two?
This shows why the quality of training data is very important for AI.