Spot the Learning Machine
You have been learning a lot about learning machines! Now it is time to put your skills to work. In this lesson, you get to be the detective. You will look at descriptions of different machines and decide: is this one learning from examples and getting better? Or is it just following fixed rules that never change? Get your detective hat on — here we go!
How to Spot a Learning Machine
Here is your detective checklist. Ask these three questions about any machine: Question 1: Does it get better over time as it sees more examples? Question 2: Did it learn what to do from data, instead of being given exact rules? Question 3: Can it handle new situations it has never seen before? If the answer to all three is yes, you are probably looking at a learning machine. If the machine does the same thing every time, no matter what — it is a programmed machine following fixed rules.
A learning machine improves from examples and handles new situations. A fixed machine does the same thing every time. When in doubt, ask: does this machine get better with experience?
Let us warm up with two easy cases. Case A: A microwave oven. You press the buttons. It heats your food for the time you set. Every time you press the same buttons, the exact same thing happens. It does not get better at heating food. It does not learn your favorite meals. Same input, same output, always. Verdict: Programmed machine. Fixed rules, no learning. Case B: A music app that recommends songs. The first week it suggests songs you do not like much. But every time you skip a song or replay one you love, the app notices. After a month, it knows your taste really well and suggests songs you love almost every time. Verdict: Learning machine. It improved from your feedback and examples!
Flashcards — click each card to reveal the answer
Now let us try some trickier cases. Case C: A robot vacuum that has a map of your house programmed into it and follows the same path every day. Hmm. It follows a fixed map. It does not get better at cleaning. It does not learn where new furniture is. Verdict: Programmed machine — even though it is a robot! Case D: A newer robot vacuum that uses sensors to notice when furniture has moved and figures out a new path on its own, getting better at finding dust over time. This one learns from its environment. It adapts. It improves. Verdict: Learning machine!
Do not let a fancy appearance fool you. A robot is not automatically a learning machine just because it looks cool. What matters is whether it gets better from experience — not how it looks!
A light that turns on at 7pm every evening automatically — is this a learning machine?
A photo app begins by misidentifying your pet but improves each time you correct it. Is this a learning machine?
Machine Detective Walk
- Go on a walk through your home — or look around the room you are in right now.
- Find at least five machines or apps (a toaster, a phone app, a television, anything counts).
- For each one, ask the three detective questions: Does it get better over time? Did it learn from data? Can it handle new situations?
- Write down each machine and your verdict: learning machine or programmed machine.
- Share your list with someone and see if they agree with your verdicts!
- Bonus: Can you find one machine that seems like it could be either? Talk about what extra information you would need to decide.