AI Explorers · lesson 4 of 24 · 25 min
You know your little brother is home before you see him, because you hear the door and smell his snacks. Machines need ways to notice the world too. They use sensors.
What is a sensor?
A sensor is a device that detects something in the world and turns it into a signal a computer can use. Your body has sensors too: eyes, ears, skin, nose and tongue.
| Human sense | Machine sensor | What it detects |
|---|---|---|
| Sight | Camera | Light and color |
| Hearing | Microphone | Sound vibrations |
| Touch | Touch screen, pressure sensor | Pressure and contact |
| Balance | Accelerometer, gyroscope | Movement and tilt |
| (none) | GPS receiver | Location on Earth |
| (none) | Infrared sensor | Heat and invisible light |
Notice the last two. Machines can sense things people cannot. Your phone knows which way it is tilted and roughly where on Earth it is.
From the real world to numbers
Computers only work with numbers. So every sensor has the same job: turn something from the real world into numbers.
A thermometer sensor might output 72 for the temperature. A microphone measures how air pressure changes thousands of times every second and records each measurement as a number. A camera measures how much red, green and blue light hits each tiny spot on its chip.
This turning-into-numbers step is called digitizing. After it happens, the computer has data. But data alone is not understanding.
Sensing vs. perceiving
Here is the most important idea in this lesson. Sensing is collecting raw data. Perceiving is figuring out what that data means.
A doorbell camera senses a grid of colors. Perceiving is deciding “that is a person holding a package.” The first part is hardware. The second part is where AI comes in.
Perception is hard. The same person looks different in sunlight, at night, wearing a hat, or turned sideways. The same word sounds different from a kid, a grandparent, or someone with a different accent. AI perception systems learn from many examples so they can handle that variety, and they still make mistakes when something looks or sounds very different from what they learned.
Sensors are everywhere in your neighborhood, too. Traffic lights can use sensors buried in the road or cameras to notice waiting cars. Automatic doors at the grocery store use motion sensors. Fitness watches use light sensors to estimate heart rate. Each one turns the world into numbers, and some of them hand those numbers to AI to make a decision, like when to change the light.
Try it
- Sit somewhere for two minutes and close your eyes. List everything you sense with hearing, touch and smell.
- Open your eyes. Pick three things on your list.
- For each one, write which machine sensor could detect it, or “no good sensor” if you can’t think of one.
- Now pick one and write what “perceiving” it would mean. For example: a microphone senses sound, and perceiving means knowing “that is my dog barking at the mail carrier.”
- Write one reason that perceiving would be harder than sensing for that example.
Check yourself
What does a sensor do?
It detects something in the world, like light or sound, and turns it into a signal or numbers a computer can use.
What is the difference between sensing and perceiving?
Sensing collects raw data. Perceiving figures out what the data means, like recognizing a face in a picture.
Name a sensor that detects something humans cannot sense directly.
GPS (location), infrared (heat), or an accelerometer measuring exact tilt are good answers.
Key words
- Sensor: a device that detects something and turns it into a signal.
- Digitize: turn real-world information into numbers.
- Sensing: collecting raw data from the world.
- Perceiving: figuring out what sensed data means.
Previous lesson: What AI can’t do (yet)
Back to AI Explorers
Next lesson: How computers see: from pixels to patterns