Networks That See, Hear, and Read
Neural networks are not just ideas in a textbook. They are already working all around you! Right now, a neural network might be helping your family's phone recognize faces in photos, helping a doctor find clues in an X-ray, or helping a reading app turn spoken words into text. Let us explore the three biggest things neural networks can do in the world today.
Networks That See: Computer Vision
When a neural network is trained to look at images, it is doing computer vision. Computer vision networks can spot objects in photos, read license plates on cars, identify plants by their leaves, and check that food at a factory is not spoiled. To teach a vision network, engineers feed it millions of labeled photos. 'This is a stop sign. This is a dog. This is a ripe apple.' After enough practice, the network gets extremely good at recognizing these things — often better than a tired human eye.
Computer vision means a neural network that looks at images. Speech recognition means a network that listens to sound. Natural language processing means a network that reads or writes text. Together, these three abilities let AI see, hear, and read.
Networks That Hear: Speech Recognition When you talk to a voice assistant and it types out what you said, that is a speech recognition network at work. Sound waves are turned into numbers, and the network figures out which words those numbers represent. Networks That Read: Natural Language Processing Natural language means the way humans talk and write — full of slang, jokes, and complicated sentences. A natural language processing network (NLP for short) can read sentences and understand the meaning, answer questions, translate between languages, and even write new paragraphs. The chatbots and AI writing helpers you may have heard about are NLP networks.
Match each type of neural network to what it does.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
These three abilities are often combined. A video-captioning tool uses computer vision to watch a video AND natural language processing to write sentences describing what it sees. An accessibility app might use speech recognition to hear a question and NLP to write back a helpful answer. Neural networks working together can do things that felt like science fiction just twenty years ago.
Over the next day, keep a mental list of tools you use that might have a neural network inside. Phone cameras, voice assistants, autocomplete when you type, spam filters in email — all of these use neural networks.
What does a computer vision network do?
What does NLP stand for?
Spot the Network
- Look around your home or school for five minutes.
- Write down every tool or device you can find that might use a neural network inside.
- For each one, try to decide: does it SEE, HEAR, or READ?
- Examples to get you started: a phone's face-unlock uses computer vision. A voice assistant uses speech recognition. A spell-checker uses natural language processing.
- Share your list with a family member or classmate. See if they can add more!