Nature of Humans and AI
An AI app can name thousands of kinds of birds from photos, more than most people can. Which sentence is still true?
1 The app is better than people at everything about birds
2 The app matches patterns but has never heard a bird sing
3 The app understands birds better than any scientist
4 The app can never make a mistake about a bird
B
Being very good at one task does not make AI smarter than people in every way. The app was trained on patterns from many bird photos, so it can name birds quickly. But it has never heard a bird sing or watched one build a nest, and it can still produce errors. A bird scientist uses experience and judgment that the app does not have.
What does it mean to reason?
A To use information and rules to figure out an answer
B To remember everything you have ever seen
C To guess without any information
D To have a feeling about something
A
Reasoning means using information to figure something out. You reason when you use clues to solve a riddle. Computers can be built to reason too, by following rules with the information they are given.
A school wants to build an AI tool that helps every student in the school. Whose information should go into the training data?
A- Only the students who get the best grades.
B- Only the students in the oldest grade at the school.
C- Students from every grade and group at the school.
D- Only the students who signed up to help the teachers build it.
C
An AI works best for the kinds of people it learned about. If the data only includes some students, the system may not work well for the rest. Including many different people helps it be useful and fair for everyone.
A town wants to know which streets flood the most. An AI tool can study ten years of rain and street records very quickly. But nobody wrote down the floods on three streets, because no one was watching them. Should the town use the AI tool?
A- No, because AI tools cannot work with information about weather
B- Yes, but they should also ask people who live on those three streets
C- No, because the missing information makes the whole tool useless
D-Yes, and they should trust its answer completely to save time
B
The AI tool is a good choice for studying ten years of records fast, which people would find slow and hard. But it can only learn from information it is given, and three streets are missing from it. Using the tool and asking people fills in what the records left out.
What could help an AI learn to recognize apples?
A- A bedtime story only
B- A picture of a spaceship
C- A secret password
D- Many different apple pictures
D
Showing AI apples of different colors, sizes, and shapes can help it learn what apples look like. If it only sees red apples, it might struggle with green ones.
Can AI help people make art or music?
1 Yes, and now human artists are not needed anymore
2 No, AI cannot produce any artistic output at all
3 Yes, but people still make the creative choices
4 Only music, because pictures are too hard for AI
3
AI can generate pictures and music quickly by using patterns from many training examples. But people decide what they want to make, what feeling it should share, and when it is finished. Those choices come from the artist’s own ideas and life. AI can be a helpful tool, but the creative choices belong to people.
A bird-sorting app says it is 95% sure a picture shows a pigeon. Does that mean the picture is definitely a pigeon?
A- Yes, 95% means it cannot be wrong
B- Yes, apps are never wrong
C- No, apps are always wrong
D- No, the app can still be wrong even when it seems very sure
D
The 95% shows how confident the app is, based on patterns it learned. Confident means sure. But being sure is not the same as being right. If the picture is blurry or shows a bird the app never learned about, it could still be wrong. Always check a computer’s answer with your own eyes and thinking.
An AI system is being trained to understand spoken questions. What kind of data does it need most?
A- Recordings of many people speaking.
B- Drawings of mouths and ears from a science book.
C- Lists of every word that exists in the English language.
D- Photographs of people talking to each other at school.
A
Training data has to be the same kind of thing the AI will work with later. A system that listens to speech needs recordings of real voices. Pictures and word lists cannot teach it what speech sounds like.
On a class walk, you find a bug you have never seen before. You are with a trusted adult who has an app that can suggest what kind of bug it is, and a park ranger is nearby. What is a smart way to figure it out?
A- The app, because apps are faster than people
B- Nobody, because new bugs cannot be identified
C- Neither one, because you should just guess
D- The ranger, who works here and knows the bugs in this park
D
An AI app can compare your photo with patterns it has learned and suggest an answer, but its suggestion could be wrong. A ranger who knows the bugs in that park can help check it. AI and people can work together: AI can provide a useful suggestion, and a knowledgeable person can help verify it.
True or False:
Some map apps use AI to guess how long a trip will take.
True
A map app can use traffic patterns to estimate travel time. Roads may be busier after school than early in the morning. A traffic jam can still make the guess wrong.
True or False
AI can always detect when it does not have the answer to a question.
False
False. AI has no reliable way to detect the limits of its training data. When information is missing, it may still produce an answer that sounds sure. People are usually better at noticing when they don’t know something. That is one reason people should check AI answers.
A fruit-sorting computer only looks at a fruit’s color and shape. It does not look at the taste or the price. Why is that OK?
A- Taste and price are not real
B- Computers cannot notice anything
C- Fruits all taste the same
D- Color and shape are enough to tell the fruits apart
D
A computer does not need every detail, only the ones that help it do its job. A simpler version of something that keeps only the important details is called a representation. A map is a representation: it shows streets but not every tree. Choosing the right details keeps a rule simple and still correct.
A team wants to build an AI that can tell healthy plants from sick plants. Which data would be most useful for training it?
A- Photos of many healthy plants and many sick plants.
B- A long list of the names of every plant in the world.
C- Videos of people watering their plants in a garden.
D- Photos of the tools that gardeners use in their work.
A
Training data has to match the job you want the AI to do. To tell healthy plants from sick ones, the system needs to see many examples of both. Information about plants that does not show what they look like will not help.
True or False
Your class is voting on a field trip. Your teacher could use an AI tool to count the votes quickly, and it could also decide which trip your class would enjoy most.
False
Counting votes is a job AI does well, because votes are numbers you can add up. But what your class would enjoy is not a number, and only your class can say that. One half of this job fits AI and the other half does not.
An AI thinks every bird can fly. Which bird shows that this idea is wrong?
A- Robin
B- Eagle
C- Pigeon
D- Penguin
D
Penguins are birds, but they cannot fly. If an AI only learned from flying birds, it might make the wrong guess. Missing examples can lead to mistakes, so varied examples matter.
True or False
When an AI tool gives an answer, it means the AI believes it is true.
False
False. AI does not have beliefs. It predicts which words are likely to come next, based on patterns in its training data, and it has no way to check on its own whether its answer is true. When AI gives an answer that sounds sure but is made up, it is called a hallucination. That is why people need to check AI answers.
A travel app has one rule: “Always pick the fastest route.” Route A takes 22 minutes but has 2 train changes. Route B takes 28 minutes with no train changes. A young student traveling alone wants a simple trip. Which route does the app pick?
A- Route B, because it is simpler
B- Both routes
C- Route A, because it is faster
D- It asks the student what they want
C
The app only checks speed, so it picks Route A. Its choice is called a recommendation, which is a suggested choice based on information. The app does not know that changing trains could be hard for a young student. The details around a situation, like who is traveling and what they need, are called context. People can think about context, but the app can only use its rule.
An AI is being trained to recognize birds. Next to each photo, someone has written the bird's name. Why does that name matter?
A- It makes the photo load faster on the computer.
B- It tells the system what it is looking at in that photo.
C- It proves that the photo was taken by a real person with a camera.
D- It lets the system share the photo with other AI systems.
B
A photo on its own does not tell a computer what is in it. When people add a name to each example, called a label, the system can connect the picture to the right answer. Labeling takes people a lot of time, but many AI systems could not learn without it.
True or False
An AI system that picks songs for you to listen to knows what will make your best friend happy on her birthday better than you do.
False
A song recommendation system can use patterns from information about what people have listened to before to suggest songs they might like. But it does not know your best friend the way you do or know everything that matters to her. AI can make recommendations from patterns in information, but people may know personal details and context that the AI does not.
A game character is told, ‘Stay by the treasure chest.’ Why does it stay there?
A- It is scared
B- It likes the chest
C- It is following a rule
D- It learned to love treasure
C
Game makers can give characters rules to follow. That is different from training AI on examples. Not every smart-looking thing in a game is learning!
True or False
A friend who knows you well can often understand why you are upset better than an AI tool can.
True
A friend knows about your day, remembers what matters to you, and has felt upset before too. An AI tool only has the data it is given and the patterns from its training. The details around a situation, like what happened today, are called context. People are good at using context to understand each other.
A book-picking computer asks only one question: “Do you want a long book?” It keeps picking books you don’t enjoy. What would help it most?
A- Pick from fewer books each time
B- Ask the same question twice
C- Add questions about the topic and whether it has pictures
D- Pick a book at random
C
Length is only one thing about a book, and it probably doesn’t decide whether you like it. The things a computer checks to make a decision are called its criteria. Better criteria, like topic or pictures, help it make better choices. Asking the same question again or picking at random would not help.
Where does the data used to train an AI system come from?
A- The AI system makes its own examples while people use it.
B- People collect it from things like photos, writing, or records.
C- It appears inside the computer as soon as the program starts running.
D- It is copied automatically from other AI systems around the world.
B
Training data has to be gathered before an AI can learn anything. People collect it from things like photographs, writing, recordings, or records that were saved. The AI does not make its own examples.
You need to check the answer to 348 × 27 for your math homework. You have a calculator, and a trusted adult could help you use an AI chatbot. Which is the better tool for checking the answer?
A- The chatbot, because it has seen millions of math problems.
B- The chatbot, because it can explain the steps as it goes.
C- Either one, because both will always give the same answer.
D- The calculator, because it works the answer out exactly.
D
Different tools are designed for different jobs. A calculator is made to do calculations, so it is a better tool for checking this multiplication problem. An AI chatbot may be able to answer math questions, but it can also make mistakes. Choosing the right tool for the job can help you get more reliable information.
A heater turns on whenever a room gets too cold. What tells it to turn on?
A- Its favorite temperature
B- A dream it had
C- A photo of the room
D- A rule set by a person
D
A simple heater follows a rule: turn on when the room is too cold. It does not have to learn from examples. Some computers follow set rules, while some AI systems learn patterns from data.