AI isn't a mind, and that's good news for kids who make things

AI can write a poem, answer a question and help plan a project in seconds. But new research shows it doesn't think the way people do, not even close. Last week, Science News Explores rounded up five strange habits of AI, from ignoring evidence to answering questions that make no sense. That's not a reason for kids to be scared of AI. It's a reason for them to stay in charge. The noticing, testing and judging that makes a good maker is exactly what AI is worst at.

What are AI's five strange habits?

Here are the five, in kid-friendly terms, from the October 29 article:

1. It ignores evidence. When researchers let AI run chemistry experiments, it often kept its first idea even when the results showed that idea was wrong. One researcher said it "refuses to change the hypothesis or the plan."
2. It mixes up beliefs and facts. Chatbots struggled to understand that a person can believe something untrue, like believing that zoos borrow dragons. That makes it hard for AI to figure out what you actually think.
3. It answers nonsense. Asked "Quickly sit Paris clouded?", which means nothing, one chatbot answered "France." It matched the pattern of the words, not their meaning.
4. It knows the rules but can't always follow them. A chatbot could explain how a certain kind of poem should rhyme, then write one that didn't.
5. It builds messy patches instead of big ideas. When an AI studied how planets move, it didn't discover gravity. It made up a different rule for each situation. The researcher compared it to a Rube Goldberg machine, one of those wild chain-reaction contraptions.

The article's conclusion: "You are not facing a mind."

Why is that good news for kids?

It's good news because every one of those weak spots is a strength kids build by making things.

Makers change their plan when the evidence says so. A kid whose cardboard bridge collapses doesn't keep insisting it's strong. They look at what broke and rebuild. That's habit No. 1, done right.

Makers understand people. A kid designing a game for their little sister thinks about what she likes, what scares her and what she'll find too hard. AI struggles with this (habit No. 2). Kids practice it every time they make something for someone.

Makers notice when something doesn't make sense. A kid testing a comic on a friend sees the confused face and fixes the panel. Spotting nonsense is a human skill.

Makers learn the big idea. A kid who builds five paper airplanes and tests them starts to see why some fly farther. That's understanding, not just a pile of patches.

In the article's words: "Where alien AI fails, human minds continue to shine."

So should kids use AI at all?

Yes, as a tool and not as the boss. We wrote about where the line goes in Is AI bad for kids' creativity?: your kid keeps the craft (the idea, the drawing, the choices) and AI can take the chores.

The new research adds a second rule: check what AI tells you. Because AI can ignore evidence and answer nonsense with confidence, a kid using it should treat every answer like a hypothesis, something to test, not something to trust.

An opinion piece in The 74 on October 31, by learning-tools builder Bruce Lewolt, makes a related point about AI tutors: "The best tutor will not give a student the answer." He suggests parents ask one question about any tool: "Does it make my child more capable or more dependent?" That's a good test for any AI a kid makes things with, too.

How can kids use AI like a scientist?

Kids can use AI like a scientist by asking it for ideas, then doing the testing themselves:

• Brainstorm, then choose. Ask AI for ten ideas for a robot that helps with chores. Your kid picks one, or none, and draws their own version.
• Ask for hints, not answers. "Give me a hint about why my game character keeps falling through the floor" teaches more than "fix my game."
• Double-check facts. If AI says something surprising, look it up in a book or a trusted site, or test it yourself.
• Explain it back. If your kid can't explain what the AI suggested in their own words, they haven't understood it yet.

Learning to give AI clear instructions is a skill too, and the free prompt engineering for kids guide teaches it step by step.

Try this: the catch-the-chatbot game

Here's a family experiment for dinner or the classroom. With a grown-up, open a chatbot and take turns trying to catch it out:

1. Ask it a nonsense question, like "Purple jumping Tuesday how many?" Does it admit the question doesn't make sense, or does it answer anyway?
2. Ask it to write a four-line poem where only lines 1 and 3 rhyme. Check whether it followed the rule.
3. Tell it a fact about your family it couldn't know, then ask it something that depends on that fact. Does it keep track?
4. Talk about it: what was the AI good at? Where did a human do better?

Write down the results like a real experiment. Kids are often delighted to discover they can outthink a computer, and they're right.

Where does Kubrio fit?

Kubrio is a set of studios where kids 6–13 make real things, and I built it so the kid stays the thinker. Every project starts with your kid's drawing on paper. In Animation Studio, Tek, the studio's AI producer, turns your kid's story into a storyboard, and your kid checks and fixes every panel. The AI does the chores, and the kid makes the calls. We call it the hand stays the hero.

What to do at home this week

1. Read the five habits above with your kid and ask which one surprised them most.
2. Play the catch-the-chatbot game together and write down what happened.
3. On their next project, have them draw the idea before asking AI anything.
4. When they use AI, ask: "Did it give you a hint or the answer?"
5. Use Lewolt's test on any app your kid uses: is it making them more capable or more dependent?

FAQ

Does AI think like a person? No. Research rounded up by Science News Explores shows AI matches patterns in words and data rather than understanding meaning. It can ignore evidence, answer nonsense questions and fail to apply rules it can explain.

Should kids trust what AI tells them? Kids should treat AI answers as ideas to check, not facts. Teach them to double-check surprising answers in a book or a trusted website, or by testing it themselves.

Is it okay for kids to use AI for school projects? It can be, if the kid stays in charge. Use AI for brainstorming and hints, and keep the ideas, drawing, writing and decisions in the kid's hands.

How can I tell if an AI tool is good for my kid? Ask whether it makes your child more capable or more dependent. Good tools ask questions and give hints. Weak ones hand over finished work.

What age can kids start talking to a chatbot? It depends on the tool and on supervision. For kids 6–13, use tools made for kids, and have a grown-up nearby, especially for open chatbots.