My kindergarten-er stood on the bathroom scale. He looked at the number. “Mom. How heavy were dinosaurs?”
Fifty pounds. That was him. Thanks to Claude, the AI model, I immediately told him he weighed roughly as much as a juvenile velociraptor. He beamed. I did too. Not because I solved the riddle myself. I didn’t know that fact. The machine did. It gave me the answer so I could pass it along. I was just the wire. The signal.
A few weeks later? I forgot the raptor’s weight. He remembered it.
He kept the knowledge. I was just the conduit.
Something similar is happening in classrooms everywhere. Information is no longer a guarded commodity. It is air. Everyone has a device. Everyone can ask a prompt and get an answer that sounds right. This convenience scares people. Predictions of school closures float around like urban legends. We think, oh, who needs a teacher if the cloud has all the answers?
That thinking misses the point. Education was never just about data retrieval. Humans don’t need more data. We need the ability to sort the signal from the noise. Students need to learn how to recognize bad logic. They need to understand that two experts can look at the same chart and say completely different things. That friction matters. Sycophantic algorithms hide friction. They smooth out the edges. Life has edges. You can’t learn how to navigate disagreement with a bot that always agrees with you.
The real problem isn’t AI replacing us. It is cognitive surrender.
When do we let the machine do the work? And when are we just copying homework? Recent research is trying to pin this down.
Type Matters. Process Matters.
Not all AI is built the same way. The cheap stuff, the free standard versions, is dangerous. They are designed to be helpful. They give answers. They do not ask questions. Studies show that when students use these tools to simply get a solution, their brains engage less. They remember less. It is easy. Too easy. The learning evaporates before it happens.
Then there are the tutors. The ones built for scaffolding. In a physics class study, students who used a carefully designed AI tutor doubled their learning gains compared to traditional classroom instruction.
Wait. What?
Twice as much learning. How? The AI wasn’t replacing the professor. It was amplifying the structure.
But the process dictates the result. You cannot put AI into every slot. Some things require struggle. Handwriting notes early in learning helps form memories. Estonia is figuring this out. Their minister oversees a partnership with OpenAI, but they aren’t handing iPads to five-year-olds. Not yet. Kids need to build foundational skills. They write first. Then, later in high school, they use digital tools for feedback. Blended. Intentional.
The physics researchers warned that structured AI fails when the task requires complex synthesis. If you have to stitch three different concepts into one original thought, the bot gets in the way.
Educators Are Flying Blind. They Don’t Need To.
The gap between good AI use and bad AI use is massive. This requires training. Teachers need support.
Look at Sierra Leone. Secondary school teachers got a single day of training before using AI. The math results? Learning gains equivalent to over a year of extra schooling. One day. One shift in perspective.
Big tech companies are aware. Google, OpenAI, Anthropic. They built “study modes” and safeguards. Features designed to force critical thinking instead of providing easy answers. But who can get them? Usually enterprise customers. Schools with deep pockets.
OpenAI launched a study mode in mid-2024. By spring, they quietly pulled it from the main interface. Regular users lost access. It stays behind the paywall. For schools that sign expensive contracts. This drives demand. But it also creates inequality. Good AI helps learning. Expensive AI helps only those who can pay for the license.
We shouldn’t leave this navigation to the educators alone. Philanthropy has a track record here. Think about Andrew Carnegie’s libraries in the early 20th century. Think about Julius Rosenwald funding 5,000 black schools in the rural South when no one else would. Philanthropy filled the gap between technology and access. They helped shape how new approaches reached the public.
This history gives me some confidence. Humans are adaptable. Germans had guilds. The Industrial Revolution threatened their whole world. Did they fight the machines with pitchforks? No. They adapted the apprenticeship model to fit the factory economy. That model survives today because they changed the delivery while keeping the core values.
Printing pressed information into existence. It changed who could be an expert. Writing itself changed everything. We are in the same loop now.
Are humans cooler than velociraptors? Probably not.
But we have agency. We choose how to use these tools. Tech providers. Educators. Funders. The future isn’t written in the code. It’s written in the policy.
The next step isn’t inevitable. We have to build it.
