A group of four-year-olds in a preschool in Shenzhen gather around a small robot named Doubao on the floor. They talk to it, question it, and make fun of the way it thinks. These aren’t kids being watched by a screen. To learn how artificial intelligence thinks, they are being trained in a planned and organized way. Even though the scene is small, what it stands for is very big.
China has made a choice for the whole country. Learning how to use AI starts in kindergarten. Beijing‘s pilot programs now require students to take at least eight hours of AI classes each school year, starting when they are young and increasing the number of hours each year. The Minister of Education in the country has said that AI is the “golden key” to changing the country. This framework will be used across the whole country in a soon-to-be-released white paper, with the clear goal of AI leadership by 2030. That’s not a goal; that’s a plan with important dates.
In the United States, on the other hand, most school districts aren’t talking about how to teach AI. It’s about whether or not students should be able to get close to it. Only 6% of American public school teachers surveyed by the Pew Research Center in 2023 thought AI does more good than harm in the classroom. In response, many districts have blocked access completely. In a way, you contain it if you can’t control it. The issue is that containment is not a plan.
Researchers who study how kids think and learn in the early years have made a strong connection. They have known for a long time that learning a language is easiest when you are young. If a kid learns French when they are four, they might never have to use an accent again. If you start at thirty, you almost never will. Some experts say that the same reasoning works for AI as well. When kids learn fluency early, they don’t just learn how to use the tools; they also get a sense of how those tools work. One educational technologist said it was like the difference between classical music and jazz. Classical is structured, has rules, and is learned by heart. Jazz is made up of improvisation and instinct, and it relies on finding patterns when things are uncertain. In that light, AI fluency is like jazz. Also, jazz needs to begin early.
Another country that is falling behind is Japan, but the U.S. may be the most surprising case. South Korea, Singapore, and Finland have all taken strong steps, such as training teachers, creating national AI curricula, and, in the case of Finland, giving all of its citizens free AI courses. Even though the US has a lot of powerful technology, it still doesn’t have a national AI education framework. There is a patchwork of pilot grants, electives that can be chosen, and workshops that happen from time to time. It’s not nothing. It’s not a plan, though.
This gap has a real cost that goes beyond test scores and being ready for work. The countries that learn AI quickly will have long-term advantages in economic growth, cyber security, and yes, military power. China’s national AI development plan from 2017 said that developing talent would be the key to reaching its global goals. That paper is almost ten years old. People are being put to death in the classrooms in Shenzhen.
It’s still not clear if the political climate in the U.S. can lead to the long-term, coordinated commitment that AI education needs. The way American education policy is set up means that changes happen slowly and unevenly across the country. There is still time to talk about whether ChatGPT helps students cheat, but the window is closing faster.
One honest ten-year-old American student said it straight out: she uses AI on her homework even though it’s technically against the rules. Why did she do it? She knows what will happen. She wants to be all set. That line of thinking is hard not to find more convincing than most of the policy debates going on above her head right now.
