Every child will get AI. Not every child will get a teacher

Today, you hear concerns that AI tools will be a privilege reserved for wealthy children. But I think that in the near future, we’ll be sounding the alarm for the opposite reason: every child will have an AI tutor, while only those whose parents can afford it will have a human teacher.

Children from low-income families are already using AI more often than others. According to data from the Pew Research Center, one in five teenagers from families with an annual income of less than $30,000 does all or almost all of their homework with chatbots. Among teenagers from families with an income of $75,000 or more, that figure is 7%.

With AI, a child can learn more – or they may lose the ability to think for themselves. It depends on the adult who is there to guide them – or isn’t.

Why an AI tutor can’t replace a real teacher

About 15 years ago, a young man who looked like a homeless person came to a branch of my IT school. We assumed he’d come in to warm up or ask for money. But he said, “I’m willing to do any job; I’ll raise the money – just let me study here.” The young man did indeed raise the money, completed the course, then taught at our school for a while, and generally got his life back on track.

When I read that AI will help bridge the achievement gap between children from poor and affluent families and will become a tutor for every child, I remember this story. Would an AI tutor have given this young man the opportunity to gain knowledge on his own? It would have. But would he have actually acquired that knowledge? He admitted that he made it this far thanks to his teachers – people who believed in him and motivated him not to drop out.

Access to knowledge isn’t the same as knowledge itself. Sal Khan, who promised a personal AI tutor for every student on the planet, admitted this year that most students weren’t using it.

An experiment by the Wharton School showed that AI can harm learning. About a thousand Turkish high school students solved math problems with regular ChatGPT, with a tutoring version of the bot, or without AI. During practice, both bots improved results. But on the exam without the bots, the group with regular ChatGPT fell behind the group without AI by 17%: the students got used to the hints and didn’t learn to solve problems on their own.

Children offload all the hard work onto AI – for example, writing an essay, which requires considerable effort. But they don’t ask it to play a video game for them or eat dessert.

I come from a long line of educators and earned my Ph.D. in education. I can say that for centuries, despite societal development and technological progress, one of a teacher’s main tasks has remained the same. A teacher must create enough motivation so that children learn not because they have to, but because it’s interesting. This task is best handled by a real person who understands the student, not by an algorithm. Yet the bet is on the algorithm.

This year, the U.S. Department of Education included AI development among the priorities for the allocation of federal grants to schools. The document specifically mentions AI tutors and assistance for students falling behind in the curriculum. During the public discussion, commentators pointed out that AI in the classroom would reduce the amount of time children spend interacting with their teacher. The Department noted that it is important to preserve the human element of learning, but did not change the final text of the priority.

Who will teach our children

For now, education systems using AI still have a human teacher present. But for how long? AI is spreading faster than the internet and personal computers once did. Its capabilities are already outgrowing the benchmarks built to measure them; moreover, a model of the same level becomes tens of times cheaper within a year.

At the same time, a live teacher is a scarce and expensive resource. UNESCO predicts that by 2030, the world will need an additional 44 million teachers, with 4.8 million of them in Europe and North America. In the U.S. alone, there are 56,000 unfilled teaching positions.

Against the backdrop of a teacher shortage and potential budget cuts, AI seems like a simple solution. Teachers are the largest expense for American public schools: nearly 30% of all costs go toward teacher salaries. Many school districts are already looking for cheaper alternatives to human teachers.

I see a danger in the fact that the development of AI and the shortage of resources are pushing us toward the next step: what if we reduce the teacher’s role to that of a platform operator who monitors metrics and steps in only when the AI system detects a problem?

During the pandemic, many parents realized that distance learning falls short of in-person instruction. But society has already come to terms with this. Online schools have even become more popular. Just as imperceptibly, society may also get used to the fact that not everyone has a live teacher.

Schools now face the decision of whether AI will free up teachers’ time or allow them to get by with fewer teachers. This choice will determine whether every child has a person by their side who believes in them – just as the teachers at my IT school once believed in that young man. Or whether AI, which is supposed to level the playing field for students, will instead entrench inequality.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com