School in the Intelligence Age
School in the Age of Intelligence
Continuing the reflections of Sam Altman on "The Intelligence Age," Dmitry Stavisky on "AI Teaching Assistants" (one, two), and Nikolai Andreichenko on the "School on Mars," I have written down my thoughts on how the field of education will change in the next ten to twenty years.
How Will School Change in the Intelligence Age?
Looking at ChatGPT query statistics, some conclusions can already be drawn. The two most popular types of requests are “write…” (text, essay, letter, presentation) and “solve this task” (usually homework). This suggests that schoolchildren and students were among the first to master “accessible intelligence” and learned to apply it for educational purposes.
We can recall how similar processes occurred during the previous industrial revolution. For example, with the advent of cars and trains, people walked less but covered exponentially greater distances. Walking is the most natural movement for humans—our ancestors are called "Homo erectus," meaning we began walking long before we developed intelligence. Nevertheless, with the emergence of new modes of transport, people did not stop walking entirely; they even invented treadmills, stadiums, and the Olympic Games to move more.
Similarly, with the emergence of "accessible intelligence," people are unlikely to stop learning. Instead, they will invent many new ways to train and enhance their intellect using AI tools. Sam Altman suggests that “this will look like magic to our grandmothers.”
Of course, there will be negative side effects, but we will find ways to counter them. With the advent of cars, obesity rates increased, and in some countries, diabetes became the leading cause of death. I believe the same will happen with intelligence in "Homo sapiens." Those who continue their intellectual development will become much smarter than their ancestors (us). Those who completely delegate decision-making to AI—from choosing socks to choosing a spouse—will degrade. And there will be those who lack access to new technologies and remain on the sidelines of progress. The intellectual gap between people will widen, potentially creating social problems, but the overall intellectual level of humanity will rise—just as the average daily distance traveled per person has increased over the last 150 years.
How to Prepare People for Life in Such a Future?
I ask myself three questions:
What should we teach?
How should we teach?
Who should teach?
1. What Should We Teach?
This is the hardest question, and we rarely ask it. The answer to “What should schools teach so that children can act independently and successfully in the age of intelligence?” seems to lie in answering another question: “Who is a person in the age of intelligence?” As Sam writes, “Many of the jobs we do today would seem like a waste of time to people a few hundred years ago, but none of us, looking back, would want to become a lamplighter.” Yet, modern humans and citizens of ancient Athens have more in common than it seems.
With the emergence of religious schools and later mass compulsory state education, the right to determine what to teach shifted from parents to politicians and religious leaders. They designed education systems to achieve their goals—determining the official language of instruction, their version of human history, and defining what is right and wrong. Even in developed democracies, influencing school curricula is nearly impossible today. Disagreeing parents must either relocate or send their children to private schools.
Who will decide what to teach in the intelligence age? Governments, private companies, and decentralized education models will all play a role in shaping curricula and policies. This is a big question. Perhaps we will create a system where everyone can customize their own education. Or we might delegate this role again—this time to artificial intelligence (i.e., its developers).
2. How Should We Teach?
This question is already clearer. Today, at a teacher's request, ChatGPT can research advanced educational methodologies, create lesson plans, generate assignments for each student, and adjust them based on individual progress. Dmitry Stavisky calls such AI agents "teacher assistants." This results in a classroom model enhanced with Iron Man’s suit. Will the system gradually modernize in this way, or will it be fully replaced by artificial intelligence? We will see in the next 5-7 years.
Many questions remain. What should classrooms look like if each student has a personalized program? What should lessons be like if each student progresses at their own pace? We are only beginning to approach these answers, and they will become clearer as AI integration progresses.
3. Who Should Teach?
For me, this is the most obvious question. Education, like medicine, is an expensive service because selecting and training specialists takes years, and their work benefits society only decades later. Meanwhile, salaries must be paid, accounting for 70-80% of any educational institution's budget. These long investment cycles create enormous costs, affordable only to governments, corporations, or the wealthiest 0.1%. As a result, there is a global shortage of teachers, and many who exist are of low quality.
Once we access technology that reduces the cost of education by hundreds of times, it will be widely adopted. Therefore, I have no doubt that AI tutors will become a commodity within 3-5 years.
AI solutions like Khanmigo or Squirrel AI cannot replace teachers entirely, but if the work is divided into small tasks and distributed among dozens of AI agents (bots using large language models), it can come close. For example, a team of AI agents from Praktika.ai helps users improve their English pronunciation, focusing on a British accent, by conversing on topics like 'How to get to the library.' One AI agent develops the lesson plan based on previous progress and goals, another generates discussion questions and assignments, a third plays a Londoner guiding you to the library, a fourth evaluates and corrects mistakes, and a fifth analyzes the entire session to suggest improvements.
These AI agent teams, specialized in one domain, are called vertical AI agents. They are designed and trained to solve problems within a specific field—such as education. Learning with such AI tutors follows the principle of “fake it till you make it”—you try something first, then learn the correct way. This mimics how young children learn by copying adults.
Today, there are 100 million teachers worldwide—yet they are still insufficient. Soon, billions of AI agents will join them, taking on some of the workload. This will increase the quality of education and the number of learning hours.
Will teachers lose their jobs? Technology made street organists, bar pianists, and silent film musicians obsolete. But top musicians gained billion-sized audiences, and listeners got access to their music.
What Will School Look Like in The Intelligence Age?
Here are my five key points:
“Accessible intelligence” is already transforming education from preschools to corporate institutes. This is an irreversible process.
Designing the school of the future today is impossible, but we can outline the key questions that need answers.
AI’s greatest advantage is that its intelligence is accessible and scalable, unlike human intelligence. The more AI agents we create, the more accessible education becomes, allowing more people to benefit.
We still need to learn how to use “accessible intelligence” correctly to achieve the best educational outcomes. I am confident that entrepreneurs will figure this out.
AI will not take teachers' jobs—teachers who use AI will replace those who don’t.
Andrew Ilingin
January 2025