AI Is Changing Radically, but Is Korean Education Standing Still?

The world is undergoing a sea change. What are we teaching?

Artificial intelligence is advancing too quickly. Should I compare it to a volcanic eruption, a nuclear bomb, or a tsunami? I want to call it a sea change. A transformation on that scale is not happening once and then ending. It keeps happening, and it is even accelerating. The economy, culture, society, and education cannot keep up. Neither can our own awareness.

In one sense, I understand. Change is so rapid that society will find it hard to absorb. But understanding that difficulty is different from saying we can keep living as before. With the world changing like this, how should Korea’s education system change? This is a worldwide problem, not just a Korean one. Today I want to examine recent AI developments one by one, then discuss what we should learn and what we should teach amid these changes.

I want us to think about this through dependent arising. Technology does not exist over there while education exists separately over here. Changes in technology change how we work; changes in how we work change the economy and society. The education we and the next generation receive within that world must change along with it. So please do more than hear the news and say, “That’s impressive.” Join me in asking what human beings need in this world.

A world where games are made when you ask for them

First, consider Google’s announcement of Playground. The screen introduces a service for making games. The central idea is simple: tell AI what game you want, and it will make it in minutes. There is no requirement to know how to code. Type what you imagine into a text box and add an image if needed. It will create a game from those inputs.

The introductory video presents a stream of playful, unconventional ideas. Someone imagines a yarn-themed demolition derby; someone asks whether their own cat can be put in the game. There is a club-sandwich tower defense game with disliked olives as enemies, and a game in which you keep jumping toward an alarm’s snooze button. There is a cowboy game where you give the bad guys the boot, and a very slow sloth platformer. Instead of a person translating each idea into a programming language, you describe it in words and a game comes out.

This presumably does not mean it can make absolutely every kind of game yet. But the direction is clear: enter what you want, have it turned into something you can actually play, and share the resulting game with other people. When I looked at the screen, the sharing area near the bottom appeared to be temporarily closed. Whatever that temporary screen state meant, the change this service illustrates is enormous. The relationship in which making a game necessarily meant learning to code first is changing.

Think about how startling this is. Until now, if you wanted to make a program, you had to learn programming. A substantial technical barrier separated the person imagining a game from the person implementing it. Now you make a request in human language, and AI writes the code. A world is appearing before us in which I can make the game I want by saying what I want.

“Toad, toad” and the genie in the lamp

When I was writing my book, I used this example: “Toad, toad, I’ll give you an old house; give me a new one.” Say that, and the toad takes the old house and makes a new one for you. Another analogy is the genie in the lamp. Rub the lamp, the genie appears, and when I make a wish, it does something for me. Our world is finally becoming a place where saying “Do this for me” gets it done.

For now, that applies only in some areas. But I think we are heading toward a world where we can make such requests about almost everything. In the past, I even imagined this: what if you wanted to project a blast like a martial-arts movie character, fire an energy wave like someone in Dragon Ball, or launch a fireball like a wizard? I would say “Fireball,” and it would fly.

I might have no idea how the fireball was generated or how it traveled. AI would make it happen. At this point, you might feel I am talking about an extremely distant future. Fair enough. But from the perspective that sufficiently advanced science will look like magic, the future really does seem likely to be a magical world. In today’s ability to make a game from a spoken request, I see a sign of that future.

From AI that explains in text to AI that shows you

The second development is OpenAI’s GPT-6 and Intelligent UI. Its introductory video makes the idea easy to understand. Someone who has moved into a new home is choosing wallpaper or paint colors, but every description is verbal: peach, a light warm orange with a gentle creamy feel, cobalt, a medium-dark royal blue. Can we immediately picture the precise color just from reading those words?

Travel presents the same problem. A San Francisco travel guide is full of text. There is plenty about where to go and what to see, but it is hard to grasp the plan at a glance. Admittedly, I wonder whether a modern guidebook would really contain nothing but words and no pictures. The same goes for instructions for folding paper or assembling a bicycle. How frustrating would it be to follow instructions that contain not a single illustration? You have bicycle parts all around you, but after reading the instructions you still cannot assemble them.

This is the video’s central point: sometimes words are not enough. So the proposal is to turn verbal information into something visual. Show a map for the trip, make bicycle assembly understandable by sight, and directly show the paper-folding process and the wallpaper colors. Instead of simply writing a longer explanation, provide an answer in a form that makes the intended task easier to understand and carry out.

The important point is that this goes beyond attaching one picture. Answers incorporate visual material and interactive elements you can manipulate. They show how things connect, make complicated subjects easier to learn, and quickly create tools for the task at hand. Even when I ask the same question, I receive an answer that is easier to understand and use.

Answers change shape for cooking, clothing, travel, and learning

Consider preparing a Sunday roast. The earlier GPT-5.6 Instant answer continued as a long stretch of text, describing what to prepare and how to cook it. With Intelligent UI in GPT-6 Instant, visual material comes first. Images accompany the menu, how to cook each dish, and how to allocate time. The overall cooking plan is much easier to take in.

The same applies to wardrobes and outfits. Suppose you are spending a week in Italy and attending a wedding over the weekend. Ask what to wear, and it presents outfits visually for each day. You can see the first day’s arrangement, then the second and third, along with checklists and preparations for a picnic. There is less need to read a long passage and reconstruct the plan in your head.

What about a road trip? It is not as though earlier answers had no images at all. But now the answer is much more visual and includes a map. My understanding was that it would use an existing mapping system, such as Google Maps, and connect it to the itinerary through programming. You can click and inspect it directly. It shows where to go each day, what to see, actual photographs of the locations, and how much money the trip will cost. After building the itinerary, it takes follow-up questions.

Learning changes too. Ask what GDP is, and instead of merely defining gross domestic product, it helps you visually understand its components. If you want to film with a drone, it shows which drone to use and how to film. Problems that are difficult to follow entirely in your head, such as the Monty Hall problem, become easier to explore. It can also create a bill-splitting calculator, a retirement savings calculator, or a retro game in the form you need, right there.

AI began with LLMs, large language models, and expanded into LMMs, large multimodal models. Even so, there were still shortcomings in how easily users could understand and use the results. This change is about producing much more visual, manageable answers to the same questions. AI arranges information in a form the user can readily understand. That too is a major change.

Benchmark scores are high. What about real accounting work?

Next is a human-baseline benchmark comparing accountants with AI. AI benchmark scores have risen rapidly in recent years. Yet there was a problem: adopting AI did not seem to be producing substantial productivity growth in U.S. economic statistics. People kept saying how capable AI was, but others felt it did not perform very well when actually assigned work. Trillions of dollars were going in without an obvious productivity increase. That was the paradox.

To investigate this gap, researchers gave simplified versions of tasks from the APEX accounting benchmark to 12 junior accountants and compared them with AI. At first, the word “junior” suggests someone just starting out. That was my initial impression too. But reading further, I found that the participants averaged about five and a half years of accounting experience and were all licensed CPAs.

An average of five and a half years could mean some had two or three years of experience while others had seven or ten. Yet they were labeled juniors. Consider how difficult it is to qualify as a CPA in Korea. It is a demanding examination, comparable in people’s minds to the Grade 5 civil service examination or other higher professional examinations. These were qualified people with real experience, and they were given the same accounting work as AI.

On medium-length, well-defined accounting tasks, frontier AI models were faster and more accurate than the junior accountants. They even outperformed the best accountant in the study. Just 18 months earlier—a year and a half—the best AI models had fallen below the accountants’ average score of about 37%. Now they were approaching 100% accuracy on the same tasks. Looking at the graph, I pointed out that this change had already happened in 2025, and it was now 2026. The interval is remarkably short.

The tasks comprised four realistic month-end close scenarios. Participants had to examine a company’s working files, find the necessary figures, calculate them, and prepare a results table. The study was originally designed to measure augmentation: how much people improved by using AI. But the models achieved perfect scores on their own, leaving no room to measure further improvement. AI had already answered all the provided tasks correctly.

There was also a cost difference. On a per-task basis, the frontier models were more than ten times cheaper than humans. That was the gap between hiring accountants to do this work and assigning it to AI. AI performed much better at carefully searching files and closely following instructions. The task designers had expected scores of around 30% from junior accountants and 55% from mid-level accountants, but even those expectations were not easy for people to meet.

What we must ask after “not fully replaceable yet”

The findings are said not to mean that the entire accounting profession can be completely replaced immediately. Not yet. Nor does progress in programming immediately lead everyone to say humans should be excluded altogether. But these results clearly raise questions about the future of work. I want to ask what comes next.

People have long expected AI to be good at accounting and to replace that work quickly, because it deals with numbers and tables. Now, in October 2026, it is outperforming accountants with an average of roughly five years of practical experience. What will happen in October 2027? What about 2028? What percentage of those earning their living as accountants will remain? I think we are reaching a situation where people will say, “Accountants are finished.”

Translators are already losing work. Simultaneous interpreters, book translators, and countless others are affected because AI translates accurately in an extremely short time. Accountants will face that change as well. The game creation we saw earlier is another example. Work that required programmers to write code can now become a program when requested in human language. What, then, happens to the jobs of the people who did that coding?

Mathematical problems and the time it takes to solve them

I have already discussed OpenAI’s earlier announcement of a solution to a difficult mathematical problem involving fluids. The new development is a set of mathematical results released on October 6, 2026. The model was given roughly 4,000 problems across number theory, algebraic geometry, analysis, computer science, and other fields, and results for about 700 were made public.

The reported achievements included finding a zero-free half-plane for the Riemann zeta function to prove a quasi-Riemann hypothesis, and obtaining a new upper bound for the matrix multiplication exponent. I am not a mathematician. Frankly, even after reading these descriptions, I do not know exactly what they mean. So how did people in the field evaluate them?

The mathematician Alex Kontorovich reportedly praised the result as worthy of a Fields Medal if a human had produced it. OpenAI’s Sam Altman described a new era of discovery. Mathematicians were said to be responding with a mixture of wonder and questions while checking the model’s claims through formalization in Lean. In other words, mathematicians themselves were examining the work.

I also think the time involved matters. AI is tackling, one by one, problems that remained difficult for humans for years. It is not spending an entire year on each problem, is it? The type and difficulty of the problem will vary, but we are talking about days at the long end and hours at the short end. AI is entering work done by mathematicians and scientists, including questions that have long awaited solutions.

What happened when researchers found problems suited to Claude

Next is news about Claude: 36 scientific manuscripts produced over three months. Anthropic published “Claude-shaped science,” a contribution by Harvard physics professor Matthew Schwartz, on its official blog. The article introduced a new approach to using frontier models in scientific research.

The approach was not to throw enormous scientific problems at the model indiscriminately. It was to identify problems well suited to current LLMs—so-called Claude-shaped problems—and apply the model closely to them. That does not mean AI is equally good at every scientific question. Among many problems, some may fit the characteristics of today’s AI. The researchers looked at what the model did well and found problems where that ability could be used.

Schwartz and Claude examined about 400 candidate problems. After considering what could be researched, he collaborated with 19 coauthors to produce 36 manuscripts across 18 fields. The starting point was the calculation of scattering amplitudes in quantum theory and particle physics. Again, the terminology is too difficult for me to understand in detail. But what follows catches my attention: tools developed in physics moved into other disciplines.

Claude reportedly noticed that the structure of Feynman integrals in physics was the same as that of Bayesian evidence integrals in population genetics and phylogenetic-tree calculations. Just hearing those terms does not tell me exactly what they mean either. Still, discovering connections between disciplines strikes me as significant. A method of calculation that works in one field becomes a tool for solving problems in another.

In ecology, the researchers reportedly used forest data from Barro Colorado Island in the Panama Canal to calculate Rampal Etienne’s neutral-theory equation, which had not been solved with actual data for 20 years. In population genetics, they solved a 30-year-old integral expression concerning the effects of natural selection on rare mutations. The account also described analyzing gnomAD and Gambian genomic data to identify the influence of gene conversion, a phenomenon missing from the standard model. Reading this does not mean I understand all the science. But it gives me a sense of the range of work underway.

The work extended to economics and linguistics. Researchers checked replication packages for 4,452 papers in five leading economics journals and systematically investigated discrepancies in digits and similar issues. They were checking whether figures in the papers matched the actual calculations. They also built AccStack, a database organizing accent systems across 6,072 languages.

The assessment was that this work demonstrates how to combine original human value judgments precisely with AI’s computational power. Humans are good at some things, and current AI is good at others, so the proposal is to combine them. That may describe the present. But here too I ask: how long will that relationship remain as it is now?

An 85-year-old Enigma cipher, deciphered in two days

I have discussed ciphers several times before. This report said GPT-6 Astra deciphered an Enigma message that had remained an 85-year-old unsolved problem, doing so in two days. Some of you may have heard of Enigma. It was the cipher used by German forces during the Second World War, the one Alan Turing confronted by developing the decoding machine called the Bombe.

The particular message was reportedly posted online in 2005 and had resisted numerous attempts at decipherment. The important point is that AI did more than supply the correct answer: it analyzed the cipher and developed the tools it needed. Without receiving detailed instructions for the decoding procedure from humans, it analyzed public material and formed hypotheses. It even built the necessary cipher simulator and Bombe itself.

The researchers assessed Astra’s two days of work as equivalent to months of analysis by a human researcher. AI is moving ahead at detecting patterns, identifying keywords, and repeatedly trying different approaches. Finding patterns is something AI is inherently good at. I believe it will become increasingly impossible for humans to keep up with AI in this field of decipherment.

Starting with one image to decipher Napoleon’s code

There was another cipher report on October 5. GPT-6 Astra reportedly solved a military cipher from the Napoleonic era that had remained undeciphered for 217 years, and did so in six hours. The starting point was a single image and a single prompt. It was shown a message that had remained on lists of unresolved historical ciphers for decades and asked to solve it.

The cipher was used by French forces during the Napoleonic Wars in 1809. It reportedly carried military instructions from the headquarters of Eugène de Beauharnais, Napoleon’s stepson and viceroy of Italy, to the French general Auguste de Marmont. A single line of ordinary French was followed by 24 lines of numbers, letters, and symbols. With 155 distinct symbols and roughly 1,300 substitution units, the message was considered especially difficult to decipher.

Astra first transcribed the symbols in the image into a form it could work with. It analyzed their structure and independently located and compared external sources, including scans of a 1969 military history book hosted on an academic platform. It then reportedly created and ran its own simulated-annealing decoder to solve the cipher. Reading an image, searching for sources, and writing a program were connected within one process.

The request amounted to: “This is a French cipher from a particular period. Try to solve it.” Given that goal, the model searched the internet, found material, and built a decoder. Building a decoder means writing a program that can make repeated attempts. This is different from having a human specify every next step for the model to follow. The entire process—deciding what tool to build, finding clues, and carrying out repeated attempts—took about six hours.

It was not just a matter of discovering substitution rules. The model had to transcribe the message accurately from the image, classify its symbols, and reportedly correct its date by comparing other letters and historical material. The decipherment also filled a gap in the historical record. In an 1865 collection of Napoleon I’s correspondence, part of a letter addressed to Marmont stopped at “a handful of.” The newly deciphered message revealed material connected with the phrase “a gathering of rabble.” The complete sentence telling him not to be intimidated by a disorganized rabble was restored.

The central point is more than the solution of one old cipher. Starting from a single image, AI performed the whole task: locating and reading the necessary material, making tools, and obtaining a result. AI is actually doing this work now.

After programming, is mathematics next?

Axios argued that OpenAI’s mathematical advances have implications beyond mathematics. Programming was an early example of what happens when AI reaches a sufficient level in a particular field. As its coding ability began to exceed the average human level, the amount of coding done directly by humans began to decrease. As we saw with game creation, you can obtain a result simply by asking, without writing a line of code. Could mathematics be next?

The material OpenAI released on GitHub on October 6 consisted of 722 manuscripts on longstanding mathematical questions, organized into 372 groups of findings. Academics and researchers were invited to examine and develop the material further. They were being asked to investigate possible errors and whether the results could be taken in a better direction.

Responses mixed excitement with unease. A Rutgers University mathematician described a result as worthy of a Fields Medal if a person had done it. Other mathematicians questioned whether these findings and another solution announced the previous month were truly original breakthroughs or relied heavily on existing human work. Still others downplayed the usefulness of AI-generated mathematics.

Stephen Wolfram said that it is easy to discover new mathematics or generate trillions of theorems, but the problem is that most of those theorems would interest nobody.

Well, one can take a pessimistic view. But mathematics, like programming, offers ways to determine whether a result is correct. Mathematicians can scrutinize proofs, and proofs can be translated into formal languages that a computer checks line by line. In that respect, it resembles checking whether AI-written code actually works.

Software engineers have already experienced this transition. What happens to other fields? When change occurs in programming, should we assume it stays within programming? When it occurs in mathematics, should we assume it stays within mathematics? AI became better at Go than humans. Things are changing one by one. Do we think the other fields will remain exclusively human?

Even I, who anticipated it, struggle to keep up

We have looked at game creation, accounting, mathematics, scientific research, and decipherment. How does it feel to you? It has been almost three years since I wrote my book about artificial intelligence. I did anticipate changes in this direction when writing it. Recently, however, the pace has become so fast that even I struggle to adjust my thinking.

In discussing creativity then, I asked whether AI was really incapable of producing mathematical proofs. Many people said AI could not be creative, could not do this, or could not do that. I wrote that it could certainly produce mathematical proofs. I thought the belief that creativity belongs only to humans might itself be a misconception, and that we should ask whether people making such categorical statements sufficiently understood what creativity is.

Now we live in a world where AI produces mathematical proofs, does accounting, conducts scientific research, and handles the programming required to make games. It also turns explanations in human language into colors, pictures, tables, and maps. Had I described today’s news a year ago, people might have said, “Oh, come on.” Yet in October 2026, this is already our world.

People using the change, and people who do not yet feel it

How much of this change do people actually experience? In the United States, only about 2.2% of people using AI are said to pay for a paid version. That means people who go beyond the free version and pay, not everyone who has tried AI at all. I suspect Korea’s figure is somewhat higher. Recent statistics also placed Korea among the countries where AI is heavily used.

Koreans may be quick to adopt new things; general educational attainment and the large number of university graduates may also contribute. In any case, hearing that the U.S. paid-use figure is so low makes me think most people have not yet fully grasped how capable AI has become or how urgently everything is changing.

Some people may have little reason to use AI as they continue their usual work. But those who have used it in areas where it has already surpassed a certain human level feel differently. It is so convenient that being told to do the same work without AI would make returning to the old way difficult. I have experienced that myself.

Building the Baroboneun Bulgyo website without a line of code

The Baroboneun Bulgyo website, barobul.org, was made with AI. It was built in the GPT-5.5 period, so I think asking for a redesign now would produce something more attractive. The current design falls somewhat short of what I want. But apart from that dissatisfaction, the process of making the site had already changed.

I am someone who can build websites and program. I can work on both front ends and back ends, as well as HTML and styling. Yet I did not write a single line of code for this website. I did none of the front-end, back-end, or HTML work directly. I only made requests in words. I said what I wanted built, and it was built.

The site’s articles are produced in a similar way. When I upload a video, YouTube recognizes the speech and creates automatic captions. Those captions are full of errors. I give them to AI and ask it to infer from the context what I originally said and correct them. I then have it write an article from the corrected captions, fixing spelling and spacing, tidying repetition, identifying the points, and organizing the writing.

Open an article on screen, and you can see an opening such as: “Hello. This is the Arhat’s live broadcast. It is Friday, October 2, 2026. I am Yoon Deok-man, the 21st-century Arhat, here to talk with you.” My spoken words are corrected and turned into writing. In these tasks, I already feel AI is much more capable than the average person. It builds the site, corrects the captions, and writes the articles. The work continues without my writing the code.

If a month’s work takes four days, what happens to hiring?

I spent more than 20 years of my life coding and programming. I worked on styling, images, HTML design, and front-end and back-end architecture. Now AI can handle more than 80–90% of that work. Anyone who has done this professionally would find that startling.

I made that site by myself with AI in four days. Doing it the old way without AI would have taken at least about a month. That would have required two or three people working together, not just me. Yet I completed it alone in four days. That is a change I have experienced directly.

Suppose I were running a web agency. Would I need to hire more employees? Unless demand for websites suddenly grew more than tenfold, the agency’s workload would not change very much. But if its capacity to process that work increased tenfold, why hire anyone new?

It does not stop at a hiring freeze. If the agency already had five coders or programmers, might it decide that some were no longer needed? If the workload stays the same while efficiency rises sharply, employment is bound to be affected. That is what has now arrived in the world. I have often discussed this in terms of economics and labor. Today I want to move from there to education.

Why do laws and administration always move late?

Social systems, laws, administration, and government are said to change last. Why? When science and technology advance and new social problems emerge, government moves to address the problems afterward. It would be good to prepare before trouble arises. But how often have you seen laws changed in anticipation of a problem that has not yet occurred? I have hardly ever seen it.

Even after multiple incidents, laws change very slowly and conservatively. Suppose a child dies in an accident in a school safety zone. How do we prevent it from happening again? The same kind of incident has happened more than once or twice. Yet it must become a public controversy and a prominent topic before legislation finally follows, much later. The system changes slowly even after something has happened.

I think our social system is extremely outdated. Government and social structures do not move ahead of events; they try to deal with the aftermath, and even that process takes a long time. Bitcoin is another example. Before it appeared, or immediately afterward, were laws put in place on the grounds that “this new thing requires advance preparation”? Or were laws made only much later, after various problems appeared and people debated how to handle them?

The same thing is happening in education. The world changes first, and institutions follow late. But the world does not pause and wait for institutions to catch up. AI has already begun taking over work done by people, yet education continues to train people for the world that preceded that change.

Society is overturned every day. Is education the same as years ago?

Society is moving toward a situation where most programmers are no longer needed. I think demand for accountants may also drop sharply as soon as next year. Many people who worked in translation are already losing work. Programmers who make games increasingly risk graduating with nowhere to find employment. What about mathematics and scientific research? We believed mathematicians’ work could be done only by mathematicians—only by humans. That premise itself is changing.

Yesterday was different from today, and tomorrow will be different again. Each new announcement turns the world upside down, to the point where it is difficult to know what the world will look like in a year. What about education? What has changed between this year’s education and last year’s, the year before that, and the year before that? There may have been subtle adjustments, but has anything changed fundamentally enough to match this transformation?

I do not feel that it has. On one side, jobs and tasks change. On the other, children are still made to study the same things and are ranked through the same examinations. They are told to repeat the old route to success without considering the world that will await them when they enter society.

Watching my third-grade son’s English lessons

My son is in the third grade of elementary school. When I was his age, Korean elementary schools had a different name. As the education system changed repeatedly in the intervening years, I was not following the details of classroom teaching. Recently, watching my son’s lessons, I saw that English begins in the third grade. Rather than gradually learning to write and read the alphabet, the children suddenly start conversations with “Hello,” and then they are given dictation.

Does that not mean they are expected to learn A, B, C, D and study English in advance at home? These are third graders. I think that is itself a problem with the way elementary education works. But I want to ask a further question: why are we teaching English now? Why must everyone speak English like a native speaker?

I have studied English extensively throughout my life and used it. I can read English. Recently, though, I do not read English texts directly. More than half the documents we examined today were English sources that I read in Korean translation. I occasionally check an original for an unusual point, but I use translation to understand the overall context and content.

What I used on screen today was not even a translation separately requested from the latest AI model; it was Google Translate. In my view, its translation is somewhat weaker than that of the newest models. Give a page to a current AI system and ask it to translate, and it produces smooth writing much like a human translation. With translation at this level, should the entire population still learn English for exactly the same reasons and in exactly the same way as before?

Of course there are reasons to learn English. I am not saying every such need has disappeared. But we should question why everyone has been taught English on the premise that they must speak it like a native speaker. Twenty, thirty, or forty years ago, such education may have helped society considerably. Now things are different. Most people can live without being good at English and can use translation tools.

We are also moving toward a situation where I can speak Korean in front of a foreign visitor and that person hears English. If not this year, what about next year? Even now, a phone can transcribe someone’s English and display a Korean translation. Wasn’t that kind of use also discussed in connection with the recent APEC event? In such a world, I am asking why the purpose and method of teaching English must remain unchanged.

Education that trained humans as tools for calculating and translating

Education needs fundamental reform. In the past, people had to be trained as tools for particular fields because only humans could do those tasks. Translating tools and calculating tools were needed. Before computers, how were calculations for reaching the Moon or predicting a rocket’s trajectory performed? Humans worked on paper to determine where something would land and how much force it would need to overcome gravity and leave the atmosphere.

Once computers began performing calculations, we stopped doing those calculations ourselves. Do you routinely work through complicated calculations on paper? Accounting also had to be done by people, so they were paid high salaries, and qualifying as an accountant required extensive study and training. Human involvement in those tasks will gradually decrease.

How, then, should education change? I think about this every day. What will society become? What will remain for humans? I am not saying I already possess all the answers. But I clearly feel that we must move beyond the standards education currently clings to.

Consider a system that compares how much people have memorized, tests their recall of the sequence of Joseon kings, chemical symbols, or the years of historical events, and rewards those with higher scores. We have assessed human abilities and made people compete by solving derivatives and integrals at university, memorizing legal provisions, and performing complicated accounting calculations. If anyone can obtain comparable results through AI, what sense does it make to rank people and divide their worth by those standards?

Does this mean we should learn nothing?

We need to start from the reality of a society living with AI. Accept that AI exists, then ask how humans can use it better and live alongside it. Even with the same tools, some people will use them more effectively and produce better results. What abilities does that require?

Does that mean “we no longer have to memorize anything”? Really? Surely not. Humans still need language to communicate and talk with one another. We must consider what we need to know and understand to ask useful questions and make requests. The question is therefore not whether to abolish all learning, but what to memorize and what to learn.

Intellectuals, thinkers, philosophers, educators, and people actually working in different fields must gather to discuss this. Education must change as society changes. We can delegate what AI does well while developing the thinking, critical faculties, judgment, and language skills humans need. We should discuss and debate how to cultivate people with broad capacities who live and work with AI.

I think this also connects to a world less governed by authority. The standards that ranked people and conferred authority because they memorized material and earned high scores are being shaken. We should ask anew how to develop people’s ability to live with others, use tools, judge, and make requests. Education cannot keep treating competition and scoring as its sole purpose.

Do not screw up your child’s life for an outdated formula for success

I know that getting angry about a system’s failure to change does not alter the fact that governments, laws, and institutions move slowly. But the present transformation is comparable to the shift from a geocentric to a heliocentric view of the universe. The world is changing on that scale, yet education still uses the old system. It keeps assigning needless studies and homework and ranking students through the college entrance examination. Why must it do that?

Children are sent to private academies and made to study all day so they can become doctors, accountants, scientists, or programmers. They are relentlessly pressured to win the competition. I wish people would stop screwing up children’s lives to fit an old education system. Quite apart from the need to reform that system, society has already changed even if the institutions have not. You cannot educate a child simply by trusting the existing system.

What responsibility could a schoolteacher take for doing something different from established practice? They will continue as before. The person writing examinations will also keep setting the same questions, regardless of how society has changed. Schools, private academies, examination setters, universities, and the education system irresponsibly keep going as before. Society has been turned upside down, but children still receive the old education.

What about parents? Do they ask, “How should my child live in this changed society?” Or are they pressuring their children in 2026 exactly as everyone else has done, as they themselves have done, as people did ten years ago? We should not make humans do tasks that calculators and computers took over in the past while preparing children for jobs that may not even exist when they enter society.

How quickly can you multiply four-digit numbers? Why should speed at calculations you do not even perform in ordinary life be so important? In my view, today’s education keeps making children do the equivalent of that. Conditions have changed, but it continues training them in functions needed under the old conditions.

People who judge, play together, and think broadly

First, the education system must change. Discussion and debate are the starting point. How will human society change as it lives with AI? What will people need in that society? What should be taught? We must ask what humans need if they are to judge, make requests, and do things themselves instead of entrusting everything to AI.

I am not saying we should teach nothing. Perhaps an era will come when people can simply relax and live without working. Even so, people need abilities in order to judge and ask for things. There is no reason to spend human time repeating work AI already handles, but developing thought, critical faculties, judgment, and language is another matter. We must cultivate people who can think more broadly and do things with AI.

We also need education for a happier life. What I mean when I think of examples from the United States and Japan is that the state should help people participate in varied sports, performing arts, and hobbies when they are children and young adults. It should provide facilities too. Instead of pointlessly sending children only to academies or sitting them down with strange homework, we should enable activities in which they participate and cooperate.

Sports are good in that respect. Even where competition exists, people can take part together as play and games. Education should help them develop social skills and learn how to live in a shared society, rather than one defined by antagonistic competition. We should think about people who can enjoy richer, more interesting lives, rather than people merely lined up by test scores.

Politicians, government officials, educators, teachers, scientists, and intellectuals should all come forward and speak about this. Yet they are not properly doing so. This rotten society is keeping quiet. I know institutional change is slow by nature. But must we also accept a situation where we are not even properly discussing how education should change?

What parents can do while waiting for institutions

Second, parents must recognize that society changes day by day, month by month, and year by year while education still follows an antiquated system. If your child or grandchild attends school, refuse to simply fit them into an inhumane system of competition and ranking. Stop treating an education that produces oddly rigid, stereotyped people as something natural and inevitable.

I want to say that school grades are not what matters. Help children think critically and become creative people, think comprehensively, and make things together with AI. What matters is helping them learn while enjoying themselves and grow into people with greater depth and breadth.

I hope parents will encourage reading and talk with their children about using AI effectively. Discuss what should not be handed over entirely to AI, what can be requested, what questions to ask, and what to ask it to do. I would like parents to offer their children and grandchildren that kind of educational environment.

In today’s broadcast, Dao Sunim was the only person participating in the chat, so we could not exchange many different perspectives. More questions and conversation would have made the discussion richer. I am not trying to determine the answer to education alone and hand it down. I want us to ask together what we should teach ourselves and the next generation who will live with AI, and what we ourselves must learn and change.

Education cannot change in isolation

When the world changes, I too am bound to change. This connects with dependent arising. We live by influencing one another. Can society’s conditions change while I alone remain fixed in my earlier thinking? I hope governments and politicians, intellectuals and scientists, philosophers, people working in the field, parents, and students will come together to discuss how we should change.

Education is, of course, intertwined with social and economic systems. If society remains competitive, can education easily change just because we tell it to stop competing? Educational change rests on the broad premise that social systems, economic systems, and culture must also change. That makes discussion all the more necessary. What must we do to move toward a society of living together and sharing?

The AI news I have presented and my frustration with education ultimately lead to this question. Systems change last. Then we must bring this issue to the forefront of society as soon as possible and discuss how government, society, and the economy should change in step with what is already happening. Some people are already being dismissed, some cannot find employment, and some are experiencing the disappearance of the usefulness of the work they used to do.

Will you turn away and pretend not to know because it has not happened to your child yet? Will your occupation forever be something that only you can do? Or will AI and robots take your place?

I hope you will join the discussion. What kind of society should we have? What kind of education should we provide? How should the education system change? The world is already changing. We must now discuss together what we need to learn and what we should teach the next generation.