Language has always changed. New words emerge, old words fall out of use, expressions once regarded as incorrect become established, and dialects or slang are sometimes absorbed into standard usage.
The language used by younger generations today also contains many abbreviations and expressions whose meanings differ from their original ones. But this alone is not enough to conclude that “modern people are losing their language ability.” Deliberately bending language and being unable to use anything else because one does not know the correct expression may look similar from the outside, but they are fundamentally different.
A person who understands formal language and chooses to simplify it when speaking with friends has a different level of linguistic ability from someone who does not know the formal expression in the first place, even if both use the same words. The problem is not language change itself. The deeper issue is that, in a future where artificial intelligence becomes embedded in almost every part of daily life, the need for people to understand language precisely, think for themselves, and express ideas in their own words may gradually diminish.
Language Is Also a Tool for Thought
Human beings do not think through language alone. We also think through images, space, sensation, equations, sound, bodily perception, and many other nonverbal forms. A large vocabulary therefore does not automatically mean greater intelligence.
Even so, language is an extraordinarily powerful tool for distinguishing abstract concepts, comparing them, and organizing them logically. Consider ideas such as causation, correlation, necessary conditions, sufficient conditions, probability, induction, deduction, and falsifiability. Someone who understands these concepts can recognize them as distinct and manipulate them mentally. By contrast, if a person processes very different phenomena through only a small set of expressions such as “I just kind of feel that way,” “that’s crazy,” or “that’s weird,” they may still be able to communicate, but it becomes much harder to classify ideas with precision.
There is a major difference between saying, “This policy is crazy,” and saying, “I agree with the objective of the policy, but the incentives created by its institutional design may work in the opposite direction from that objective.” In the latter, distinct concepts such as goals, institutions, incentives, and causation are separated and examined individually. Linguistic ability, then, is not simply the ability to write elegant prose. It is also the ability to break down vague internal impressions, structure them, and convert them into a form that can be examined and tested by others.
Will AI Improve Language Ability?
As artificial intelligence becomes widespread, the amount of information accessible to individuals will increase dramatically. If you encounter an unfamiliar word, you can ask about it immediately.
If part of a difficult text is unclear, you can have it explained. Papers written in foreign languages can be translated. A single concept can even be explained at the level of an elementary school student, a university student, or a specialist. In this sense, humanity may never have had an environment more favorable to learning. Yet having access to information is not the same as possessing that information as knowledge of your own.
The easier it becomes to obtain an answer through AI, the less cognitive work a person may perform along the way. Imagine a high school student who develops a question about a historical event. In the past, the student might have opened a textbook, checked the index, searched for related material in a library, read about the periods before and after the event, encountered unfamiliar people and events, and gradually assembled the relevant information. Along the way, the student would inevitably encounter knowledge unrelated to the original question. Finally, they would organize what they had understood into their own written explanation. With generative AI, it is possible to ask a question, receive an answer, and stop there. If we compare only the final explanation, the AI-generated answer may even be better. But the amount of work performed by the human brain is very different.
Information Encountered Is Not the Same as Knowledge
This distinction is important when thinking about the future of education. Being exposed to more information is not the same as gaining more knowledge.
Humans turn information into part of their own cognitive system by searching for it, understanding it, retrieving it from memory, connecting it to other knowledge, and reusing it. But when AI performs the search, summary, comparison, reasoning, and writing, much of the cognitive activity that once took place in between can be skipped. And from the individual’s perspective, that can be a perfectly rational choice.
If one person must spend thirty minutes writing an essay while another can type, “Summarize this in 400 Japanese characters at a high-school level,” and finish in thirty seconds, yet both receive the same evaluation, it is natural that more people will choose the latter.
The problem is not the use of AI itself. The question is how far the human brain will continue to develop abilities that no longer need to be performed internally.
The Externalization of Cognition
Humanity has long moved cognitive functions outside the brain. Writing externalized memory, and books externalized large stores of knowledge. Clocks externalized time management, calculators externalized arithmetic, and computers externalized information processing. Search engines reduced the need to remember where information was located. Smartphones externalized much of everyday memory, including phone numbers, schedules, maps, photographs, and contact information.
This does not mean that modern people are intellectually inferior to those of the past. On the contrary, external memory has enabled humanity to construct more complex societies, technologies, and scientific systems. Generative AI, however, has a somewhat different characteristic from earlier tools. AI does not merely store information. It organizes, interprets, compares, reasons, and expresses it. What is now beginning to be externalized is therefore not only memory, but part of the processing that occurs during thought itself.
Writing made it possible not to remember everything. Calculators made it possible not to calculate everything. Search engines made it possible not to know where information was stored. Generative AI is beginning to make it possible not to understand, organize, and articulate everything by oneself.
Those Who Use AI and Those Who Let AI Replace Them
Because of this distinction, human ability in the AI era may not simply rise or fall. It may diverge. Some people will think through and write their own material, then ask AI to evaluate it. They will ask about concepts they do not understand, have AI generate objections, check for logical inconsistencies, and then reconsider their own position.
Used in this way, AI does not replace thought but expands it. Because individuals can access volumes of knowledge and perspectives that were previously beyond their reach, both linguistic and reasoning ability may increase dramatically.
On the other hand, it is also possible to let AI write instead of writing oneself, summarize instead of reading, provide answers instead of researching, and recommend choices instead of making judgments. In such cases, the quality of the output may improve while the abilities of the person producing that output do not improve to the same degree. Future society may therefore enter a seemingly paradoxical state in which the linguistic capacity available to humanity is the greatest in history, while the linguistic ability possessed independently by each individual declines.
A World Where AI Exists from Childhood
An even more important distinction is that adults beginning to use AI today are not in the same situation as children who grow up with AI from birth. Most adults using AI today developed basic reading, writing, memory, arithmetic, and logical reasoning abilities in an environment where AI did not exist. They are adding AI as a new tool on top of already-developed cognitive abilities.
Future children may be different. If they do not understand something, AI will explain it. If they cannot write, AI will write for them. If they do not know a character, AI will complete it. If they cannot recall a word, AI will supply it. If they do not know what to say in a social situation, AI will formulate the response. That may become the normal environment from early childhood.
The human brain is highly plastic and develops the abilities required by its environment. Conversely, there is no reason for it to continue allocating the same amount of cognitive resources to abilities that are rarely required. The question, therefore, is not whether AI will make humans lazy. It is a developmental question: if a capability is always available externally, will the brain still need to build and maintain the same capability internally?
What Remains When AI Is Gone?
One way to think about this problem is to imagine a world in which AI suddenly becomes unavailable. Suppose people from different future generations are given the same task. They cannot use the internet, AI, or smartphones, and they cannot consult reference materials. They are then told, “Think about this issue for two hours and write a 3,000-character essay.”
If the education system remains essentially unchanged while cognitive substitution by AI continues to advance, later generations may perform worse at this type of unaided task. Yet if the same people are allowed to use AI, the result may be reversed. Newer generations may access more advanced information, compare multiple sources, read foreign-language literature, analyze enormous datasets, and produce extremely high-quality writing.
In other words, two changes may occur simultaneously: ability including AI may rise, while ability without AI may decline.
BCI Erases the Boundary
If Brain-Computer Interfaces, or BCIs, become practical, the issue becomes even more complex. Today, there is still a clear sequence of operations between humans and AI. We think, open a device, type, read the answer, and understand it. But if the brain and AI become directly connected through a BCI, that boundary may become blurred. The moment you try to remember a name, candidates could appear. The moment you hear a foreign language, its meaning could be supplied. When you begin composing a sentence, AI could suggest how it continues. During a conversation, an appropriate response could be generated in real time. When thinking about a complex concept, related information could be retrieved simultaneously. If such a state continued all day and throughout a lifetime, the distinction between “what I know” and “what AI knows” might cease to matter. As long as a person remained connected, they could potentially use vastly more knowledge than any human can today.
They could work across multiple languages, use advanced mathematics, and gain immediate access to specialist knowledge in medicine, law, engineering, and other fields. Functionally, such a person might possess the greatest intellectual capability in human history. Yet the moment the connection is removed, that same person might have a limited vocabulary, struggle to construct long passages, fail to remember names, be unable to perform mental arithmetic, not retain directions, and have difficulty sustaining complex reasoning using the biological brain alone.
Would Humans Really Have Lost Their Abilities?
This raises another question. If AI is always available, should its capabilities really be excluded from what we consider human ability? We do not usually define the ability of someone who wears glasses solely by their unaided vision. We do not define the mobility of someone who uses a car solely by the distance their legs can carry them. Nor do we define the true research ability of a scientist using computers by what they can calculate with paper and pencil alone.
Likewise, if AI is constantly connected, it may be reasonable to regard “human plus AI” as a single cognitive system. In that sense, human ability may not have declined at all. Rather, part of human intelligence may simply have moved from the biological brain to an external system. But this structure introduces a new category of risk.
A Single Point of Failure for Cognition
Externalized abilities depend on the external systems continuing to function. Power can fail. Networks and servers can go down. The companies operating these services can collapse. Wars can occur, and governments can restrict access. Large-scale cyberattacks can disable AI accounts. BCIs themselves can malfunction. If humans no longer retain sufficient internal capability when such failures occur, the result would not merely be a service outage. The cognitive capacity of society itself could decline at the same time.
The more AI becomes critical social infrastructure, the greater this problem becomes. Electricity and communications can already become single points of failure for modern society. If memory, knowledge, translation, calculation, writing, decision-making, and communication also become dependent on AI, then AI will become infrastructure supporting human cognition itself.
What Should Remain Inside the Human Mind?
None of this means that we should continue learning everything in the old way. There is little reason to preserve the ability to memorize dozens of phone numbers simply because earlier generations had to do so. Nor is there much need for humans to perform large multiplications rapidly in their heads. Externalizing abilities that tools can perform, and using the cognitive resources saved for more advanced tasks, is also part of the history of human progress. The real question is what should be externalized and what should remain internal.
Vocabulary, reading comprehension, logical construction, and the ability to put ideas into language are particularly important. So are the ability to critically evaluate other people’s claims, reason about causation, and detect contradictions. These abilities matter not only because they allow humans to perform tasks without AI, but because they are also required to recognize when AI is wrong.
If a person cannot construct a logical argument independently, it becomes difficult to judge whether an argument generated by AI is sound. If a person does not understand the meaning of words deeply enough, subtle errors hidden inside fluent and plausible AI-generated prose may go unnoticed. The more capable AI becomes, the more unnecessary human ability may appear to be. Yet at the same time, the minimum level of ability required to evaluate AI may become even more important.
The Role of Education Will Change
For this reason, “education without AI” may take on a different meaning in future schools. Today, AI is often prohibited in examinations primarily to prevent cheating. In the future, however, activities such as writing without AI, thinking without reference materials, reading long texts, performing calculations mentally, explaining ideas from one’s own memory, and discussing problems directly with other people may be used as training to preserve cognitive ability. It may be similar to exercising despite the existence of elevators in order to maintain physical strength.
The existence of machines capable of lifting heavy objects does not mean that humans should stop using their bodies entirely. Likewise, the fact that AI can support thought does not necessarily mean that humans no longer need to think independently. Education in the AI era may need to cultivate both the ability to use AI and the ability to think without it.
Will Human Ability Decline, or Simply Change?
It would also be inaccurate to assume that the human brain of the future will simply become weaker than it is today. When one ability becomes unnecessary, another may become more important. Rather than memorizing large numbers of facts, it may become more valuable to determine what needs to be known. Rather than producing answers, it may become more important to define the right problem. Rather than remembering one piece of information, it may become more important to understand relationships among many pieces of information. Rather than generating text, it may become more important to judge whether generated text is actually valid. The value of these abilities may increase. AI may therefore change not simply the level of human intelligence, but its composition. And how far that change proceeds will not be determined by technology alone. Education, social institutions, culture, what society chooses to define as ability, and the way we ourselves use AI will all shape the future.
The Outsourced Mind
Artificial intelligence began as a useful tool outside the human mind. But its role is gradually changing. From something that searches, to something that answers. From something that answers, to something that thinks. From something that thinks, to something that makes decisions alongside humans. And if technologies such as BCI continue to advance, the boundary between human thought and external artificial intelligence may eventually become unclear. In that world, people may be able to use far more knowledge than they can today. Language barriers may shrink, specialist knowledge may become easier to access, and the amount of information an individual can handle may increase dramatically. Human ability including AI will probably become greater than it is today. But at the same time, another question will emerge: what remains when AI is removed? Our own memories. Our own vocabulary. Our own judgment. Our own logic. And our own words. How much of these must remain inside us? Until now, humanity has used technology to extend physical ability and memory beyond the body.
From this point forward, we may begin externalizing thought itself. The result will not necessarily be a humanity that has lost its intellectual capacity. On the contrary, a humanity connected to artificial intelligence may possess forms of intelligence we can barely imagine today. But at the same time, it may also become a humanity capable of doing less with the biological brain alone than people can today. The problem of the future is not only whether artificial intelligence becomes more intelligent than humans. It is also how much of the thinking that humans no longer need to perform should still remain within us. And how we can preserve the ability to think while also gaining the enormous advantages that come from externalizing thought. The central question of the age of artificial intelligence may not be whether humans or machines are more intelligent. The more important question may be : How much of human intelligence still needs to belong to the human mind itself?