AI has rendered traditional writing skills obsolete. Education needs to adapt.

Many college students today know what our education system hasn’t yet acknowledged: Artificial intelligence (AI) has rendered traditional writing skills obsolete. It’s a change I see firsthand as a teacher of undergraduate writing classes at UCLA. 

AI can perform most writing tasks 

Over the past several years, educators have produced a seemingly endless string of task force reports, policy statements, op-eds, and other forms of hand-wringing about the role of AI in writing. But few have recognized what is abundantly obvious to pretty much everyone under the age of 25—that today’s young people will inhabit a future where the vast majority of writing will be produced using AI. 

Of course, there are exceptions. Novelists, reporters, screenwriters, columnists, attorneys, academics, and others in writing-intensive professions will still need to be skilled writers. For them, AI will complement and amplify—but not fully replace—human writing expertise.  

But the vast majority of students will not choose to lead lives centered on writing. They will write much more sporadically and for more utilitarian purposes. They will write progress reports, meeting summaries, user manuals, business plans, newsletters, letters to city councils, cover letters, and instructions to a neighbor on how to feed the cat when they are away. With only a modest amount of human supervision, AI can produce these documents quickly and effectively. 

Historical context 

And what of the argument that the process of writing is far more than putting words on a page, and that the mental discipline needed for non-AI-assisted writing has a long list of benefits? 

It’s a valid argument. But it’s not going to win the day when weighed against the efficiencies made possible by AI. And it echoes analogous arguments made in the past with respect to other abilities that we largely no longer cultivate. 

Consider good handwriting. Penmanship, wrote H.C. Spencer in a book published in 1866: “Puts into full requisition all the higher powers of the mind. Under this impulse the faculty of perception is called into vigorous exercise, memory is made more tenacious of its treasures, judgment is at work in determining relations, proportions, and distance; while taste, ever alive to the forms of beauty, whether in nature or in art, is busy with all those nicer discriminations of shade, color, outline, and finish. . . .”

These are very substantial benefits, and it is true they are lost when we write using a keyboard or keypad. But on balance, far more is gained, which is why the past half century has seen nearly a complete transformation from pen strokes to keystrokes. 

So it will be with AI and writing. We are currently in a liminal space where it is still possible to imagine that the hard-earned rewards accessed through the labor of writing will be sufficient incentive to keep AI at bay. But that is an illusion that is ebbing with each new advance in AI—and those advances seem to be occurring every few weeks. 

The democratization of good writing 

While people who have spent years cultivating their writing skills might bemoan the arrival of AI-assisted writing, there is also a much more optimistic way to view these changes. Until now, the ability to write well was inherently elitist. People fortunate enough to have the time and financial capacity to pursue higher education were better positioned to produce excellent writing. 

In the blink of an eye, that has changed. AI is enabling anyone, regardless of education level, to create well-written documents and do so in pretty much any language. It’s a profound change, and it’s also profoundly threatening to institutions such as colleges, which are collecting tuition used in part to teach traditional long-form writing skills to students who will rarely use them after they graduate. 

But to lament that good writing will no longer be the exclusive province of elites is, well, elitist. A far better response is to celebrate the arrival of a technology with the promise of truly democratizing written communication. 

What about hallucinations? 

As documented by a growing list of news stories, generative AI systems sometimes output false information. In fact, there is some evidence that hallucinations are an inherent aspect of large language model (LLM) outputs. But hallucinations are an easily solvable problem. 

First, anyone using AI to write can examine the resulting text and remove any assertions of fact that are not easily and reliably verifiable. The people who have ended up in hot water for inadvertently turning in documents containing AI hallucinations have failed to successfully do this checking. 

Second, with the passage of time, I expect that AI hallucinations will come to be viewed a bit like slow internet service in previous decades. Just as technology advances have rendered slow internet largely a thing of the past (with access from airplanes and off-grid locations still notable exceptions, though those exceptions are vanishing thanks to Starlink), further advances in AI will mitigate hallucinations. 

Even if future LLMs can’t be entirely prevented from outputting some degree of false information, it will be easy to add back-end software to check the accuracy of all publicly verifiable assertions of fact. And we won’t need to wait decades before we see AI systems that are mostly hallucination-free. They should be widely available within a few years. 

Adapting our education system 

Today’s writing curricula are overseen by people who came of age in the pre-AI era. There is a natural bias in wanting to teach students to thrive in the world that most educators know best—where the ability to produce a well-written document without computer assistance opened the door to otherwise inaccessible professional opportunities. 

But that world is already receding into the past. It is far better to redesign writing education to make full, unapologetic use of the power of AI, because that is how nearly everyone will write. This means replacing classroom policies that prohibit the use of AI in writing assignments with policies promoting its responsible use. It also means teaching students that AI-generated plagiarism is still plagiarism, and that AI sometimes produces hallucinations.  

And it means helping students become proficient at using AI as a force multiplier to improve the depth, versatility, and speed of their writing. Among other things, this requires teaching students how to evaluate writing for flow, organization, clarity, and logical and stylistic coherence. This will enable them to examine and further polish (often using AI) the AI-produced text they will use in nearly all of their writing. And contrary to what might be expected, a person doesn’t need to be a good writer to be good at evaluating writing—just as someone who can’t play piano can nonetheless distinguish between skilled and unskilled piano players. 

For current college and high school students, the capabilities of AI are no more surprising than internet access was to the people who were young two decades ago. Instructing them not to use AI to help them write makes as little sense as telling students in 2005 not to use the internet for research. Today’s young people know that when it comes to writing, the technology landscape has undergone a tectonic shift, and they have already found their new footing. Those of us involved in teaching them need to do the same.