ChatGPT and the death of ideas
ChatGPT can’t write. Despite the hype around generative AI, it has the choppy style of a high schooler, the bland voice of an editorial board, and the intellectual depth of a cactus. It’s not a real threat to human writers — at least not to the average ones. But growing reliance on generative AI as a shortcut and substitute for writing could transform writing from a common marker of professional competence to a niche skill. By diminishing the public’s writing ability, generative AI could threaten our ability to express and develop original ideas.
ChatGPT’s primary use case so far is, predictably, in academic settings. According to Google Trends, interest in ChatGPT first surged in late 2022 before mysteriously plummeting in late December; it eventually peaked in April and has steadily declined since. While large language models (LLMs) like ChatGPT are weak writers, they can proficiently (but not eloquently) organize and articulate information, making them ideal for many writing assignments. ChatGPT is a convenient shortcut for time-consuming, formulaic writing tasks and it can plausibly serve a similar function in professional settings.
But greater efficiency is no reason to celebrate generative AI, which could degrade the public’s writing skills. Writing, like any other skill, requires frequent practice for proficiency and eventual expertise. Using LLMs as a shortcut for writing tasks means people will have fewer opportunities to refine their writing skills — if they continue to write at all. With less exercise, their writing abilities will atrophy and increase their dependence on LLMs to produce adequate content. LLMs, now shortcuts, will develop into necessities.
The narrowing of writing ability would be a sharp departure from recent historical trends. Today, the global adult literacy rate stands at 87%, a dramatic change from just centuries and even decades ago, when literacy was restricted to the elite in many places. The rapid expansion of literacy has enabled both the widespread distribution and expression of ideas, allowing the public to meaningfully engage in political and cultural discourse. By reducing the public’s writing ability, LLMs threaten to reverse that progress and narrow societal discourse.
Even with heavy dependence on LLMs, most people would probably retain some semblance of basic, formulaic writing ability. But good writing and good thinking aren’t formulaic. At its core, writing is an act of translation, and translation is a difficult, messy art. We formulate ill-defined thoughts, then decide how to best express them. The goal of translation isn’t perfection, because perfection is impossible. Instead, translation aims to maximize expression of the original idea or content, capturing its nuances as much as possible while maintaining clarity. Reliance on LLMs simply diminishes our intellectual and expressive potential.
Writing is a powerful intellectual exercise because it enables us to not only express ideas but also formulate them. At its best, it’s an intellectually boundless process that forces us to grapple with our own and others’ ideas. Creativity and originality don’t exist in a vacuum; they emerge from the deep contemplation that any good writing process requires. Consequently, a decline in writing ability diminishes our ability to both express ideas and create new ones.
Of course, people won’t stop thinking even if they write less; ideas and stories are essential to human nature. We’ll still have original ideas and still seek to share them, but we’ll increasingly lack the ability to do so. Instead of enduring an arduous writing process, people will opt for the ease of LLMs, perhaps touching up that content with a few edits. Although that might seem a convenient tradeoff, we lose control and autonomy over our ideas by outsourcing expression. The idea expressed may have originated in a human mind, but it no longer belongs to that individual. To the extent that the idea still exists, it exists meaningfully only within that person’s mind — where the idea is least valuable. Ideas shift from part of a shared societal discourse to part of an internal monologue.
That shift will probably benefit the few skilled writers who remain; writing ability will become more scarce and thus more valuable. But changes in an attribute’s value for an individual don’t affect its value to society. Though thoughtful, skilled writers may find more lucrative opportunities than they do today, that can never compensate for the societal losses involved. We’ll miss out on brilliant ideas because people never learned to express them properly. We’ll lose valuable contributions to political and cultural discourse that could’ve changed the trajectory of entire societies. And worst of all, the collapse of writing skills will undermine freedom and democracy.
Free societies revolve around the open, accessible exchange of ideas. Essentially, they depend on the risky and noble proposition that individuals will cherish and exercise their rights. But when we outsource our speech to opaque and unaccountable LLMs, we essentially grant the corporations that control them autonomy over our speech and ideas. Generative AI could thus allow corporations to directly shape and filter political discourse on an unprecedented scale, hindering genuine free speech by inserting their agendas and biases into “our” speech and ideas. That would shatter the ideals of equality implicit in free speech and radically alter the scope of societal discourse. True freedom and democracy are impossible in a society that can’t independently form or articulate ideas.
The most important step we can take to preserve a vigorous exchange of ideas isn’t regulating or restricting AI development. Given the rapid pace of development and open-source nature of some models, that approach risks yielding AI leadership to bad actors. And since AI is a wide-ranging field featuring models with vastly different capabilities, vague talk of “regulating AI” is mostly meaningless anyway. Instead of pursuing rash policies and mistaking vapid political catchphrases for solutions, we need to rethink our institutions for the AI age.
Our education system is an obvious starting point. Virtually every aspect of modern American education, from essays to college admissions, can be gamed to varying degrees. A system that can be gamed is a system that incentivizes shortcuts, but some of those shortcuts pose an existential risk to an intellectually vibrant society. Written assignments should push students to interact with ideas instead of organizing and restating information. In short, we should shift away from rubric and grade-centric models of assessing writing toward qualitative evaluations.
To cope with a surge of AI-generated content, social media platforms should significantly expand verification programs. Verification should be free and accessible, aimed simply at ensuring that users are real people. There’s one crucial drawback to this approach: it marginalizes anonymous users, many of whom might be hesitant or unable to share their ideas without anonymity. Platforms should thus continue to allow unverified users while algorithmically prioritizing content from verified users. It’s an imperfect compromise, but it at least supports human ideas while striving to ensure privacy and free expression.
Finally, lawmakers around the world should deny intellectual property protections to AI-generated content. The US Copyright Office has taken mixed steps in this area, rejecting copyrights for some AI-generated images while indicating its openness to issuing copyrights for some AI-generated content. The Copyright Office and other relevant agencies should instead deem AI-generated content categorically ineligible for IP protections. And given the increasingly global nature of consumer markets, agencies in other countries should take similar measures. To preserve creativity, it’s vital to ensure that AI-generated content doesn’t have a market edge over human content.
As important as it is to adapt these and other institutions to AI, the preservation of ideas ultimately depends on our own choices. We should remember that while there’s value in brilliance, there’s also value in mediocrity. Mediocrity may be uncomfortable, but it’s a necessary stage in the development of any skill and one we should embrace. It doesn’t have to be our ultimate destination, which is what it will become if we opt for the ease of LLMs. By valuing our autonomy and creativity over convenience, we can choose a better path.
