AI and Writing Predictions
literally 100% guaranteed
I expect that large language models will get better, but that it will probably be a long time before they replace “writers.” However, long before they do, they will probably change the nature of writerly employment. Here are some predictions about how that might shake out.
—Many bullshit jobs for writers, such as entry-level copywriter, or mid-level copywriter, will dry up quite quickly. This is a shame in some ways: I funded my travels while writing my first book with such jobs, and I was, and am, grateful for that. On the other hand, the disappearance of hack jobs will force writers, perhaps, into realer jobs, which is good, although some will not like it. I learned much more from waiting tables and mixing drinks than I did at my bullshit writing jobs. People with good verbal skills can generally figure out some kind of employment.
—In the short term, Buzzfeed and other sites will take advantage of advances in LLMs to pump out tons of written content. In the long term, such content will become completely devalued. Do you want a summary of the virtues of a particular wristwatch? If Gary Shteyngart hasn’t written a review, you might as well just ask an LLM yourself.
—The writers who do best will thrive because they will, essentially, become media platforms; they will cultivate community, found social movements, exert influence, collaborate with other artists, commission young talents, sell products. To an extent, this is already happening. Scott Alexander, for example, makes a handsome living partially because he is a social lodestone as well as a public intellectual.
—Some writers will be replaceable by LLM in theory but will remain successful as entities, because of their character and social position. Essentially, they will become “private label” providers of LLM products. Some of these people will be celebrities, some will just be weirdos who are good at cultivating sensational facades, provoking strategic outrage, etc.
—There is already Midjourney kitsch. Similarly, for a while, tics of LLMs will be noticeable if writers simply copy-paste their output, especially when a lot of people do this. Some readers will mind this, and some will not.
—Because of an increase in easily produced, competent pablum, formal experimentation and voice will be more valued. More writing will be like Anne Carson: stylistically mixed, collage-y, personal in the sense of reflecting the skew of a particular interiority. As well, writing that feels convincingly raw and diaristic—less polished than GPT, but clearly intelligent, like this—will see another renaissance. As in the era of Tao Lin clones, some will be really good, and a lot will be terrible. To some extent, previously reliable signifiers of rawness and personality will be mimicked; new signifiers of verisimilitude will evolve.
—Also because of this, the value of boots-on-the-ground journalism will go up, especially if it takes advantage of digital affordances by including footage, artifacts, tape, and so on. We will increasingly appreciate writing that has the gritty actuality of coming from a contingent perspective in recognizable space. And, again, new signifiers of this will need to be figured out.
—Writers who can use LLMs as assistants for research and generation while maintaining their voice will be the most successful. You will see some writers, previously less prolific, produce astoundingly good work at the pace of, say, Matt Levine. This will be extremely good news. The quantity of great writing out there could increase dramatically if we take up the challenge.
—As well, this will allow some people who are less gifted writers, but who have unique ideas, perspectives, and information, to express themselves better. Prose ability won’t be as much of a barrier to communicating, or it will be more easily surmounted. This is also extremely good news.
—Those who believe in AGI, or some other singularity, could then predict that we will see a golden era of human writing lasting some number of decades, before it is taken over as a medium by other intelligences, freeing up humans to do ???. I myself have no idea about this, really, although all of the near-term intelligence explosion narratives seem far-fetched, from my dilettantish perspective.
Sasha's 'Newsletter' is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
I've been trying to game this out in my head over the last twelve months, and I agree. The one thing I would add is a fear for a market for lemons type situation, where the growing flood of LM enabled content makes ppl less willing to engage with stuff from unknown sources, partly killing public spaces, which could make it much harder for new voices to break through and privilige ppl with preexisting networks or who live in influential geographic areas.
But I'm also optimistic for what it will do to writing. "How I can I write the compliment of what GPT does" is a fairly good rule of thumb if you want to write well.
tom wolfe was more right about the value of experiential writing than he ever could have anticipated