Writer’s Diary #29 – Large Language Models as Writing Assistants and First Readers?

Years ago, when I first discovered Wendell Berry, I remember reading a 1988 Harper’s article by Wendell Berry titled “Why I Am Not Going to Buy a Computer.

When the article was published, I was five years old. That was the same year my mother, a professor, brought home an Atari ST 1040 from work. I cherished that computer, using it for the better part of the following decade. In 1997, my parents purchased a Macintosh Performa 6400. On both machines, I learned to type, program, hack, tinker, and write.

However, I‘m unsure whether my writing process has been the best. Two years ago, Peter Elbow’s book on freewriting was a revelation.

Anyway, in that 1988 article, Berry writes:

As a farmer, I do almost all of my work with horses. As a writer, I work with a pencil or a pen and a piece of paper.

He continues:

My wife types my work on a Royal standard typewriter bought new in 19~6, and as good now as it was then. As she types, she sees things that are wrong, and marks them with small checks in the margins. She is my best critic because she is the one most familiar with my habitual errors and weaknesses. She also understands, sometimes better than I do, what ought to be said. We have, I think, a literary cottage industry that works well and pleasantly. I do not see anything wrong with.it.

He goes on, noting that a friend suggested he buy a computer, and change his workflow. As of 2019, Berry hasn’t purchased a computer.

His reluctance to buy a computer, and his support of his wife as a first reader, sparks professional envy.

One the one hand, Berry is a gifted and prolific writer, and has crafted a thoughtful, respected body of work around agrarian themes without the aid of a computer. Yet, of course, this success can be attributed, in no small part, to the work of his wife, Tanya[^2^]. I suspect, in the way, a good first reader makes a writer strong.

While I hesitate to form opinions about a couple who have lived together for sixty years, creating a thriving literary cottage industry, there is a feminist critique that many acclaimed male writers have relied heavily on uncredited female secretaries, assistants, wives, and family members.

I’m thinking of Eleanor Marx, who edited many of Karl Marx’s manuscripts. Simone de Beauvoir, whose un-attributed efforts shaped Jean-Paul Sartre’s work. Zelda Fitzgerald, F. Scott Fitzgerald’s wife, was the source of many of his ideas, and even, it seems, of un-attributed scenes, in other words plagiarism.

While such unacknowledged collaboration raises questions about attribution and recognition, it also raises questions about writing as a process, involving first readers and, frankly, the work of a typist.

Since the 1990s, society has replaced secretarial staff with computers. Many writers, including myself, have effectively become our own first readers, typists, and copy-editors. My method of writing hasn’t changed much since childhood: I draft, rewrite, revise, correct spelling and grammar, and revise, revise, and revise. I might go through a few dozen drafts, fixing and adding spelling errors as I go.

For my first book, I employed a copy editor and a substantive editor. Friends and family members read my earliest drafts, and offered suggestions.

As I attempt to write my new book, juggle my responsibilities as a professor and a father, I find not only that there‘s too much going on and the writing isn’t happening, but that I have a much smaller pool of people who can help out.

Part of this issue, as Cal Newport‘s work suggests, is that administrative and secretarial roles that used to exist in offices and white-collar jobs were replaced by technology in the 1990s. Funds that previously paid for support staff now end up in IT departments, ultimately benefiting Big Tech.

Berry’s critique of this transition is prescient, as he starts his argument by expressing his disdain for dependency on oil companies.

Part of the issues, is I’m older, and everyone is busier.

But, the challenge of writing and finishing remains. Finishing requires a combination of hard work, creativity, and perseverance. Some elements of the process are iterative. Some of it requires a good first reader.

The work of a first reader, editor, and copy editor is hard to fill. In my experience, I have no one to fill this role and have done this work myself.

But, over the last six months, I’ve begun to see artificial intelligence and large language models from OpenAI and Anthropic as competent copy editors and first readers, when prompted carefully.

While AI will drastically disrupt many industries and jobs wreaking havoc on society, for my writing, I can see a roll for it.

Not to generate words, ideas, scenes, narratives, or text, which is plagiarism. but, to copy editor.

Perhaps some writers will use AI to cross a boundary of what is moral. Perhaps what is seen as moral will change? But, in the mean time, I’ve found a role for AI to clean up my typing, as a first reader, and as a copy editor.


Writer’s Diary #28 – Anthropic’s Claude-2-100K

Yesterday I listened to the latest Hard Fork episode, which features an interview with Dario Amodei, one of a group of OpenAI employees who left to start Anthropic, OpenAI’s competitor. The interview is quite interesting for someone like me who doesn’t know this area well. But it made me want to check out Anthropic’s new [Claude 2 AI] (https://claude.ai/login).

Claude 2 is not available from Anthropic in Canada, but I got access through Quora’s Poe app.

I’m an anthropologist, a writer, and I’m interested in these big language models and teaching and writing. On the one hand, there is a lot of disruption. I’m on sabbatical, but I don’t know what that means for student essays. But I’m fascinated by AI as an editorial assistant. Many writers rely on editors, but for a long time those editors were unpaid, often female relatives. AI provides me with a free editor and critical reader, for the first time not a family member or friend.

Working with Claude 2-100K yesterday, I came to a couple of realizations.

First, 100K makes Claude much more useful and less stupid than ChatGPT. A few weeks ago I had the idea of using ChatGPT 4 to rearrange a book of fragments. Much of my work involved condensing fragments into summaries for ChatGPT 4’s short memory, or at least the short memory I have access too. With Claude 2-100k, I could give the AI the whole draft. We discussed several possible outlines. With the ChatGPT 4 model I have access to, it always felt like I was at the limit of the AI memory and capacity. With Claude-2-100k, the answer was much slower to come, but was feedback on the entire damn manuscript. My first book took 8 months. This took under a minute.

From there, I had a long, free-flowing conversation with Claude 2 about my book, its ideas and theories, and then we turned to writing and the differences between how Claude 2 writes and how I work. Frankly, Claude 2 just felt much more thoughtful and interesting to talk to than ChatGPT 4.

ChatGPT feels like it has less memory, is dumber, and frankly more prone to nonsense. It seems smart at first, but on further review it is often very formulaic.

Maybe Claude 2 will seem that way to me, but yesterday it seemed like a much more interesting interlocutor than I’ve had on this book, ever. What it says about me that I say this about a large neural net running on servers, converting text into tokens and then running complex vectors and matrices, leaves me very confused. What does it mean to be human, I wonder?…