Hallo friends, and welcome, new readers!

I hope you’re managing to navigate safely whatever weather challenges may have come your way this year.

In today’s newsletter:

• The “I only use AI for…” fallacy

• ICYMI: What “America” meant before 1776

• New-ish podcast series to check out

Feel free to forward this to a friend, student, or colleague who might enjoy it!

The “I only use AI for…” fallacy

Last week Humans: A Monstrous History got a mention in the Washington Post. Science journalist Tom Zeller Jr. opened his opinion article, “Don’t quit this whole-brain workout” (PDF), by describing the conversation with which I open the “Machines” chapter of Humans (that chapter segment was also excerpted on Lit Hub).

Zeller asks: “What happens when we begin to outsource one of the brain’s most cognitively integrative activities [i.e. writing]?” Parts of his essay align with my own thinking, but his suggestion that genAI is worth using for writing-adjacent had me shaking my head.

Occasionally people who make things with words say: “I don’t use AI for writing, I only use it to….” For Zeller, “LLMs [are] an incredibly useful tool for interrogating ideas, compiling evidence, testing assumptions, identifying holes in my logic and even proofreading my work.”

Such attempts to assure readers that the writer did their own homework ignore the fact that the ethical problems don’t end with whether what they send into the world are words they thought up, chose or typed.

The problems with writing-adjacent uses of “genAI” fall into three categories:

• Risks to the world (the planet, democracy, society)

• Risks to the writer (AI dependency and AI psychosis; absorbing magnified biases and misinformation; loss of confidence from loss of skills and/or lack of practice doing something)

• Risks to the piece of writing (errors of fact and framing; magnifying biases and harms in the source data; replicating the biases of the designers; insincere or revolting prose style)

If you’re on Bluesky you may have noticed that I call genAI products things like “ecocide-plagiarism-IP-theft-psychosis-harassment-authoritarian-surveillance-radioactive waste machines.” To keep this newsletter from turning into the 7,000-word frenzy of bullet-points and idea stubs sitting on my laptop — material I should craft into op-eds and essay pitches but is infuriating to face — I’ll focus on one idea:

There are numerous safe alternatives to slop products, both for writing and for writing-adjacent activities that some define as falling outside whatever they call “writing”. To use an LLM product instead of an alternative is to add to harms to the world, to one’s self, and to the body of human knowledge.

Alternatives include:

• Putting your work aside for a while so that you come back to it almost wearing a new brain

• Asking a friend or colleague to try to poke holes at your draft writing — and you can poke holes back at their own drafts in exchange

• Heading to your local public or university library (many have public membership options), or speaking to or emailing a subject librarian or curator

• Borrowing library books on writing, research, organizing your work/life… 

But what about accessibility aids? I am, of course, entirely in support of aids to accessibility (more on that in a minute). And the homogenizing label “AI” is only “a marketing term,” as computer scientist Dr Timnit Gebru puts it. “AI” encompasses varied activities and products from a bunch of different subdisciplines. Machine learning isn’t the same thing as generative AI products based on large language models (LLMs).

I don’t accept some AI boosters’ framing of the accessibility problem: as a zero-sum game between people who have accessibility needs and thus supposedly need whatever LLM-based products Silicon Valley wants to barf at us with no oversight, and people who (currently) don’t have accessibility needs and thus supposedly have the luxury of being AI-haters and refuseniks.

Instead, imagine if the billions that venture capitalists have sunk into genAI products¹ had been spent on non-ecocidal accessibility measures, assistive technologies, and education. Imagine if the millions billionaires regularly spend on lobbying against taxes on billionaires and against independent tech oversight were used for the public service debts their businesses incur, from the schools and universities that educated their employees to the health of people living near polluting power plants and data centres.

Just imagine if that money were spent retrofitting older buildings to make them more accessible, or on affordable computer hardware, software, and office furniture for accessibility needs. Or on air purification systems in public buildings, including classrooms (remember COVID, anyone? People are still catching this and other respiratory illnesses). Or was available to pay more (and pay better salaries to) teaching assistants, writing coaches, proofreaders, copy editors, teachers, professors, librarians, curators, developmental editors, and ghost writers. Or for schools, public transit, and medical services in Global Majority countries that have long paid the price of climate change and colonial extraction.

Organizing, editing, and brainstorming are part of writing-while-human. They can be hard, even scary, but that doesn’t mean they could or should be done by machines. If you need a burst of inspiration or a jolt to throw you out of a rut, ask a question out loud, pick a word at random in a physical dictionary, and use that as a prompt to magick up an answer. Or ask a person.

But for anyone trying to earn a living by writing, the world that tech has been allowed to build is a terrible one. (I’m glaring at government and institutional enablers.) Media outlets’ revenue streams are dwindling into trickles. Their subscriber bases are shrinking (I wonder how many potential readers are paying for ChatGPT subscriptions that could have covered magazine and newspaper subscriptions instead). And as more people click on AI summaries rather than on websites, news outlets and other businesses receive less ad revenue even though their content is stolen and warped into those very summaries. All of these mean job cuts for writers.

Here again is a place where genAI boosters push a false binary between people who supposedly “need” genAI in order to churn out words fast enough to earn a living wage, and people who refuse to use it.

Tech executives promote products that dismantle education and obstruct direct access to high-quality sources of information. (Plug AI into your university, and you can fire faculty! HUH.) I don’t know whether these executives are doing this out of greed, for sinister reasons, or because they are functionally illiterate, galactically ignorant, and don’t realize how many of their products add nothing to the universe.

Whatever the cause, the effect is the same: the planet gets hotter; people’s brains become worse (as if that will help them come up with climate-crisis solutions!); human knowledge is poisoned with the verbal equivalent of cat vomit (would you pick through that for the “good” bits?). And all this is built on intellectual property theft, exploited and traumatized data workers, and surveillance systems that enable the worldwide descent into fascism.

It’s as if someone said, hey, instead of fixing problems, let’s add new ones on top!

You can be sure that once tech companies have made more people, businesses, governments, and educational institutions organizationally and psychologically dependent on their products prices will enter the stratosphere, and users will be royally [insert your profanity of choice]. (Yes, this is infuriated Surekha trying not to type a swear word. Think of The Muppet Show’s Animal crossed with Miss Piggy.)

Given the existential risks of widespread, unfettered AI use, it’s vital that we remember, learn, use, promote, and fund these alternatives instead of lining the pockets and equipping the bunkers of tech billionaires. If a particular energy-guzzling, exploitation-derived, or private-data-consuming AI product isn’t doing this in service of climate change abatement, pandemic modelling, or other socially beneficial activity that other tools or human experts cannot perform, then it is redundant and harmful.

Don’t let the tech industry’s most nefarious boosters impoverish how you understand who and what humans are; what they are capable of (time, energy, resources, and bandwidth-permitting); and the sorts of lives they deserve to live. There are other solutions to creative, literary, and logistical conundrums: books, education, communities, professional experts, infrastructure, and a more equitable distribution of wealth.

ICYMI

Looking ahead to the July 4 weekend in the US, if you missed my LA Times essay on the first map on which “America” appeared, you’re in for several surprises: check out “What ‘America’ meant before 1776, and who ‘Americans’ are today” (PDF)

New-ish podcasts to check out

American Medieval, hosted by Professor Matthew Gabriele. Funny and thought-provoking is the episode in which Professor Sonja Drimmer talks about medieval printing and contemporary AI-generated images: how “the political and economic project of so-called AI”, namely “upward wealth transfer,” poses a “threat to civil society… dismantling any kind of grounds for consensus about truth that we have.”

The Useful Knowledge podcast from the American Philosophical Society: a Philadelphia-based learned society founded in 1743. For this weekend, the interview with Professor Joyce Chaplin on her latest book, Ben Franklin’s Stove: An Unintended American Revolution, would be an apt one to try.

I recently recorded episodes for both these shows, so watch/listen out for them in the coming weeks!