Amazon’s new AI writing culture is selling short our powers of productive thinking

How can I tell what I think till I see what I say?’” wrote EM Forster in Aspects of the Novel. What he was getting at was that the act of writing can uncover thoughts and beliefs that may not be fully formed or understood until you try putting them into words.

As a university teacher, I used to put it more bluntly: there are two kinds of writer; those who know what they think and write it; and those who find out what they think by trying to write it. In a long career surrounded by academics and journalists, I’ve only known two people who fell into the first category; everyone else joined me in the second.

What brings this to mind is news of an apparent cultural shift inside the tech giant Amazon. For a long time, writing was sacrosanct in the corporation, which is an odd thing to say about such an organisation. And it came from the very top – its founder, Jeff Bezos.

“The company began each big meeting with six-page narrative describing the product or feature,” writes Kristi Coulter, the author of the piece, “typically written by the project lead, read in silence before anyone spoke. The writing’s purpose was to crystalise thinking and anticipate every scenario.” PowerPoint, the conventional emotional crutch of corporate discourse, was verboten in Amazonia.

“My perfect meeting,” Jeff Bezos told the Lex Fridman Podcast, “starts with a crisp document. The document should be written with such clarity that it’s like angels singing from on high. I like a crisp document and a messy meeting. The meeting is about asking questions that nobody knows the answer to and trying to wander your way to a solution. When that happens, it makes all the other meetings worthwhile. It feels good. It has a beauty to it, an aesthetic beauty, and you get real breakthroughs in meetings like that.”

You do, and Amazon did. But then two things happened. One: Bezos stepped down as chief executive to spend more time with his pneumatic girlfriend (now his second wife) and his spaceship venture. Two: AI arrived and, according to Coulter, Amazon’s leadership is now encouraging employees to let AI do the writing for them. Its internal marketing for Cedric, the company chatbot, says that it produces “six-page narratives in seconds”.

At this point, an old engineering adage comes to mind: if it ain’t broke, don’t fix it. Writing is a cognitive process through which clarity emerges rather than merely being recorded. Bezos, not exactly a sentimental humanist, built an entire corporate culture on that insight. His company invented an innovative way of ensuring that meetings would be creative rather than procedural and that decisions emerging from them would engender a sense of ownership in the participants.

In corporate and bureaucratic life, this is pure gold dust. And yet Amazon has apparently decided to turn its back on it. Coulter spoke to 15 Amazonians, a mixture of Bezos-era veterans and newcomers, to see what they made of the change. She found most of them fretful.

“Writing is thinking,” said one veteran. “That was the whole point of Amazon’s writing culture. I can’t tell you how many times I changed my mind when writing a narrative. And even when I didn’t, my arguments were more precise for having written them down. Now we have chatbots writing six-pagers to be summarised by other chatbots.”

In a strange way, the anxieties of these employees are also reflected in university humanities classrooms. What conscientious teachers are mourning isn’t necessarily the essay as a genre but the absence of the cognitive effort that writing imposes. When a student uses an AI to create an essay, they are avoiding the very experience that the assignment is designed to create. And in that sense, they are cheating themselves as well as the institution that provides the credential they crave.

All of which is just a way of saying that AIs – and large language models (LLMs) in particular – are enabling and disabling. On the one hand, they augment human capabilities and give you the opportunity to access a sizable chunk of the written knowledge that humans have accumulated over centuries. They are, as Berkeley professor Alison Gopnik put it, “cultural technologies” – such as books and libraries.

On the other hand, undue reliance on them may produce longer-term cognitive decay. Just as GPS has rendered us incapable of navigating with traditional maps and incessant typing leads to handwriting deterioration, outsourcing writing to LLMs will reduce the productive effort of trying to say precisely what you mean, and discovering through failure that you didn’t yet know what you meant.

Both Amazon’s six-page-memo culture and the humanities essay were, in their different ways, technologies for forcing that encounter. The ultimate irony is that what makes AI so appealing – that it spares you the struggle – is precisely what makes it cognitively worthless as a thinking tool.

In Forsterian terms, it’s the writer that knows what it thinks and writes it.

Informed opinion
Populism Is an Information Problem, But Not in the Way That You Think is an original take by Andrew Curry on the UK’s problems.

The price of knowledge
Interesting insights from Ted Gioia on his blog can be found in How Jeffrey Epstein Became a Public Intellectual.

Varsity blues 
The Misuses of the University is a sobering essay by François Furstenberg on what’s gone wrong with America’s elite academic institutions.

Photograph by Andreas Rentz/amfAR/Getty Images