A guest post published on Jane Friedman's site this week argues that generative AI is quietly flattening the way writers sound, and it offers memoirists a set of exercises meant to protect what makes their prose distinct. The advice lands alongside separate reporting from The Verge on AI's environmental cost and on the industry's long pattern of asking for regulation it has not fully gotten.
Guarding a memoir voice
Writer and editor Lisa Cooper Ellison, in a guest post on Jane Friedman's site, wrote that the risk for memoirists has shifted. Where writers once worried about imitating a literary hero too closely, Ellison wrote that AI's widespread use is now reshaping everyone's prose toward sameness, whether or not a writer directly uses the tools. Ellison defined voice as a mixture of natural style and sentence-level choices, including tone, rhythm, syntax and word choice, that distinguish one writer from another. She wrote that voice is mostly built through sustained work rather than innate talent. To help writers isolate what is theirs, Ellison described several exercises. One involves rewriting a scene in the style of a favorite author, then removing every phrase or rhythmic pattern that belongs to that author, leaving behind what is left of the writer's own style. Another asks writers to name everyday household objects in their own words, arguing that instinctive word choices reveal a writer's worldview and should be tracked in a personal vocabulary list. Ellison also encouraged writers to experiment directly with AI, using two published tools that let a person rewrite an article in an AI-influenced style and then compare it against a human-written version. She linked to an August 25, 2026 article responding to Substack's collaboration with the detection tool Pangram, which she said lets writers see how Pangram distinguishes AI-generated text from human writing. Ellison wrote that reading both versions aloud helps a writer notice what sounds flat and what sounds true, and that the goal is not to change a writer's style but to sharpen it by increasing specificity. Ellison also noted that some writers already use AI to generate text outright, sometimes out of convenience and sometimes because they believe it writes better than they do, a trend she said is prompting some editors to rewrite contracts.
The cost behind the tools
Separately, The Verge reported that the AI boom is producing far more electronic waste than previously estimated. A report from the nonprofit Basel Action Network, published Wednesday, projected that AI-related e-waste could total enough to fill 23 million shipping containers by 2050, or roughly enough 40-foot containers to circle the globe six times if lined up end to end. The Verge reported that BAN's estimate is higher than earlier studies because it counts the full infrastructure supporting data center servers: power supply and distribution, cooling systems, backup power and networking equipment, along with a category BAN calls "AI Waste Contagion." That category covers telecommunications infrastructure and personal devices that become obsolete sooner because of AI advances, according to The Verge's account of the report. BAN projects global e-waste will triple to as much as 211 million metric tons a year by 2050, attributing 15 to 20 percent of that to AI, The Verge reported. The organization's founder, Jim Puckett, said in a press release cited by The Verge that AI depends on enormous amounts of specialized hardware and that today's AI buildout could become a larger toxic waste crisis than the one already underway. The Verge noted that a 2024 study had estimated AI e-waste at 1.2 million to 5 million tons by 2030, and a February study had put AI server waste at 131,000 to 225,000 tons annually by the end of the decade, figures BAN's report suggests undercounted the problem substantially.
A familiar pattern from executives
A second Verge article traced years of public statements from AI industry leaders calling for regulation, framing recent comments from Sam Altman, Dario Amodei, Demis Hassabis, Satya Nadella and Elon Musk as part of a pattern rather than a new development. The Verge documented Musk warning about AI's dangers as far back as 2014, Sundar Pichai writing in the Financial Times that AI needs to be regulated, and Sam Altman telling Congress that regulatory intervention would be critical, alongside a 2023 open letter warning that mitigating AI's extinction risk should rank alongside pandemics and nuclear war.
The Verge reported that these calls have produced few binding results, noting that California Gov. Gavin Newsom vetoed the state's SB 1047 AI safety bill despite Musk's public support for it.
None of the sources connect these developments directly. But together they sketch a widening gap between how AI is marketed to writers as a craft tool, what its infrastructure costs, and how little enforceable oversight has followed years of industry statements about risk. For memoirists, Ellison's advice offers a starting point: the September 25, 2026 Substack-Pangram comparison tool and the article-rewriting exercise she links to remain available for anyone who wants to test, in their own ear, where their prose ends and a machine's begins.
Sources
Every quotation and figure in this article is checked against the source reporting linked above before publication. How we report