A profile of Emily M. Bender, a computational linguist at the University of Washington who co-wrote a paper in 2020 to illustrate what LLMs can and cannot do
Nobody likes an I-told-you-so. But before Microsoft's Bing started cranking out creepy love letters; before Meta's Galactica spewed racist rants … Mastodon: @ben@werd.social , @emilymbender@dair … , @sundress@eigenmagic.net , @dingemansemark@scholar.social , @paul@social.paulkedrosky.com , and @KevinMarks@xoxo.zone Tweets: @_zeets , @_zeets , @mariehermanova , @lucianabenotti , @elizab0t , @musa_algharbi , @davetroy , @pt , and @pt Mastodon: @ben@werd.social : “One fired Google employee told me succeeding in tech depends on “keeping your mouth shut to everything that's disturbing. ” Otherwise, you're a problem. ” … Emily M. Bender / @emilymbender@dair-community.social : Thank you, @lizweil for digging into the topic of “AI” in our current moment and highlighting how humanist (and linguistic) perspectives can bring as society grapples with important discussions. — https://nymag.com/... … Alice Boxhall / @sundress@eigenmagic.net : This is an extremely thorough piece on the *actual* risks of the current generation of “AI”. The final paragraph in particular doesn't pull any punches. — https://nymag.com/... Mark Dingemanse / @dingemansemark@scholar.social : Very nice profile of @emilymbender in this thought-provoking and hard-hitting NY Mag piece — on text generators posing as humans and humans calling themselves parrots https://nymag.com/... #chatgpt #stochasticparrots #nlu Paul Kedrosky / @paul@social.paulkedrosky.com : Good piece on AI ethics, LLMs, humanity, language, and the deterministic boys (mostly) with toys driving things with little regard for consequences. #xp — You Are Not a Parrot — https://nymag.com/... Kevin Marks / @KevinMarks@xoxo.zone : Bender defined... dehumanization as “the cognitive state of failing to perceive another human as fully human... and the experience of being subjected to those acts that express a lack of perception of one's humanity.” … Tweets: Zito / @_zeets : It's incredible how so much tech is built and maintained on cheap and traumatic labor https://twitter.com/... Zito / @_zeets : This reminds me of an old essay on how technologists constantly compare the human brain to whatever is the dominant technology at the time. Once the brain was supposed to be like a clock, then a train, then a computer etc. https://twitter.com/... Marie Hermanova / @mariehermanova : We've learned to make “machines that can mindlessly generate text,” Bender told me when we met this winter. “But we haven't learned how to stop imagining the mind behind it.” (The best, most nuanced critique of the so-called AI boom that I've read so far!) https://nymag.com/... Luciana Benotti / @lucianabenotti : “LLMs are tools made by specific people [..] It's not about all of us. It's about some of us becoming a superspecies. This is the darkness that awaits when we lose a firm boundary around the idea that humans, all of us, are equally worthy” https://nymag.com/... Elisa Lindinger / @elizab0t : Alright so the LLM topic is kind of exhausted by now but did you know about the trend* of saying SALAMI instead of AI? https://nymag.com/... *and yes I know that it's not a trend, but please let's make it one https://twitter.com/... Musa al-Gharbi / @musa_algharbi : “We go around assuming ours is a world in which speakers mean to say what they say and expect to live with the implications of their words... We've learned to make machines that can mindlessly generate text, but we haven't learned how to stop imagining the mind behind it.” https://twitter.com/... Dave Troy / @davetroy : This story is well worth your time. “From here on out, the safe use of artificial intelligence requires demystifying the human condition.” https://nymag.com/... Parker / @pt : I see this sentiment a lot as a way of elevating the perspectives of critics over builders. It misunderstands the situation. Builders get excited about making ideas real & then invest time/money, they don't invest money & then decide to get excited. https://nymag.com/... https://twitter.com/... Parker / @pt : Are you OK, @NYMag? https://twitter.com/...
Context & Ripple Effects
When Emily M. Bender co-wrote her 2020 paper on what LLMs can and cannot do, it was an academic argument; by March 2023, when New York Magazine profiled her, the failures she warned about had names — Microsoft's Bing producing creepy love letters and Meta's Galactica spewing racist rants before being pulled. The profile converts a critique that was easy to dismiss in 2020 into an I-told-you-so moment with a public face.
The arc since then shows the critique hardening into an institution: Bender coined "stochastic parrot", went on to argue in her own essay that the rogue-AI-versus-actual-harms framing is ahistorical, and by 2025 was interviewed by the Financial Times about her skepticism and the book The AI Con. The profile is the hinge where academic linguistics became a mainstream counter-narrative to AI industry optimism.
First-order effects
- Bender's 2020 paper gets retroactive validation: Bing's and Galactica's public failures make her "what LLMs cannot do" argument legible to a general audience rather than just linguists.
Second-order effects
- The critique camp immediately fractures — Bender's schism essay pushes back on existential-risk framing while Platformer's counterargument insists skeptics are underestimating how real and dangerous AI is, forcing the debate to define what "harm" means.
Third-order effects
- If the pattern holds, scrutiny shifts from model outputs to the machinery behind them — the hidden tasker workforce at companies like Scale AI and the industry speech culture captured in the fired Google employee's account — with critics institutionalizing into books, columns, and standing media presence.
The trend: LLM criticism is moving from academic papers to mainstream media profiles, even as the skeptic camp splits over whether the real danger is hype or the technology itself.