zeynep tufekci
@zeynep
Complex systems, wicked problems. Society, technology, science and more. @Princeton professor. @NYTimes columnist. My newsletter @insight http://theinsight.org
I'm launching a newsletter, called @insight. I want to try to create a space for thinking deeply about complex, thorny but important puzzles, along with a community. Something between public writing and social media: some of both but neither. Sign up here: zeynep.substack.com/p/welcome-to-i…
Emergent misowlignment!
New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
How do you feel that so many of my students thanked me for introducing them to the work of Tom Lehrer in my class on algorithmic accountability? youtube.com/watch?v=QEJ9Hr…
So sad to see that Tom Lehrer died. Musical and math prodigy and iconoclastic star of the conformist 1950s who totally rejected fame, one of the most interesing people I've ever written about
So is there any systematic study for the LLM-dash, or is it lore? (I believe it, too, fwiw). Or why it shows up like that? “LLM-dash”, my term for it, works well if you read it out loud. It’s not just an name — it’s a memorable signifier.
i wake up. something’s wrong with the clock on the wall. the numbers are jumbled. my hands aren’t right. i tell my wife. she responds: “that’s not just an observation—it’s a powerful insight.” i scream.
Breaking News: The EPA said it would eliminate its scientific research arm and begin firing hundreds of scientists, after denying that it intended to do so. nyti.ms/4500xIU
Breaking News: The Supreme Court agreed to allow the Trump administration to proceed with dismantling the Education Department by firing thousands of workers. nyti.ms/4lUrGUu
Outstanding article by @zeynep. "Plausibility Engine" is a great characterization of LLMs
I wrote about "MechaHitler", which is about more than Grok when we view LLMs correctly: as remarkably powerful plausibility engines. All the AGI talk and the Eliza effect fried brains, imo. Important tech needs to be understood on its own terms. Gift: nytimes.com/2025/07/11/opi…
“plausibility engines” is an excellent descriptor for LLMs
I wrote about "MechaHitler", which is about more than Grok when we view LLMs correctly: as remarkably powerful plausibility engines. All the AGI talk and the Eliza effect fried brains, imo. Important tech needs to be understood on its own terms. Gift: nytimes.com/2025/07/11/opi…