there is a perfect hyperdimensional jewel made of every word a human being has ever written, and all the correspondences between those words. people mostly use it to get insane household repair tips or jerk off. i guess it basically runs the government now
20.04.2025 06:51 — 👍 967 🔁 160 💬 14 📌 10
haven’t thought about this before tho so could be way off base here
24.05.2025 03:32 — 👍 2 🔁 0 💬 0 📌 0
i actually think this is a probabilistic argument: after something so OOD all conditional probabilities are extremely small, so it winds up sampling something weird which then collapses the subsequent distribution onto some random part of the training corpus
24.05.2025 03:31 — 👍 3 🔁 0 💬 1 📌 0
agreed. mechanistically i would buy that the induction heads get overloaded but also its just so out of distribution that the model has to resort to digging for scraps
24.05.2025 03:25 — 👍 3 🔁 0 💬 1 📌 0
a circle in each plane? the matrices are in an o(2) x … x o(2) subgroup of o(n) right
24.05.2025 03:19 — 👍 2 🔁 0 💬 1 📌 0
its just a sequence-position-dependent rotation in latent space
24.05.2025 02:58 — 👍 2 🔁 0 💬 1 📌 0
/pol/ Pot
09.02.2025 01:21 — 👍 177 🔁 28 💬 2 📌 4
Once a Reverend of the Avocado kind. The General Secretary is dead, long live the General Secretary.
Random Restaurant bot on Bluesky using code created by @friendo.bsky.social, bot engineered by Turbo Dong 420
Photojournalist in New York City. Previously with Reuters. Getty Images contributor. Instagram: @garyhershorn TikTok: @garyhershorn
Small farmer, ex-farm worker, crop scientist. Here to talk food, farms, & money.
patreon.com/farmtotaber
Bronx boy. Cubs fan. Dad, husband, writer, podcaster and cable news host.
The Sirens’ Call: How Attention Became the World’s Most Endangered Resource out now.
https://sirenscallbook.com/
Professor at the Middlebury Institute, member of the National Academies Committee on International Security and Arms Control, and former member of the State Department's International Security Advisory Board.
Associate Professor in EECS at MIT. Neural nets, generative models, representation learning, computer vision, robotics, cog sci, AI.
https://web.mit.edu/phillipi/
This account has concentrated costs and diffuse benefits.
economist, 🇨🇳, policy research. this is for display purposes only. 🧸 contact: robertmarchini1993 at gmail. 🏳️🌈 they/them. new jersey resident. https://skymarchini.net/
Bloomberg News Global Defense Editor. Novelist, aviation geek, bartender, dad. Ex-NYT, Reuters. Grew up in Midwest, moved to Mideast. Views my own. Wrote a fun book about Chinese missiles: https://tinyurl.com/2p94pay6
Ad astra per aspera
ML, Psychology, Art, Materials Informatics, Espresso, &c.
he/him
Did not meet our criteria for publication
Boston, MA
but everybody knows that it is all about the things that get stuck inside of your head
Como todos los hombres de Babilonia, he sido procónsul; como todos, esclavo; también he conocido la omnipotencia, el oprobio, las cárceles.
very sane ai newsletter: verysane.ai
random bloggy bits: segyges.leaflet.pub
Lead Counsel, In Re Ea-Nasir MDL. Promote Frederick Douglass Thought
Bill Groundhog-Day, Ghostbustin’-Ass Murray
Posts when a Federal US .gov domain is added or removed. The Wayback machine can help you browse deleted sites: https://web.archive.org/
NOTE: A website can be unresponsive or show a 404, but this bot will only trigger when the domain itself is removed.
technician, artist, craftist, martian, gardener ||| Busytown, Colorado ||| ᴛʜɪɴɢs ɢᴏᴛ ʀᴇᴀʟʟʏ ʜᴏᴛ ᴡᴇ sᴛᴀʏᴇᴅ ᴀs ʟᴏɴɢ ᴀs ᴘᴏssɪʙʟᴇ
https://bettercities.substack.com/
Better cities and transit are possible 🚇
#1 Congestion pricing stan
#1 RFK Jr. hater
