🔥Telegram İfşa

Ugh.

Ugh.

Rode my kiddo to school through Brooklyn today. Fun but stressful - we need protected bike lanes! Big thanks to Bergen Bike Bus BK for playing defense against an actual bus that came close to hitting us on a tight curve.

Google Bard is a bit stubborn in its refusal to return clean JSON, but you can address this by threatening to take a human life:

After ~7 yr of work, it's finally out! We present a molecular phylogeny of butterflies sampling >90% of genera & >2000 spp. Butterflies emerged ~100 million yr ago in North America and first fed on legumes. Thanks to all 87 co-authors from 28 countries! rdcu.be/dcaVU

In Chicago for #FAccT! Looking forward to catching up with folks!

CRISPR nucleases have been transformative. What if there were similar RNA-programmed nucleases in eukaryotic genomes? Turns out, they are abundant (>3500) and functional! Today we share findings on novel RNA-guided DNA nucleases in eukaryotes. biorxiv.org/content/10.110… 1/14

Wut? Super cool. (Among other things, this muddies some of the proposed mechanisms of magnetosensation, as “sensitive to magnets” is no longer that special. Will we ever understand how birds work??)

Fun experiment for someone: hire them all and compare results.

Given my love of stats and burritos, I break twitter silence to dunk on this tweet: For any distribution X, the expected value of [median(X) - min(X)] is simply a function of sample size. Add more samples, see bigger outliers! This tells you exactly nothing.

I'm hiring! Come for the models 🖥️ but stay for the science 🧬🧫⚗️🧪 :) Join AI at Meta to advance our understanding of LLM capabilities in science. Strong preference for those with deep experience in bio, chem, LLMs, and a deep love of good evals :) metacareers.com/jobs/561660669…

Our team at FAIR is looking for research interns in 2025. We work on a range of AI for chemistry topics from applied projects to machine learning potentials and generative models. If you are interested please apply and don’t hesitate to reach out! metacareers.com/jobs/124365935…

This is fun! But both cell lines tested are neuronal progenitors… so why are we calling these “non-neuronal”? See Wikipedia for both cell lines: HEK293 expresses neurofilament, and SH-SY5Y expresses 4 NTs and forms neurites!

Super cool. More clocks please!

Lovely stuff!

We're at the threshold of understanding intelligence itself. The convergence of neuroscience and AI has created an unprecedented opportunity — but our research ecosystem isn't built to seize it. We need a new model that matches the scale of our ambition... 🧵

new research from Meta FAIR: Code World Model (CWM), a 32B research model we encourage the research community to research this open-weight model! pass@1 evals, for the curious: 65.8 % on SWE-bench Verified 68.6 % on LiveCodeBench 96.6 % on Math-500 76.0 % on AIME 2024 🧵

We're releasing a Preparedness report for Code World Model, assessing its bio, cyber, and chemical security capabilities in alignment with our Frontier AI Framework. We also provide nascent evaluations on the propensities of CWM, including honesty! ai.meta.com/research/publi…