Daniel Severo

Researcher at Meta Superintelligence Labs - FAIR / PhD @UofT @VectorInst / Previously @GoogleAI @UFSC 🇧🇷

[Daniel Severo @_dsevero]

More Research Scientist positions opening up in Montreal!

We're looking for candidates with experience in video gen, RL, world models, and adjacent topics. Come join our World Modelling team at our Petite Italie / MILA office.

https://t.co/Pb9CqTQATB

Feel free to reach out!

[Daniel Severo @_dsevero]

@giffmana I'm 5 vesting events away. Muse will save me.

[Daniel Severo @_dsevero]

@levelsio @sobedominik Why in the world are you coming to Floripa for the malls lol. Did you check out the southern bit of the island? It's insane. Go to Florida if you want a mall

[Mark Zuckerberg @finkd]

Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple.

[Daniel Severo @_dsevero]

@BlackHC It’s like going to the gym for ur brain. I started with coding but quickly ended up doing random problem sets from books I like.

Maybe it’s worth it to learn a new coding paradigm? Something that helps u develop new abstractions. The code itself isn’t that important anymore. Idk

[Artem Zholus @artemZholus]

Introducing RoboJEPA, the first scaling law for multiembodiment world models.

We scaled JEPA predictors up to 8B params, showed scaling laws extrapolate and correlated compute scaling with planning performance.

We release all checkpoints and the code.
https://t.co/nSRBjasDBk

[Build Canada @buildcanada]

Losses were dominated by public sector jobs (particularly in education and health care) and self-employment.

Private sector jobs however, rose by 24,100.

📊

[Daniel Severo @_dsevero]

@BNNBloomberg Wrong. Read it again and fire your intern who made this post.

[Aran Komatsuzaki @arankomatsuzaki]

Meta presents RoboJEPA

Scales JEPA predictors from 22M to 8B parameters (largest model trained to date), using data across 12 robotic embodiments, and studies how latent prediction error can forecast downstream planning performance

https://t.co/i3EH6coQXX

[Daniel Severo @_dsevero]

@DimitrisPapail I have a hard time understanding who thinks the opposite. My job has become 10x more productive and 100x more fun. We're literally living in the future. I'm mind blown every day by the ease of experimenting. I can just make any tool I need in minutes. What a time to be alive.

[Daniel Severo @_dsevero]

Gotta admit this is great. Goverment admin work really needs a rehaul. AI could make things so much more efficient and reduce the cost of goverment.

Can we get this too @MarkJCarney, maybe @cohere can help :D

[Department of State @StateDept]

SECRETARY RUBIO on https://t.co/pt9rQXC44r: This is really going to be a revolution in the way that government works for the average American.

[Daniel Severo @_dsevero]

$49 per page 😂 what a joke mafia. People should be ashamed.

[Daniel Severo @_dsevero]

Why are "government approved translators" for documents still a job. So useless.

[Daniel Severo @_dsevero]

Very cool way to recruit!

[Vector Institute @VectorInst]

The standard graduate application process is backward.

You spend months researching programs, navigating faculty pages, and emailing professors who may or may not respond. You pay hundreds of dollars in application fees. Then you wait to find out if a lab actually has space, funding, or any interest in your research.

Great AI research starts with the right supervisor match. The process should too.

🧵

[Daniel Severo @_dsevero]

@giffmana I love that the page is clearly 100% AI generated

[Daniel Severo @_dsevero]

@sarahookr Highly recommend Cali -> Santiago -> Florianópolis, over the Andes

[Daniel Severo @_dsevero]

@VictorTaelin Dahora d+. Parabens!

[Taelin @VictorTaelin]

Bend 2 is here!

It is a new programming language that blocks AI mistakes via *proof checking* - the same technique big AI labs used to solve open math problems, like Navier-Stokes.

It is also very fast, and runs on GPUs.

Watch the video. Link in the comments.

[Daniel Severo @_dsevero]

@sarahookr Bom d+