Chris Sweeney
@the_shweenz
Followers
748
Following
372
Media
75
Statuses
2K
Meta Reality Labs Research. Research Scientist. I read binary at a 6th grade level.
Joined October 2010
No teleoperation. No simulation. No RL. Multi-fingered robot manipulation policies directly by watching videos of humans with Aria glasses on. It was super fun working on this with @irmakkguzey's lead!! 1/n
4
30
258
Almost 30 years later and everybody is still sleeping on The Miseducation of Lauryn Hill
0
0
0
How come that popcorn place I bought from 12 years ago is in better email contact with me than my college roommate
0
0
0
The best part of waking up is NOT having Folgers in my cup. I am drinking Intelligentsia Black Cat Analog at the moment actually
0
0
0
I honestly do not know anybody who does not work for Google but regularly uses Gemini. My experience is all these big company models are excellent art doing at least one thing. Folks who use it - what is Gemini best at? What should I try with it?
This is Gemini 3: our most intelligent model that helps you learn, build and plan anything. It comes with state-of-the-art reasoning capabilities, world-leading multimodal understanding, and enables new agentic coding experiences. 🧵
0
0
0
Imagine inventing the lightbulb before electricity
30 years ago: Transformers with linearized self-attention in NECO 1992, equivalent to fast weight programmers (apart from normalization), separating storage and control. Key/value was called FROM/TO. The attention terminology was introduced at ICANN 1993 https://t.co/m0hw6JJrbS
12
17
349
They say a man dies three times. Once when he takes his last breath. Once when his name is spoken for the last time. And once when his tokens produce their last usable gradients.
1
0
1
Last night my kid said she was scared and I told her she doesn't know fear until she's been chased by a Yeti playing SkiFree on your grandma's computer
1
0
1
If you really want to feel old remember that when you started your PhD you were writing makefiles because python and cmake weren't popularized yet and caffe was just the thing at the uni that gave you diarrhea
If you really want to feel old, remember that when you started your PhD you were writing Caffe neural net layers in C++ and constantly messing up the backprop equations such that you’d get no gradient flow, or that you used intermediate VGG-16 “feature maps” as embeddings for all
0
0
4
The Project Aria Team is excited for The Fourth Hands-on Egocentric Research Tutorial with Project Aria 😎 TOMORROW Monday, Oct 20th 📅Monday, Oct 20th 🕐8am-12pm 📍NOW IN in room 328 Featuring inspiring talks from @rapideRobot, @LingniMa, @danfei_xu, @bowenwen_me and more!
0
2
3
Very interesting business model
Introducing RTFM (Real-Time Frame Model): a highly efficient World Model that generates video frames in real time as you interact with it, powered by a single H100 GPU. RTFM renders persistent and 3D consistent worlds, both real and imaginary. Try our demo of RTFM today!
0
0
0
People can debate who would win a 1v1 between Michael Jordan and @KingJames but if it were 5 MJs vs 5 Lebron's there's no doubt which team is winning right?
0
0
0
AFAIK these models are generating pixels using color and optical flow during training. Due to projective ambiguity more than one 3d configuration could project (render) to the same image. It just so happens that the pixels these models generate often appear physically plausible
0
0
0
The laws of physics staying intact? My man does a beautiful back flip with a twist from a hanging position on a basketball rim!
This is one of the more mind-blowing sora 2 samples. When the model makes a “mistake,” it’s often because the internal agent being simulated messes up while the laws of physics stay intact.
2
0
0