←
Patrick Rim44 posts
Patrick Rim avatar

Patrick Rim

@PatrickRim

Co-Founder & CEO @buildpbc Prev. @GoogleDeepMind, @NVIDIA, @Meta, @Caltech, @Yale

◷ Joined June 2024▧ 6 Photos and videos
Eddy Xu avatar
⟳ Patrick Rim retweeted
Eddy Xu@eddybuild·Oct 18

a great mentor once told me that a startup is an external reflection of a founders inner world

Patrick Rim avatar
Patrick Rim@PatrickRim·Oct 18

It's my great honor to lead Build into our next chapter: massive research bets and long-term partnerships in pursuit of physical superintelligence. If I do my job, @eddybuild will go down as the greatest visionary of our generation.

a year ago, @buildpbc made a bet that human data scaling was the bottleneck to solving physical labor it turns out this was a pretty good bet. the next stage of this competition is driven by research. as such, i've promoted my co-founder and chief scientist @patrickrim to ceo! patrick is a crazy man dressed up as a thoughtful researcher. once you talk to him, you'll see what i mean :) data companies with exceptional researcher CEOs have critical advantages across hiring, sales, and data quality. in the last 60 days, patrick: - pushed sota on 3d pose (from scratch) - built an outlier eng & research team in lightspeed - assembled the largest academic consortium in physical ai - won the largest contract in build's history if things unfold as i predict, patrick will go down as an N-of-1 founder of our generation. -eddy founder & president, build ai

Post attachment
Patrick Rim avatar
Patrick Rim@PatrickRim·Oct 6

Incredible, wouldn't have bet against @kevinbanghe and glad to be proven right

Today we introduce Kairos 1: the world's first model built to simulate individual human behavior. Frontier models only understand population averages, ignoring the individual differences that make us human. Kairos beats Persimmon, Astra, Fable, and Gemini across 13 simulation benchmarks.

Post attachment
Patrick Rim avatar
Patrick Rim@PatrickRim·Sep 23

oom more data coming in hot 🔥

Introducing Midcentury. We’re building the data and simulation infra for physical AI. Today, we’re coming out of stealth with a $15M Series Seed to scale robotics beyond polished demos. We’re already supporting frontier labs with: → The world’s largest egocentric dataset: 2M+ hours, 50+ environments, 20,000+ tasks → Matrix: a frontier simulation platform scaled with our real-world data to evaluate and post-train policies at scale

1:22
Patrick Rim avatar
Patrick Rim@PatrickRim·Sep 17

throwback to when I collected a million distorted and occluded digits off football jerseys models don't fail on clean data, they fail on the wrinkles. now we collect wrinkles at scale!

Post attachment
Kingston Kuan avatar
⟳ Patrick Rim retweeted
Kingston Kuan@kstonekuan·Aug 23

Gotta give props to Build AI for holding the record for so long on this too; it was an awesome early release for the open source community

Build AI avatar
⟳ Patrick Rim retweeted
Build AI@buildpbc·Aug 11

pre-trained on Egocentric-1M data!

Today we are introducing Dyna-2, a world-action model pre-trained on one million hours of human video. At this scale, for the first time, we discovered several new scaling laws: • world-action models exhibit scaling law on human data across four orders of magnitude, from 1000 to 1,000,000 hours, • this human data scaling law implied a scaling law on never seen robot data, • both data and objective matter; world modeling and scaling on video data are essential for cross-embodiment scaling transfer to emerge 🧵

1:42
Patrick Rim avatar
Patrick Rim@PatrickRim·Aug 24

I’ve left DeepMind to lead research at Build AI Our first bet: scale monocular vision. We haven’t even scratched the surface :D

Steven Cheng avatar
⟳ Patrick Rim retweeted
Steven Cheng@stevencheng·Aug 12

1M hours of human video sounds insane. That’s roughly 170 years of continuous waking experience. If you had people wear cameras specifically to collect it: 1,000,000 hours ÷ 8 hours/day = 125,000 person-days That’s ~343 people collecting 8 hours a day, every day, for an entire year. That’s a massive data operation. So I’m really curious: how exactly did Dyna-2 collect its 1M+ hours of egocentric human video?

Dyna-2, a World Action Model (WAM) — not a Vision-Language-Action (VLA) model — pre-trained on one million hours of human activity data. WAM is starting to show promising products. Excited to see its generalization capabilities improve dramatically.

1:42
Patrick Rim avatar
Patrick Rim@PatrickRim·Jul 27

lowering the cost of intelligence increases its consumption - creating jobs we can’t yet fathom