Introducing GEN-0, our latest 10B+ foundation model for robots
⏱️ built on Harmonic Reasoning, new architecture that can think & act seamlessly
📈 strong scaling laws: more pretraining & model size = better
🌍 unprecedented corpus of 270,000+ hrs of dexterous data
Read more 👇
I’m excited to announce @generalistai We believe the path to general-purpose robots starts with precise, fast, and resilient manipulation. What you see here are end-to-end AI models, trained from scratch, doing some very hard tasks. This is pixels in, actions out.
While I’m proud of how far we’ve pushed the frontier in a year, this is only the beginning of the end-to-end AI era in robotics. This is not our “ChatGPT moment” — this is what the “GPT-2 of robotics” looks like.
Robotics is at an inflection point right now that feels analogous to NLP pre GPT-3. IBM Watson, sentiment analysis, translation, and image classification were all specialized, hardcoded tasks. Robotics has picking, fulfillment, and dozens of other specialized, hardcoded tasks. LLMs and ChatGPT proved the market was bigger than just those NLP tasks. We are proving a Generalist robot can be more.
Last Spring I took off from Google DeepMind, and I've been heads-down building since with an amazing team. Excited to share more today -- introducing Generalist.
It's felt to me for a couple years, since we started bringing multimodal LLMs into robotics, that a subset of the ingredients for creating truly general purpose robot intelligence seem to be falling into place.
But what's been needed is a new focus at the intersection of data, models, and hardware. No amount of downloading data from the internet, by itself, will create the level of fast, fluid, precise, reactive layer of intelligence in being able to interact with the physical world.
In due time we'll be excited to share more, but what we're sharing today is about what the models have grown to be capable of.
We think we've hit a new point on the frontier of general purpose real world intelligence – new levels of simultaneously fast, smooth, precise, reactive, bi-manual coordinated dexterity.
Looking forward to sharing even more. Super proud of the team we've put together, and where we're headed. Reach out if you'd like to chat about working together!
@HM__Clare Rough and relatable. It took me digging through records, researching a lot on my own, and condensing everything into a one-pager to enable an MD to see the whole story, get dx and treated. (Obviously we shouldn't have to put in this much work, but I definitely did.)
@jack_turban fought really hard to make sure we didn't do this at athenahealth -- having worked at epic too there's a good chance that other replies are right, and (in 12 months with an expensive upgrade & lots of politics) if could be fixed
@johnprendergass Moving data around in healthcare is thwarted by economic incentives, not lack of a decentralized ledger. Cool tech doesn’t change the fundamentals of the market.
@MsPackyetti I’m really sorry you’re in this process (especially right now). It sounds like you’re doing everything you can think of. Your pain is not your fault. Not responding to meds the way doctors expected is not your fault. Not having an answer yet is not your fault.