@mendonca_rl

Robotics @GoogleDeepMind Prev - Optimus AI @Tesla, PhD student @CMU_Robotics

Joined February 2020
  • Tweets185
  • Following267
  • Followers3.8K
14Media
Argon can also plan effectively in world models! Here it went through an iterated loop of 1) writing out plans for how to achieve a goal 2) using an omni world model to simulate consequences of actions 3) reflect on produced rollouts to modify the plan. This is the resulting video plan. More details in our blog - developers.googleblog.com/ge…
7
6
3
100
7,818
Russell Mendonca retweeted
Jay Chooi
@chooi_jeqOct 14
Gemini 4 Argon achieves SOTA on real robot capabilities, tied with Opus 5.5 and above GPT-6 Astra. 🧵
29
59
21
808
78,812
Russell Mendonca retweeted
Google
@GoogleOct 13
In robotics, Gemini 4 Argon is speeding up real-world physical learning. We let Argon watch a video of a human playing cornhole. It mapped the environment, calculated whole-body joint movements, and autonomously built the simulation code to practice. When transferred to Apptronik’s Apollo humanoid, the robot sank seven straight shots in a real-world trial.
32
139
76
1,115
227,706
Argon has updated my priors for the role LLMs have in solving robotics. Reasoning can be a powerful mechanism for generalization. Even for low-level control, not just for semantic planning. Watch Argon describe how it learned to play cornhole (shoutout to @RoboticistRohan !)
5
1
75
3,928
Russell Mendonca retweeted
Logan Kilpatrick
@OfficialLoganKOct 13
Gemini
1,050
387
517
6,642
1,263,580
This model has been a huge productivity boost in my workflows. Surprised by its competence on more open ended long running tasks, excited for more people to try!
Google DeepMind
@GoogleDeepMindSep 30
Introducing Gemini 4 Argon – our new frontier model. It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
1
2
12
1,076
Very agile!
Skild AI
@SkildAISep 23
We trained a robot to play football. How? Self-play for 140 years in a virtual World Cup. Meet the #Messinator:
1
16
1,587
Video prompting definitely feels like a more natural way to specify tasks for robots, compared to language. Very cool results!
Skild AI
@SkildAIAug 25
Introducing S1, our new foundation model that learns from one example. It can be taught 10-minute long tasks that it has never seen before, from one video prompt without any fine-tuning. Watch S1 operate in real-time via in-context learning:
1
2
30
2,456
Exciting progress!
Generalist
@GeneralistAIAug 19
Introducing GEN-1.5, a one-shot learner. It can learn new tasks in a few seconds. Show it what to do, and it generalizes. This capability emerged from pretraining on physical data at scale, as a step towards our mission of building general intelligence for the physical world.
1
6
806
Very interesting trend, likely due to the new algorithmic improvements in 3.7 flash
1
16
1,938
This model does exceptionally well for its size. Beyond the benchmarks, I’ve found it to be quite useful in daily work.
Logan Kilpatrick
@OfficialLoganKAug 13
Introducing Gemini 3.7 Flash : ) - it is fast! - 50% lower price than 3.6 flash (through end of year) - strong intelligence increase in only ~3 weeks (thanks to some awesome algorithmic improvements) - available in the API, AI Studio, Antigravity, and more!
16
1,502
This direction will be the consensus view for robot learning fairly soon. Scaleable pre-training is the highest order bit to unlock generalization.
Dyna Robotics
@DynaRoboticsAug 10
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
1
29
3,233
Russell Mendonca retweeted
Demis Hassabis
@demishassabisJul 30
Gemini Robotics 2 is here, with our new suite of models, robots can now reason through every movement to manage tasks that weren’t possible before, like tying delicate knots - and even team up to solve complex workflows. Huge congrats to the robotics team on this great milestone!
Google DeepMind
@GoogleDeepMindJul 30
One brain. For any robot. 🤖 We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.
305
516
48
4,533
504,209
Check out what colleagues at gdm have been up to! Looking forward, I’m particularly excited about what multi-fingered hands can enable, especially for learning from human data.
Google DeepMind
@GoogleDeepMindJul 30
One brain. For any robot. 🤖 We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.
1
12
2,266
Russell Mendonca retweeted
Google DeepMind
@GoogleDeepMindJul 29
75
202
66
2,240
409,012
Russell Mendonca retweeted
Sundar Pichai
@sundarpichaiMay 19
Gemini Omni doesn't just build scenes that look real, it reasons about what should happen next. It combines an intuitive understanding of physics with Gemini's knowledge of history, science, and cultural context. Rolling out today starting with video outputs to Google AI Plus, Pro and Ultra subscribers globally through the @Geminiapp + Google Flow, and @YouTube Shorts this week.
377
737
146
7,606
763,886
Excited for the potential of this paradigm for robotics!
Google DeepMind
@GoogleDeepMindMay 19
We’re dropping Gemini Omni: our first step towards a model that can create anything from anything - starting with video. It combines Gemini’s intelligence with our generative media systems - representing a leap forward in world understanding, multimodality, and editing 🧵
7
1,360
Russell Mendonca retweeted
Boston Dynamics
@BostonDynamicsMay 18
Everyone asks if Atlas can bring them a drink, but this robot can bring you the whole fridge. Using AI-driven behaviors, Atlas is doing hard work and coordinating its whole body to manage heavy objects, balancing complex contact points with accuracy and reliability.
466
1,098
367
6,931
1,419,911
Russell Mendonca retweeted
Google DeepMind
@GoogleDeepMindMay 12
We’re reimagining a 50-year-old interface - the mouse pointer - with AI. 🖱️ These experimental demos show how people can intuitively direct Gemini on their screens using motion, speech, and natural shorthand to get things done 🧵
449
1,030
526
8,451
1,687,622
Russell Mendonca retweeted
Sumir Meghani
@sumirmeghaniApr 16
Today, we're opening @Instawork Robotics Lab (IRL), connecting the robotics industry with > 10 million skilled workers. IRL is offering the first certified workforce program for the physical AI economy ... with 20,000 @Instawork Pros already certified! 👷 Read more from @alistairmbarr @BusinessInsider businessinsider.com/robotics…

The AI boom is making training robots the hot new gig

The race for robotics data is transforming gig work, as Instawork turns its workforce into a key engine for training AI-powered machines.

businessinsider.com
Sumir Meghani
@sumirmeghaniApr 16

The Human Advantage in the Robotics Revolution

Article · @sumirmeghani
17
20
11
121
74,055
Load more