Ted Xiao
@xiao_ted
Followers
17K
Following
8K
Media
312
Statuses
2K
Building something new. Previously Gemini and Robotics @GoogleDeepMind. Posts about frontier models, physical AGI, and scaling. Opinions my own.
San Francisco
Joined October 2013
Fundamental spatial and temporal understanding is the bedrock upon which robots will learn motor control. These types of Embodied Reasoning capabilities enable policy learning but also inference abilities like image or video conditioning. We just released a new SOTA ER model!
5
43
301
I like the analogy of human emotion as an example of weak superalignment. Evolutionary pressures selected for biochemical value functions that generalize well (“trust your instincts”) and survived millennia until the advent of pharma (GLPs) and tech (ads, socials, vice).
1
0
10
Great to see the increasing acknowledgment of large model pretraining as a distinct scientific discipline! Proper laddering, high quality evals, model health monitoring, stabilizing training dynamics… maybe not as flashy as post-training but even more critical to get right.
Consider joining us
5
3
98
APOL1 Genetic Testing at No Cost to Patients! Are you incorporating APOL1 genetic testing into your workflow? Discover No-Cost APOL1 Genotyping Program for eligible patients sponsored by Vertex Pharmaceuticals. Learn more today!
18
35
208
“In this pivotal moment, the challenges we face require a historic national effort, comparable in urgency and ambition to the Manhattan Project” - Genesis Mission, DoE. AI for science and discovery at the nation state level. Are you paying attention yet?
3
1
47
Dhruv is one of the sharpest, funniest, and kindest researchers I’ve worked with. Highly recommend considering his new lab for PhD and postdoc opportunities!
My group @Princeton is hiring! We are looking for strong postdoc and PhD candidates to join our quest for intelligent robots in open-world environments. Read more below and get in touch 🤖🐅🧡 https://t.co/7o35pwPZCz
1
7
138
Reasoning 🤝 Native Image Generation Nano Banana Pro is an incredible model, the best illustration yet of the endgame of what general positive transfer across modalities and capabilities can look like. Painful to get omnimodality right, but when it works, it feels like magic!
We just dropped Nano Banana Pro, built on Gemini 3. 🍌 With state-of-the-art text rendering, vast world knowledge and studio-quality creative controls, Gemini 3 Pro Image can create and edit more complex visuals, infographics and more. Here’s what’s under the hood. 🧵
0
0
12
Great seeing the amazing research progress at UZH, thanks @RobertoPelleri9 @davsca1!
Great to have @xiao_ted stop by the lab today! We had a fantastic conversation about the bright future of robotics and the potential of VLAs✌🏼
0
1
18
This is hands down the best robotics startup launch I have seen 🔥 Full-stack innovation from hardware design to betting big on UMI to smooth and impressive policies. After 1.5 years of stealth, instantly launching as a top-5 robot foundation model team. Congrats!👏
Today, we present a step-change in robotic AI @sundayrobotics. Introducing ACT-1: A frontier robot foundation model trained on zero robot data. - Ultra long-horizon tasks - Zero-shot generalization - Advanced dexterity 🧵->
8
19
278
Gemini 3 models from @Google @GoogleDeepMind have made a significant 2X SOTA jump on ARC-AGI-2 (Semi-Private Eval) Gemini 3 Pro: 31.11%, $0.81/task Gemini 3 Deep Think (Preview): 45.14%, $77.16/task
193
624
4K
Gemini 3 is out!! The jump from 2.5 to 3 is the largest step function improvement I’ve seen from my time on the Gemini team. This is a really, really smart model with significant breakthroughs across pre-training and post-training.
I’m really excited about our release of Gemini 3 today, the result of hard work by many, many people in the Gemini team and all across Google! 🎊 We’ve built many exciting new product experiences with it, as you’ll see today and in the coming weeks and months. You can find it
2
5
83
Great stopping by the @mimicrobotics office! Very impressive hardware and teleop stack, even better in-person than the videos.
giving @xiao_ted and @ChongZitaZhang a tour of @mimicrobotics!! Always happy to show that Europe has real robot learning!
0
1
29
VLA + RL finally unlocked with pi0.6! 🚀 Stabilizing RL to work at VLA scale is not easy, exciting to see some of the most compelling results in the last 1.5 years. I like their formulation of advantage-conditioning a lot, it’s very compatible with web-scale VL training.
We developed a general recipe that allows VLAs to improve from experience. RL is back. (yes, this is 13 hours of coffee making)
4
7
110
It is a travesty that all AGI labs are sprinting away from the physical world and towards digital companions, customer service bots, enterprise software, and coding agents. That the real world is difficult and messy and unforgiving is exactly why we need to be prioritizing it.
14
9
138
Visiting Zurich🇨🇭 this week! Would love to learn what researchers in AI and robotics are excited about. Who should I chat with?
9
1
45
Futures Traders: Now you can get 30% off for life, never pay an activation fee, and get 100% initial test refunds with your first withdrawal! Use code NOFEE100 to claim this limited time offer.
0
6
38
Seeing SIMA’s evolution the past years has been an inspiring case study on the unique strengths of utilizing a frontier model as a digital agent: advanced reasoning, multimodal understanding and conditioning, generalization and task/concept transfer. Congrats to the team!
SIMA 2 is our most capable AI agent for virtual 3D worlds. 👾🌐 Powered by Gemini, it goes beyond following basic instructions to think, understand, and take actions in interactive environments – meaning you can talk to it through text, voice, or even images. Here’s how 🧵
2
7
69
After 8 unforgettable years, I have decided to leave Google DeepMind. I feel immensely grateful to have had the opportunity to help transform the dream of general-purpose robot learning from a heretical fringe idea into a normalized technology roadmap. It has been the honor of a
46
8
472
Very fun discussion with @shahdhruv_ at @BuzzRobot! We covered a pretty broad spectrum of topics in robotics + AI, and shared why we are so excited about the technical breakthroughs in Gemini Robotics 1.5 🤖
Had the pleasure of discussing with @xiao_ted and @shahdhruv_ from Google DeepMind the SOTA robotics model, Gemini Robotics 1.5. Our guests shared the novelty of the model and its new embodied reasoning capabilities. As mentioned in the discussion, robotics today feels like it’s
0
0
24
Post-training scaling laws on Task A may not correlate at all for Task B
0
0
8
[0] Generalist GEN-0 blog post https://t.co/TsK1CenNaj [1] OG Scaling Laws paper https://t.co/pUeylAwG4s [2] GPT-4.5 Pretraining https://t.co/0lCbXVM3He [3] Off-Policy Evaluation https://t.co/sp09kc6Gpt [4]: SIMPLER https://t.co/Nri87KvRGq [5]: WorldGym
simpler-env.github.io
Project page for Evaluating Real-World Robot Manipulation Policies in Simulation
0
0
3