Atil Iscen

Scientist, researcher, engineer @Google. Interested in reinforcement learning, robotics, evolutionary algorithms and multiagent learning.

[Jitendra MALIK @JitendraMalikCV]

I want to offer some unsolicited advice to computer vision researchers jumping into robotics. Don't focus too much on VLMs, VLAs etc. That's fine, but the real action is at the sensorimotor level. Most of the open problems in robotics are in manipulation, which is about hand-object interaction, and contacts and forces are central. Proprioception and tactile sensing are as important as vision. Don't get seduced by cherry-picked demos. You can't do robotics without doing robotics.

[naveen manwani @NaveenManwani17]

🚨Paper Alert 🚨

➡️Paper Title: SAS-Prompt: Large Language Models as Numerical Optimizers for Robot Self-Improvement

🌟Few pointers from the paper

🎯Authors of this paper demonstrated the ability of large language models (LLMs) to perform iterative self-improvement of robot policies.

🎯An important insight of this paper is that LLMs have a built-in ability to perform (stochastic) numerical optimization and that this property can be leveraged for explainable robot policy search.

🎯Based on this insight, authors of this paper introduced the SAS Prompt (Summarize, Analyze, Synthesize) -- a single prompt that enables iterative learning and adaptation of robot behavior by combining the LLM's ability to retrieve, reason and optimize over previous robot traces in order to synthesize new, unseen behavior.

🎯Their approach can be regarded as an early example of a new family of explainable policy search methods that are entirely implemented within an LLM. They evaluated their approach both in simulation and on a real-robot table tennis task.

🏢Organization: Interactive Robotics Lab, @ASU , @GoogleDeepMind

🧙Paper Authors: @asurobot , @lgraesser3 , @atiliscen , @Evo_Daveo , Saminda Abeyruwan, @AlexBewleyAI , @yfzhoucs , @KamaleshKali777 , @Swarooprm7 , @pannag_

📝 Read the Full Paper here: https://t.co/MPIWpt0rfe

🗂️ Project Page: https://t.co/xl07yokRSm

🧑‍💻 Code: https://t.co/vJk50P0waD

🎥 Be sure to watch the attached Technical Summary Video - Sound on 🔊🔊

Find this Valuable 💎 ?

♻️QT and teach your network something new

Follow me 👣, @NaveenManwani17 , for the latest updates on Tech and AI-related news, insightful research papers, and exciting announcements.

[Kevin Zakka @kevin_zakka]

The ultimate test of any physics simulator is its ability to deliver real-world results.

With MuJoCo Playground, we’ve combined the very best: MuJoCo’s rich and thriving ecosystem, massively parallel GPU-accelerated simulation, and real-world results across a diverse range of robot platforms: quadrupeds, humanoids, dexterous hands, and arms.

Best of all? You can get started today with a single command: pip install playground

https://t.co/t6pZCNeOSK

[Laura Graesser @lgraesser3]

Tired on a Friday? Take a break and watch a 25min highlight reel of our robot's table tennis matches. 🍵🏓

💫 Highlights: https://t.co/aHhY90M4jp
💫 Website & paper: https://t.co/F4nz89pjnL

Want to know one of the best things about this project?

[Pannag Sanketi @pannag_]

We @GoogleDeepMind simply couldn't have achieved this breakthrough without the powerful #MuJoCo simulation engine. Its flexibility, ease of use, and advanced modeling capabilities were absolutely critical to our research velocity. 🚀 #AI #Robotics #TableTennis

[Atil Iscen @atiliscen]

I am very proud to share that our work on robot playing competitive table tennis is public now.

[Google DeepMind @GoogleDeepMind]

Meet our AI-powered robot that’s ready to play table tennis. 🤖🏓

It’s the first agent to achieve amateur human level performance in this sport. Here’s how it works. 🧵

[Gabe Margolis @gabe_mrgl]

Traveling to Japan for ICRA? Consider showcasing your recent or upcoming work at the Agile Robotics workshop!

The deadline to contribute an extended abstract is 𝗠𝗮𝗿𝗰𝗵 𝟮𝟵.

Then, join on Monday, May 13 for an exciting slate of invited speakers: https://t.co/tJejqM592u

[George Mayer @GeorgeMayer]

One of the biggest reasons an organization isn't innovative is that managers looks to parse out who did what on a project.

This causes ownership anxiety and hoarding behavior, especially for ideas.

But an individual can only have the start of a good idea. It takes a team to make it whole. So, most ideas die.

Managers think they're doing the right thing—being fair and giving credit where credit is due. But it creates a toxic culture that snuffs out ideas before they even break the soil.

The side effect (ideas dying) is actually a counterfactual, which makes it hard to measure or even see. It's not a strict tradeoff or measurement of opportunity cost. There's no regret baked in ("Oh we should have done that") because the idea died—*poof*—in someone's head.

A strong performance culture needs to recognize that when you look back at the best ideas, it's actually hard to remember who came up with which parts, because they had so many contributions and evolved so much over time. And if a team produces great things based on creative ideas, there should be broad-based recognition.

I don't know how to change this type of culture. But if you feel like your organization is lacking innovation, perhaps consider the performance incentives that are baked in.

[Google DeepMind @GoogleDeepMind]

Introducing MuJoCo 3.0: a major new release of our fast, powerful and open source tool for robotics research. 🤖

📈 GPU & TPU acceleration through #JAX
🖼️ Better simulation of more diverse objects - like clothes, screws, gears and donuts
💡 Find out more: https://t.co/KbzxeALTMV

[Atil Iscen @atiliscen]

You can find more details in the blog post above as well as the paper: https://t.co/QPBoLL5Szy and our
video: https://t.co/rqnRMEPKis

[Atil Iscen @atiliscen]

I'm very excited to present our latest work on benchmarking agility where we have a quadruped robot tackle an obstacle course. This was many years of work with so many people working together.

[Google AI @GoogleAI]

Presenting Barkour, a robotic agility benchmark that tests the implementation of low-level locomotion skills useful for movement through real-world environments, along with an ML-based example generalist locomotion policy for quadruped robots. Learn more → https://t.co/GFHauFkVXI

[Atil Iscen @atiliscen]

and present 'Locomotion-Transformer Architecture' to combine different skills such as running, climbing and jumping over obstacles. This allows us to train one policy that achieves diverse set of skills.

[Agility @agilityrobotics]

Digit working autonomously in our replica work cell. #MadeForWork

[TunnelBear @theTunnelBear]

In light of internet censorship, following the devastating earthquake, we have made the decision to add Turkey to our bandwidth program.

Anyone connecting from Turkey will have 10GB of data added to their account.

[New York Times World @nytimesworld]

Twitter has been blocked on several networks inside Turkey, according to NetBlocks, a group that tracks internet outages — taking out a key communication channel for coordinating relief efforts after Monday’s devastating earthquake. https://t.co/Qf6IZySGFY

[Arif Cem Gundogan @arifcemgundogan]

157 earthquakes in the last 10 hours... two of which were 7.7 and 7.5 magnitudes. The extent of the disaster is immense in Türkiye. I made below charts to show the scale. It is very cold, winter time. Please consider donating to relevant campaigns.

[Marine Blue Films @marinebluefilms]

#AnadoluKadimDoğa belgeselimizin çekimlerinde 20 ayı geride bıraktık, kaldı 4 ay. İlk tanıtım filmimizi çekimlerin henüz başında paylaşmıştık, yeni bir film zamanı geldi. “Korumak için tanımak lazım” sloganıyla yola çıktığımız belgeselin ikinci tanıtımını beğenmeniz dileğiyle.

[Atil Iscen @atiliscen]

https://t.co/er5Pz2SoZp