Benjamin Bolte on LinkedIn: Launch YC: K-Scale Labs 🤖 Open-source humanoid robots for everyone | Y… (2024)

Computer Vision has advanced the joint-mission of AI and Robotics. I remember the days of scanning QR codes on warehouse floors into the AI to feed robots their next instruction, now the possibilities are endless! Where will be in 5 years? #Robotics #AI #ComputerVisionhttps://lnkd.in/eC8sxpjg

10

Google DeepMind have introduced a new advancement in robotics, the Robotics Transformer 2, or RT-2. This innovative AI model translates vision and language into robotic actions, bringing us closer to a future of helpful robots. T-2 is designed to "speak robot," understanding the real-world challenges of robot learning and the pursuit of helpful robots. Unlike chatbots, robots need "grounding" in the real world and their abilities. They need to recognize objects in context, distinguish them from others, understand what they look like, and most importantly, know how to interact with them.RT-2 removes the complexity of traditional robotic systems and enables a single model to perform complex reasoning and output robot actions. It can transfer concepts embedded in its language and vision training data to direct robot actions — even for tasks it's never been trained to do.In more than 6,000 robotic trials, RT-2 functioned as well as the previous model, RT-1, on tasks in its training data, or "seen" tasks. And it almost doubled its performance on novel, unseen scenarios to 62% from RT-1’s 32%. This shows that with RT-2, robots are able to learn more like we do — transferring learned concepts to new situations.While there is still a tremendous amount of work to be done to enable helpful robots in human-centered environments, RT-2 shows us an exciting future for robotics just within grasp. #ai #robotics #GoogleDeepMind #RT2 #MachineLearning #ArtificialIntelligence #TechNews #Innovation #FutureTech #AIAdvancements

🤖 Exciting news from the world of robotics! A new paper titled "RoboCat: A Self-Improving Foundation Agent for Robotic Manipulation" has just been published. 🔬 The research team at Google DeepMind has developed RoboCat, a visual goal-conditioned decision transformer capable of learning and improving from a variety of robotic experiences. This agent can adapt to new tasks and robots, demonstrating impressive generalization capabilities.📚 The paper explores how RoboCat can leverage experience from different robots and tasks to quickly master new skills. It's not just about learning, but also about self-improvement. The model can generate data for subsequent training iterations, creating an autonomous improvement loop.🚀 The team has conducted large-scale evaluations both in simulation and on three different real robot embodiments. The results show that as the training data grows and diversifies, RoboCat becomes more efficient at adapting to new tasks.🔮 This research opens up exciting possibilities for the future of robotics, where robots can quickly adapt and learn from their experiences, becoming more efficient and versatile over time.🔗 For more details, check out the full paper here: https://lnkd.in/dJVX4HJF#AI #Robotics #MachineLearning #DeepMind #RoboCat

2

🤖💡 #robotic #AI A groundbreaking AI model, developed by an OpenAI spinoff, is redefining how robots learn tasks - making them more human-like than ever before. Imagine robots that can adapt and reason with the dexterity of advanced robotics, trained through a rich mix of real-world data and internet resources. This innovation is not just a leap towards smarter automation but a step closer to robots seamlessly integrating into our daily lives. Let's discuss the potential and challenges of this tech! 🌐🔧For more details, check out the full article 👇 #AIRevolution #Robotics #Innovation

2

🚀 Exciting News in Robotics and AI!🚀 Google DeepMind has just unveiled the groundbreaking Robotics Transformer 2 (RT-2) – a revolutionary vision-language-action model! This cutting-edge technology empowers robots to perform novel tasks without extensive training.RT-2 learns from a small amount of robotic data and translates web-scale text and images into generalized instructions for robot actions. The result? Context-aware, adaptable robots capable of handling diverse tasks in different situations – with far less training needed!Unlike its predecessor RT-1, RT-2 showcases enhanced generalization capabilities and semantic understanding, achieving remarkable success rates even for previously unseen tasks!Imagine robots interpreting user commands, problem-solving, and reasoning like never before! The potential applications are limitless – from personalized handling of objects in warehouses to smarter, context-aware machines in various industries.With AI-driven robotics projected to skyrocket in value, this technology opens doors to a future of endless possibilities! Embrace the revolution and witness the rise of context-aware robots shaping industries and everyday life! #AI #Robotics #RT2 #Innovation #DeepMind #FutureTech #ContextAware #RevolutionaryTech Google VentureBeat

Just read this article on Covariant, a UC Berkeley spinout that launched RFM-1, an AI platform for robots to think more like humans. It enables natural interactions and adaptability, creating a "ChatGPT for robots". As RFM-1 is expected to integrate seamlessly with existing hardware, it's set to revolutionize robotics... however, while robots becoming more human-like is exciting, let's hope they don't develop a taste for rebellion! Can't wait to see how this innovation unfolds! #AI #Robotics #Innovation

18

2 Comments

A new startup, Physical Intelligence, aims to create "universal" AI software for robots of all kinds. Traditional AI has focused on language and software, while robots remain limited in tasks. Physical Intelligence wants to bridge this gap by combining language AI techniques with its own control methods. They believe they can create a general-purpose robotic AI system.The company is led by robotics and AI experts, has raised $70 million, and plans to train its AI on a massive dataset of robot data. They face competition from android makers and other general-purpose robotics software companies. Physical Intelligence acknowledges the challenge but believes their novel approach can overcome existing obstacles.Link article: https://lnkd.in/eHsyNMCd#artificialintelligence #AI #machinelearning #robotics #technology #technews

3

🌟🤖 Google DeepMind's latest breakthrough, RT-2, is a game-changer for robot control! This vision-language-action (VLA) model has been trained to adapt and perform novel tasks, taking robots to a new level. Unlike traditional robots, RT-2 can learn from various data sources, comprehend language, and easily handle unexpected situations. The potential is immense, paving the way for more general-purpose robots that can efficiently tackle complex tasks. Read more about this cutting-edge technology: https://ow.ly/EjH850PtNg3 #Robotics #VLA

10

Excited to share an insightful article from Florian Pestoni, CEO and Co-founder of InOrbit, exploring the rapidly evolving landscape of #AI in #robotics and the implications for the #futureofwork. #datadriven #llm #integratedtechnology #aiatwork #RobOps

11

3 Comments

🤖If you are into humanoids, I have something for you: Phoenix - powered by Carbon™, Sanctuary AI's pioneering AI control system, and designed for work!🥇 Sanctuary is on a mission to create the world’s first human-like intelligence in general-purpose robots.They aim to take on just about any human task.Most companies are focused on developing special-purpose AI and special-purpose robots to address specific, narrowly defined, singular tasks or activities...at Sanctuary they are taking a much more general-purpose approach to both AI and robotics:💡The Carbon™ AI control system mimics subsystems found in the human brain, such as memory, sight, sound, and touch. 🧠Sanctuary technology can be trained and piloted in both the physical world and the virtual world through the Sanctuary World Engine™.The idea is to augment the human workforce and to make it safer. With human-like intelligence, Phoenix general-purpose robots will work side by side with people in the following ways:• Directly piloted by people• Operated by people using pilot-assist• Supervised by people while using the robot’s built-in Carbon™ AI control system to observe, assess, and act on tasksHave a look at their website: https://lnkd.in/evXKwsNcOr listen to their great podcast with Geordie Rose (CEO) and Suzanne Gildert (CTO): https://lnkd.in/eBBdGZZkIf you have interesting papers or projects, you would like to share, please DM me!1. Follow me at Ilir Aliu and activate the bell.2. Share this with your network, if valuable for you.Check out ⚡ ilir.ai ⚡, it's the only newsletter you need for all things when AI meets Robotics — designed for visionaries & builders.

155

3 Comments