Jie Tan’s Post

There's no doubt AI is going to fuel robotics. I love this example from Google DeepMind which shows the power of AI. The video shows a robot remembering a detailed tour of a building and then responds to a query to navigate the space to find an area that is best for drawing - taking the user to a whiteboard. I wish I had something like this for when I forget where I leave my keys! 😂 But in all seriousness, imagine a mini version of this attached to the clothes of someone who has dementia, helping them navigate through spaces and tasks, it could transform their quality of life! #ai #innovation #robotics #googledeepmind

Google DeepMind How long can its battery last? Two universal and basic elements to make the machine work are: battery (electricity) and metadata. Without battery, the machine won't move. Also, battery accounts for 40% of the total cost of an electric machine, e.g., electric car. For AI applications, the electricity bills of a C-level friend of ours has doubled because of using AI. Metadata is an enabler. It's a digital content, NOT a technology. It's like a treasure map, curated by human for treasure hunting. Without metadata, NO data can be found/retrieved, even by the most advanced technologies, like AI, quantum computers, etc. Thus, the machine won't be able to do even a simple task like to move and to sense the situation. https://lnkd.in/g-aJFnXR We are using our expertise/IP, a copyrighted multilingual metadata, to do what AI, like ChatGPT, can't do in data analytics NOW. Do you or any contacts of yours need our expertise/IP? Thanks. 關於 AI 人工智能的一些事實 (Chinese version): https://lnkd.in/eeVbdcza Some fact findings about AI (English version): https://lnkd.in/gwcPNUPP

Introduce Mobility VLA - Google's foundation model for navigation - started as my intern project: - Gemini 1.5 Pro for high-level image & text understanding - topological graphs for low-level navigation - supports multimodal instructions co-lead Zhuo Xu, Lewis Chiang, Jie Tan

Discover Advanced Robot Navigation with Gemini 1.5 Pro The Gemini 1.5 Pro model, with its impressive 1 million token context window, has redefined robot navigation. This advanced system enables robots to recall detailed environments from video tours and follow complex instructions, from finding a specific desk to remembering favorite drinks. Learn more about how Vision-Language Models and topological graphs make this possible in the latest research paper: ["Mobility VLA: Multimodal Instruction Navigation with Long-Context VLMs and Topological Graphs"](https://dpmd.ai/4bUobbj). #AI #Robotics #Innovation #TechResearch #MachineLearning

Check out Google DeepMind Gemini 1.5 Pro model, with its 1 million token context window, can navigate busy environments by recalling details from a video tour and following complex instructions. Whether it's finding a specific desk or remembering a favorite drink, this technology is pushing the boundaries of what's possible. #google #ai #robotic

Introduction to Jupiter service robot，it can perform face recognition, interact with people through voice, and guide visitors to designated locations to provide fixed-point explanations.

New Atlas video from Boston Dynamics looks very impressive ! A couple things that caught my attention: 1. Compared to other humanoid demos (mostly teleoperated), the robot is much faster at walking with the load and balancing the body during manipulation. In my opinion, this where Boston Dynamics is ahead from the competition. Most of the humanoid demos thus far demonstrated dexterous manipulation while standing still or moving very slowly. 2. The hand design looks quite different from the competition. While others have opted for human-like hands with many degrees of freedom, BD has opted for a more functional and robust design with 3 fingers. This should be sufficient for most tasks in logistics and manufacturing that require handling rigid and heavy objects. Would love to hear your thoughts on the robot in comments below.

Seeing the geometries overlaid in real time with the world shows how smart our system is about what we're handling, how it affects our dynamics, and how we calculate the perfect poses for grasping, moving, and placing this rather heavy strut. Absolutely amazing demonstration of what humanoid robots can do for the future of manufacturing. Way to go Atlas! 💪

This morning, I was captivated by this video showcasing a humanoid robot from Boston Dynamics, effortlessly demonstrating strength and precision. It's moments like these that remind me of the incredible pace of innovation we're experiencing in 2024. As I reflect on these advancements, I can't help but acknowledge the mixed emotions they evoke : On one hand, there's the awe-inspiring potential to enhance efficiency and improve lives. Yet, there are also concerns about job displacement and #cybersecurity threats. However, amidst these challenges there are tremendous opportunities, I'm optimistic about our ability to #adapt, #learn, and #collaborate. By embracing change and fostering ethical practices, we can navigate the complexities of this era while creating a world that's not only technologically advanced but also compassionate and inclusive. #ai #humanoid #bostondynamics #robots #TechRevolution #Innovation #FutureForward

Looks like Atlas knows this particular thing and hence knows how to grab that thing and put it into place. Can Atlas handle arbitrary objects? Anyway, the way it makes steps is more impressive to me for some reason.