Muhammad Zain Khawaja

Exciting developments in robotics! 🤖✨ Advancements in robot hardware, compute, data infrastructure, and especially generative AI are creating the perfect conditions for robotics to thrive. Here are some key insights: Data Challenges in Robotics: Not Enough Data: Training AI models for robots requires vast amounts of multimodal data (like LiDAR, camera, and tactile inputs), which is hard to collect. Innovative methods like using simulations, synthetic data, and better translating human actions into robot commands are helping overcome this scarcity. Too Much Data: Once robots are deployed, they generate massive amounts of data from various sensors, making it challenging to process and analyze. New platforms are emerging to manage this data deluge effectively. AI Accelerating Robotics: Faster Time to Market: AI is speeding up robot development by enabling rapid iteration and improvement without extensive coding. Versatility and Adaptability: Generative AI models help robots understand and interact with their environment more naturally, allowing them to generalize tasks and adapt to new situations. Ease of Deployment: AI advancements make it easier to program and deploy robots through natural language prompts and adaptive control systems. Hardware Advancements: Declining Costs: The price of robot hardware, including sensors and actuators, is decreasing, making robots more affordable and accessible. Better Performance: Improvements in sensors, processors, and compute systems enable robots to perform complex tasks more efficiently. Rapid Progress: Quick Iterations: Companies are developing new robot models and capabilities at unprecedented speeds, indicating a promising future for the industry. Now is a thrilling time for robotics, with technology overcoming previous limitations and opening up new possibilities. The intersection of AI and robotics is set to transform industries and everyday life. #Robotics #AI #Innovation #FutureTech

10

Foxglove Actuate is just around the corner! Join our co-founder and CTO Ilia Baranov next Wednesday for a rousing panel discussion on when sim is useful (and more importantly, when it isn't) in the robotics development process. Excited to see you all there 🤖🦾 #robotics #simulation #roboticsdevelopment #actuate2024

13

1 Comment

Spot on: "In 2015, the cost of a LiDAR unit was no less than $75,000, and today a 3D LiDAR system costs as much as up to $7,000. Significantly cheaper than a decade ago, yet still not cheap enough. This is where we see startups like Sonair developing ultrasound-based 3D distance sensors which provides autonomous robots with omnidirectional depth sensing, at a much lower cost vs LiDAR." - MMC Ventures

16

Many people ask me "How will Parallel Studios' AI Wayfinder (https://www.wayfinder.ai/) agent and Bittensor work together?" ... Here's how: Given the capabilities of both Wayfinder and Bittensor, there are several intriguing ways these two technologies could potentially collaborate or integrate. Here are five possibilities: 1. AI-Driven Subnet Management: Wayfinder's AI agents (Shells) could be used to autonomously manage and optimize Bittensor subnets. These agents could analyze subnet performance, adjust staking strategies, and even propose new subnet configurations based on real-time network data and the upcoming dTAO (Dynamic TAO) upgrade. 2. Cross-Chain Intelligence Marketplace: Leveraging Wayfinder's omni-chain functionality and Bittensor's decentralized intelligence market, a cross-chain AI services platform could be created. This would allow AI models on Bittensor to be accessed and utilized across multiple blockchains through Wayfinder's interface, expanding the reach and applicability of Bittensor's AI capabilities. 3. Enhanced Oracle Services: Bittensor's focus on creating decentralized AI oracles could be augmented by Wayfinder's AI agents. These agents could act as intelligent intermediaries, gathering, verifying, and delivering data from various sources to Bittensor's network, improving the accuracy and reliability of AI-driven oracle services. 4. Automated TAO Token Management: With the introduction of dTAO, Wayfinder's AI agents could be programmed to autonomously manage TAO token holdings. They could optimize staking, participate in governance decisions, and even engage in yield farming strategies across different subnets based on predefined parameters and market conditions. 5. AI-Powered Subnet Creation and Optimization: Wayfinder's AI could be used to analyze the Bittensor network and automatically propose and create new subnets that fill gaps in the ecosystem. These AI agents could continuously monitor subnet performance, suggest improvements, and even help implement changes through the RAO net upgrade, ensuring the Bittensor network remains efficient and innovative. These integrations could potentially create a powerful synergy between Wayfinder's AI-driven blockchain navigation capabilities and Bittensor's decentralized machine learning network, fostering innovation in both AI and blockchain technologies.

3

🎙️ Making Spot Spray Technology Accessible With Niqo Robotics | Future of Agriculture with Tim Hammerich 🎙️ "We need to stop talking about AI and focus on who is servicing it, what is the downtime like, and how many units are out there. I think that commercial narrative is finally beginning to come in." - Jaisimha Rao, Founder of Niqo Robotics. From 0 to 50 autonomous sprayers in India, Niqo's robots have covered 120,000 acres and served 2,000+ farmers. A decade in the making, they're redefining the future of ag automation. 👂🏽https://bit.ly/3OjbqgK #AI #NiqoRobotics #Agriculture #AgTech #FutureofAgriculture #Sustainability

54

2 Comments

🚀 v0.5.0 is here, and it’s a big one! We’re thrilled to announce that fully deterministic replay/resimulation is now implemented. This has been a cornerstone of Copper’s vision since day one. If you replay a log in Copper using the same code, you’ll always get the exact same results. 🎯 This may sound straightforward, but surprisingly, many robotics runtimes still struggle with this level of determinism. For roboticians, it’s a game-changer—enabling the creation of large, reliable test datasets with complete confidence in the replay. What’s even more exciting? Hooking up log replay with our simulation interface turned out to be quite seamless. And just like that, 🎉 deterministic resimulation is here! This milestone reflects Copper’s long-term vision coming to life. We didn’t stop there—v0.5.0 also includes a nifty sensor fusion code generation and a simplified API to build your tasks. Check it out! #robots #determinism #copper #rust https://lnkd.in/gN73BUDN

6

🚀 Fantastic to celebrate the start of #WorldSpaceWeek yesterday with an evening of insightful debate on the New Space Economy courtesy of #Codex, the Institute of Physics and a distinguished panel of #spacetech entrepreneurs and investors – Katie King, PhD, Nick Shave FRAeS, Andrew Bacon, Mark Boggett, Sean Sutcliffe & Sanjeev Gordhan, ably moderated by Joanna Hart. It’s hard not to be excited and inspired by the opportunities being unlocked both in space and on earth by developments in spacetech and the new space economy. Lower launch costs – thanks largely to SpaceX – combined with the prospect of lower return costs too, are opening up a whole new world of commercial possibilities from in-orbit manufacturing of pharmaceuticals (BioOrbit) and semiconductors (Space Forge), through on-orbit servicing and debris removal (Astroscale), to solar farms and datacenters in space. 💫 Precisely because the scale of the opportunity is so exciting, my attention was drawn to a couple of important themes underlying yesterday’s debate. These are recurring threads that many in the broader UK tech sector will recognise but they are magnified in the context of the new space economy: 1.    A record number of UK space start-ups are being founded every quarter, but a lack of growth capital constrains their progress commercially – what needs to change so that more of these companies can fulfil their potential and grow into global champions from a UK base? 2.    Skills shortages threaten to constrain growth in the new space economy. Whilst there are specific shortages now – Katie highlighted the difficulty filling roles that require both biotech and space engineering expertise, for example – we will need many more physicists, mathematicians, computer scientists and engineers of all descriptions to deliver on the promise of the new space economy. Against this backdrop, it was shocking to hear from Institute of Physics’ Group CEO Tom Grinyer that some 400 UK secondary schools don’t currently have a dedicated physics teacher. What more can we collectively do to address these issues now so that we can all benefit from the advances in spacetech tomorrow? 🚀

21

DeepSeek-V2.5 Released by DeepSeek-AI: A Cutting-Edge 238B Parameter Model Featuring Mixture of Experts (MoE) with 160 Experts, Advanced Chat, Coding, and 128k Context Length Capabilities Read our full take on this: https://lnkd.in/g4MdekFs DeepSeek-AI has released DeepSeek-V2.5, a powerful Mixture of Experts (MOE) model with 238 billion parameters, featuring 160 experts and 16 billion active parameters for optimized performance. The model excels in chat and coding tasks, with cutting-edge capabilities such as function calls, JSON output generation, and Fill-in-the-Middle (FIM) completion. With an impressive 128k context length, DeepSeek-V2.5 is designed to easily handle extensive, complex inputs, pushing the boundaries of AI-driven solutions. This upgraded version combines two of its previous models: DeepSeekV2-Chat and DeepSeek-Coder-V2-Instruct. The new release promises an improved user experience, enhanced coding abilities, and better alignment with human preferences. Key Features of DeepSeek-V2.5 🔰 Improved Alignment with Human Preferences: One of DeepSeek-V2.5’s primary focuses is better aligning with human preferences. This means the model has been optimized to follow instructions more accurately and provide more relevant and coherent responses. This improvement is especially crucial for businesses and developers who require reliable AI solutions that can adapt to specific demands with minimal intervention. 🔰 Enhanced Writing and Instruction Following: DeepSeek-V2.5 offers improvements in writing, generating more natural-sounding text and following complex instructions more efficiently than previous versions. Whether used in chat-based interfaces or for generating extensive coding instructions, this model provides users with a robust AI solution that can easily handle various tasks. 🔰 Optimized Inference Requirements: Running DeepSeek-V2.5 locally requires significant computational resources, as the model utilizes 236 billion parameters in BF16 format, demanding 80GB*8 GPUs. However, the model offers high performance with impressive speed and accuracy for those with the necessary hardware. For users who lack access to such advanced setups, DeepSeek-V2.5 can also be run via Hugging Face’s Transformers or vLLM, both of which offer cloud-based inference solutions. Model: https://lnkd.in/g84ADnNV

33

🔹 The Challenge for Modern Users: Disaggregated Compute and DAGs Today, many users face persistent issues in orchestrating DAGs (Directed Acyclic Graphs) and managing disaggregated compute: 1️⃣ Disconnected Experimentation: Experimentation often happens in silos on separate frameworks, with DAG orchestration as an afterthought. This makes scaling from experimentation to production tedious and error-prone. 2️⃣ Painful Debugging: Debugging across disaggregated workflows is challenging, as it’s difficult to pinpoint issues. Self-service is complicated, with users frequently hitting roadblocks. 3️⃣ Cost Control Challenges: Costs quickly escalate, as isolated systems can’t leverage shared contexts to optimize effectively. These challenges drove us to create Flyte and Union — but rather than attempting to unify everything on a single framework, our solution is unique. It enables “best-of-breed” tools to work seamlessly within a hyper-converged orchestration system. Here’s what it delivers: 🌟 Rapid Product Development: Teams have seen timelines shrink from quarters to days. 🌟 Unified User Experience: A single destination simplifies onboarding and self-service. 🌟 Cost Efficiency: Ephemeral infrastructure is provisioned just-in-time, enabling “magical autoscaling” without relying on serverless environments. Even on-prem clusters can scale dynamically by oversubscribing resources across teams. 🔹 Case Study: Flyte & Wayve Wayve leveraged Flyte to achieve: • scalability 🤯 • reliability 👏 • Single-click execution over manual interventions 😍 • Weeks to days time savings ⌛ • Dramatic cost reductions using spot instances 💵 Proud to see the impact of Flyte and Union in transforming workflows and efficiency for teams around the world.

26

Stability AI have introduced significant improvements through the development of Scaling Rectified Flow Transformers (RFTs) a key component of the StableDiffusion 3 architecture. This research enhances generative models for text-to-image synthesis by leveraging rectified flow models, which connect data and noise in a straight line, offering better theoretical properties, prompt stability, and conceptual simplicity compared to traditional diffusion models.   The team shows improved noise sampling techniques for training rectified flow models, biased towards perceptually relevant scales, which demonstrate superior performance over established diffusion formulations. A novel diffusion-transformer based architecture is presented, using separate weights for image and text modalities, enabling a bidirectional flow of information. The backbone, MM-DiT, is designed to handle different domains using separate sets of trainable model weights, outperforming existing transformer-based diffusion backbones.   The study also showcases the scalability of this architecture, with predictable scaling trends correlating lower validation loss with improved text-to-image synthesis. The largest models outperform current state-of-the-art models and improves upon text comprehension, typography, and overall image quality, as confirmed by human preference ratings.  Arxiv: https://lnkd.in/e7dhHAt8

4

1 Comment

Our humanoid robot has come far since its launch two years ago - find out about all the newest features, biggest milestones and perks of becoming part of our community in this summary of pib's birthday party. Thanks RobotsBlog for featuring us in your blog! 👉 Read the full article here: https://lnkd.in/dNHynQhj #pibrocks #robotics #OpenSource #humanoid #AI #HumanMachineInteraction

15

2 Comments

Designing and deploying robots is characterised by seeming contradictions.🤖 There's not enough multi-modal data at design time to train AI models for robotics, but when robots are deployed there's too much data coming in too fast from many different sensors. Hardware costs have declined significantly but still not enough. Robots have been to all sorts of exotic locations such as space or mines or oceans but have yet to make it in a big way to the final frontier: our homes. Advika Jalan and Alexandra Reinert explore the technological developments that are helping overcome a number of robotics challenges in the first part of our series on robotics, where we answer a key question: Why is robotics an interesting area to invest in now? Read part I now 👇 https://lnkd.in/e_BnfX-h #robotics #AI #startups

64

12 Comments

NEW RETHINKX INSIGHT 🤖 How much should we invest in humanoid robotics? What are the expected returns on investment? What is the cost of inaction? Explore our latest blog and robotics ROI calculator 🚀 https://lnkd.in/d6CgD4fa

21

2 Comments

Robotics and AI will touch all aspects of life and society. This one “could” be a path in reducing or eliminating the herbicides used on crops. On the technical side, from the article, this is astounding. Who would have thought even 10 years ago GPUs would be utilized in this manner. “Powered by 24 NVIDIA GPUs, Carbon Robotics’ vehicles use deep-learning-based computer vision models to autonomously identify and eliminate weeds via CO2 lasers. LaserWeeder, as the product is called, processes 4.7 million high-resolution images per hour, zapping 5,000 weeds per minute, according to the company. To date, its dataset includes 25 million labeled plants and more than 30,000 crop and weed models” #AI #agritech #robotics #nvidia

3

1 Comment

What if you could cut product design engineering cycles from days to seconds? Physics AI is revolutionizing engineering design with real-time simulation feedback. Imagine getting instant performance insights from just a 3D model - no meshing, no complex setup. Our latest blog explores: - How Physics AI is transforming product development - The secret behind training these powerful AI models - Our collaboration with NVIDIA for real-time virtual wind tunnels - Why this matters for every engineer and designer 🎯 Bonus: See how we're working with NVIDIA to tackle the biggest challenge in Physics AI. 👀 Read the full story: https://lnkd.in/gu7xhXeb

39

In the most recent Bg2 Pod episode Brad Gerstner discusses his interview with Jensen Huang. When asked, Jensen agreed that the #AI inference market is going to be many orders of magnitude larger than the training market and that #NVIDIA is well positioned for the inference market as well. He stated that while the major training deployments will switch over the latest #Blackwell based systems, the older systems will be repurposed for inference. The key insight is that there is still a lot of performance that is relevant to memory bound inference workloads that can be unlocked in #CUDA and other parts of the software stack. Case in point is the following work by Neural Magic where they have used mixed-input quantization to boost inference of the largest frontier models. There will be a lot of such work coming up in the future as well. https://lnkd.in/gW-rRjSR

2

What a time to be alive. Yes, LLMs are reaching initial plateaus. But then again, the underlying compute is getting 2-3x cheaper and more energy efficient every year. Our models today, are around 40.000x more compute efficient than comparable numerical models. Already today, we can build with 5M or 10M investment, what used to take 200M+ just in infrastructure before. In one year, we can build the same for 3M. In two years for 1.5M. In three years for 0.75M (or less). In other words, a global, high resolution, high precision world model, that would have taken billions to build just 5 years ago, will be built with a few million of investment in the next 5 years. This is our exponential journey towards climate sentience.

48

4 Comments

Hey, robotics developers! Did you know a new CAD-based toolkit for robotic motion planning, modeling, and execution is in town? We are talking about the SwRI Workbench for Offline Robotics Development (SWORD). Learn more about it at the link. https://lnkd.in/e6_Z_F45. #RoboticsDeveloper #RobotDesign #EngineeringExcellence #RoboticsInnovation #IndustrialAutomation

1