Katherine Scott

Following up on the recent post, let's explore the QuestBuilder - a smart quiz creation and discovery solution powered by generative AI and machine learning techniques. How Does It Work? - AssemblyAI processes the video to detect chapters and generate transcripts. - FFmpeg splits the video into chapters, and OpenCV extracts the clearest still images. - Amazon Rekognition analyses the video for additional context, such as objects, scenes, and sentiment. - The video stills, transcripts, and insights are combined with the Da Vinci Quests Knowledge Base to form a comprehensive context and prompt message. - Anthropic Claude, integrated via the Amazon Bedrock API, takes this rich context to generate engaging quiz questions and answers. - A rigorous quality assurance process ensures the accuracy and relevance of the generated quizzes. - Amazon Translate ensures the quizzes are accessible in various languages. The Impact QuestBuilder streamlines the creation of interactive learning experiences by automating a time-consuming process. This allows us to focus on delivering high-quality educational content and enhancing the learning journey for our users. Previous post: https://lnkd.in/ekPAgQdN Stay Tuned for More! I'll be sharing more details about QuestBuilder's capabilities and its impact on our products in the coming weeks. #QuestBuilder #GenerativeAI #MachineLearning #Speech2Text #Image2Text #VideoAnalyse #ImageProcessing #Education #Innovation #DaVinciApp #AWS #SageMaker #Bedrock

6

In a June 10, 2024, keynote presentation at #Apple’s World Wide Developers Conference, Ted Kremenek, Apple's director of languages and runtimes, argued that #Swift is the best programming language to replace #C++. Swift was introduced by Apple in 2014 as the successor to Objective-C, and Kremenek believes it will overtake C++ as well. “Swift’s safety, speed, and approachability, combined with built-in C and C++ interoperability, mean Swift is the best choice to succeed C++,” Kremenek said. C++ and C have drawn fire from the White House, which released a report in February urging developers to move to memory-safe programming languages. Apple is committed to adopting Swift in its own C++ code bases, Kremenek said. Prior to Swift, software on Apple’s devices was written using C, C++, and Objective-C. Apple has announced plans to migrate Swift to a dedicated GitHub organization, github.com/swiftlang. Swift 6, which is planned for release this year, promises to make concurrent programming safer and easier through full data race safety. A data race happens when different parts of code try to modify and access the same data at the same time. “Swift 6 eliminates these kinds of bugs by diagnosing them at compile time,” Kremenek said. Also planned for Swift 6 is expanded #Linux support, covering the Debian and Fedora Linux distributions, and improved support for #Windows. Generics also are eyed for improvement in Swift 6, with a new subset planned for targeting constrained environments such as OS kernels and microcontrollers. With the creation of github.com/swiftlang, Apple believes it is creating an even more conducive environment for collaboration and innovation. #Programming #Technology #WWDC #Github https://lnkd.in/gQHG7Pxr

1

We present ClutterGen that tackles two important research problems in robotics: 1) how to generate physically compliant cluttered scene for robot learning? 2) how to stably place objects instead of the current "pick and drop" paradigm? Our key idea is to formulate physically compliant cluttered scene generation as a sequential object placement problem, such that we can look at scene generation as a reinforcement learning problem. The learned environment generator can be easily applied to solve truly "pick and place" tasks. Our robot can figure out how to place an object in a stable manner into a cluttered environment. Please check out our project website with video, code, and paper to learn more! https://lnkd.in/ecswQ9yg Great work done w/ Yinsen Jia at Duke General Robotics Lab (https://lnkd.in/g_4JhxV2)

36

Another great post by Dr. Lorenzo Nicoletti explaining some of the simple concepts of Physics, since Inertia is a property of a shape that determines the "degree of resistance to change" the object with the largest Inertia accelerates the slowest. Nice way to visualize that!

15

3 Comments

Using ROS? Need real-time processing on Linux? Take a look at the cactus-rt framework https://lnkd.in/dc7a5c6a The blog posts about real-time on Linux by Shuhao Wu linked there provide many insights - even for people already experienced in the topic

187

1 Comment

We are running the next ASTRA-sim tutorial at MICRO 2024! In this tutorial we will introduce MLCommons Chakra - a distributed ML workload schema that we have been collaboratively developing since the last 1+ year. We will present mechanisms to collect execution traces from real multi-node AI clusters and feed them to ASTRA-sim to study the impact of next generation compute and/or network fabrics. We will also introduce recent updates to ASTRA-sim: including (1) Chakra-ASTRA-sim interface, (2) the ability to evaluate *custom* collective algorithms on diverse network fabrics, and (3) study LLM-specific optimizations. Resources: - Tutorial Website: https://lnkd.in/eGy9Gs-z - ASTRA-sim: https://lnkd.in/eX-2KzJN - Chakra: https://lnkd.in/e968Xsga Presenters: Tushar Krishna (GT) William Won (GT) Joongun Park (GT) Taekyung Heo (NVIDIA) Vinay Ramakrishnaiah (AMD)

67

1 Comment

Interesting insights on building robotic simulation in NVidia Omniverse USD specifically IsaacSim. Lot of concepts of AutoCAD combined with 3D viewing development, lighting and texture. Interestingly the more core custom development uses C++ and python coding platforms. https://lnkd.in/gx4P9Qv9

10

Kudos to #Composable_kernel team and please check out the AMD engineer Haocong Wang presents the ROCm Composable Kernel library which allows to write high-performance kernels for machine learning with HIP C++. #Composable_kernel #ROCm

39

1 Comment

A key message for teachers attending the Coding and Robotics Education (CARE) program at S4 Integration, was the necessity to contextualise the content with real world examples wherever possible. #Rocksi (https://rocksi.net/) a free robot simulator that runs in any modern browser and allows users to program a 3D seven axis robot arm using blockly visual programming was used to demonstrate this very clearly. Often teachers refer to the construction of cars using robots and may show a picture or video without realising that most learners have little to no idea of what is a simple robot. Rocksi provides for a constructivist approach to engaging learners in new and exciting ways developing a broader range of skills especially in a three-dimensional world while manipulating variables and seeing the outputs. The teachers will be tasked in a future training session where their challenge will be to assemble a stack of blocks as their assessment. I have little doubt that they will be totally engaged and trust that their excitement will be transferred into the classroom. Thanks to Ettienne Gerwel for arranging the training at the Wirk premises in Gqeberha. #Rocksi #Coding #Robotics #TeacherTraining #YESA #stem

30

9 Comments

🤖 Transform Your Robot with Advanced Object Detection! Are you diving into the world of robotics and grappling with object detection challenges? We've got just the breakthrough you need! Discover how to elevate your robot's perception capabilities using custom ROS messages. Our latest blog post is a treasure trove of insights and step-by-step guidance on publishing your object detection results over ROS, making your robot smarter and more responsive. Why is this a game-changer? 🚀 >> Seamless Integration: Learn the nuts and bolts of using custom ROS messages tailored to your project's needs. >> Enhanced Performance: Elevate your robot's object detection capabilities to new heights with practical, easy-to-implement strategies. >> Community and Support: Join a thriving community of innovators, and leverage the collective wisdom of experts in robotics. >> Dive into the blog post here: Custom ROS Messages for Object Detection Let's push the boundaries of what our robots can do. Share your thoughts and experiences with object detection and ROS in the comments below! 👇 #ROS #Robotics #ObjectDetection #MathWorks #Innovation #TechCommunity #matlab

3

We're living in incredibly exciting times as far as ML/AI is concerned with NVIDIA's latest launch, which includes the release of Nemotron, fine-tuned on LLaMA 3.1 70B. This has been making waves, garnering significant attention due to its phenomenal capabilities for synthetic data generation. In today's AI landscape, foundation models have become a commodity. The true competitive edge lies in the developer experience and the innovative applications built on top of these models. NVIDIA's ambition is clear: they want to dominate at every level of AI product engineering and management, offering not just powerful models but also the ecosystem for developers to unlock their full potential. This is where the future of AI innovation will flourish, and NVIDIA aims to be at the heart of it. https://lnkd.in/gEM5XAjH

6

🌭🌭HOP: Hand-Object interaction Pretraining🌭🌭 👉Berkeley unveils a novel approach to learn general robot manipulation priors from 3D hand-object interaction trajectories. Using in-the-wild videos to generate sensorimotor robot trajectories. 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅General manipulation prior from in-the-wild ✅Simulator as videos-sensorimotor intermediary ✅Fine-tuning via RL and behavior cloning (BC) ✅Sample-efficient adaptation to downstream tasks #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper https://lnkd.in/dhRFQara 👉Project https://lnkd.in/dCa77u7P

168

3 Comments

New paper published: Benchmarking the Parallel 1D Heat Equation Solver in Chapel, Charm++, C++, HPX, Go, Julia, Python, Rust, Swift, and Java https://lnkd.in/gCVufi_M

38

8 Comments

[The Fundamental Advantages of 3D Over 2D #2] Previous Post ( https://lnkd.in/gN_VWPx7 ), Before diving into this, it's necessary to categorize 3D virtual reality into two approaches: 3D on 2D: This refers to accessing 3D virtual reality through a 2D display, such as playing Roblox on a smartphone or Overwatch on a PC. 3D on 3D: This involves accessing 3D virtual reality through VR headsets like Meta Quest 3. Many doubt that 3D offers tangible benefits over 2D, particularly when comparing 3D experiences on 2D screens to standard 2D apps like Google or Amazon. However, the full advantages of 3D are most evident through VR devices, despite needing significant further development. For example, when Oculus (now Meta Quest) was first released, it was revolutionary, and subsequent devices like the Apple Vision Pro have continued to advance, though more progress is still needed. The comparison between 3D on 3D and 2D on 2D clearly shows distinct benefits for each. I believe 3D on 3D will fundamentally change our world, while 3D on 2D, due to current technological limits, remains somewhat constrained. In developing a web-based 3D engine, Adler has explored nearly all areas where 3D could be effectively utilized, finding that 3D on 2D holds advantages in specific industries and applications like digital twins, architecture, urban planning, gaming, and VTube viewing. Nevertheless, the utility of 3D on 2D over traditional 2D applications is generally limited except in specific contexts. Moreover, for 3D on 3D to truly find its place, further advancements in VR technology are necessary, addressing issues such as motion sickness, resolution, weight, processor performance, battery life, and cost—particularly challenges seen with high-end devices like the Apple Vision Pro. Addressing these hardware issues might take years, and the substantial investment required for 3D ventures poses significant challenges for investors and entrepreneurs, especially during times of high interest rates and economic downturns. Like during the dot-com bubble, terms like "metaverse" once garnered immense interest and investment. It’s important to clarify that not all virtual spaces need to be 3D to be part of the metaverse, as seen with platforms like Gather.town or even simpler concepts like Pokémon games. Yet, the common expectation ties the future of the metaverse closely to the development of VR, anticipating revolutionary 3D virtual worlds like Roblox and Zepeto. I am committed to engaging in the 3D virtual world sector with Adler for at least the next 30 years. Understanding and leveraging the fundamental advantages of 3D are crucial. Like the enduring giants of the web era, such as Amazon and Google, which survived by grasping the essence of their platform, the same holds true for the future of 3D virtual worlds. In my next post, I will delve deeper into what constitutes the essence of the 3D virtual world.

2

LUNARC has released 0.9.12 of our on-demand framework GfxLauncher, which powers the LUNARC HPC Desktop environment. The framework enables graphical applications (hardware and non-hardware accelerated) and notebooks to be executed through a job scheduler. New features: More fine-grained control of resource selection. Application scripts can use tags to specify launcher behavior. Command-line tools for resource usage, gfxusage and gfxnodes. https://lnkd.in/dhv8UGSQ

19

1 Comment