Gene D.

Sometimes robots need to ask operators for help when obstacles are in their way, a process we call “an assist”. With our new Remote Assist Clearing feature, robot operators can now resolve assists directly from BrainOS Mobile by simply clicking “Resume Route” in the app! If the obstacle (e.g., a grocery cart) that was temporarily blocking its path has been moved, the robot will safely resume its operation at your command, allowing the operator to quickly return to their other tasks. #Robots #AI #CleaningRobots #MobileApp

3

RAG (Retrieval Augmented Generation) is very cool and a great early use case for LLMs. It makes it easy to give AI agents access to your own data and opens a ton of possibilities. At the same time it feels like the pre Web 2.0 days when Ajax was a novel new technology and the use cases were simple things like auto-suggest bolted on to the same old UIs. The big shift will happen when people start adopting an "Agent First" approach, where AI is the central orchestrator for applications, not just another API in the toolkit. There are still a lot of roadblocks like unit economics, security, tooling, etc, but those will be solved in time. It's really exciting because even though many of us are convinced this is the future, not one person actually knows what it's going to look like in 5 years. It's a great time to be a builder.

12

1 Comment

Performance is key! As we cannot share the detailed work with (former) customers of Rockset (acquired by OpenAI), we decided to implement the Rockbench benchmark to check how streaming ingest between CrateDB and Rockset (acquired by OpenAI) compares. The results are very insightful. We see: ✅ CrateDB achieves 6-9x lower latencies than Rockset for streaming ingest. ✅ CrateDB outperforms Rockset on the same hardware while saving about 20% costs. ✅ When volumes increase, the latency increased linearly in Rockset, while remaining mostly flat in CrateDB. As a plus: 1️⃣ Achieve full high availability due to the distributed nature of CrateDB (benchmark results are achieved including replication in CrateDB) 2️⃣ Achieve higher throughput on the same amount of CPUs as Rockset (at 50% of memory needs) 3️⃣ Full open-source solution that prevents you from market shifts On side note: If you are thinking to switch to Elastic/OpenSearch, this benchmark has been created to compare Rockset vs. Elastic, so CrateDB is also faster than Elastic for streaming ingest cases! Read the full comparison here: https://lnkd.in/d_nEMDkx

29

1 Comment

Did you know that this company that has surpassed Apple was once near bankruptcy? The story of NVIDIA's ascent is one for the ages: 🔎 In 1993, Nvidia was founded by dreamers Jensen Huang, Curtis Priem, and Chris Malachowsky. Their quixotic quest? Crafting processors to supercharge video games. But their 1995 chip NV1 was a brutal reality check - a flop that nearly bankrupted them. 🔎 Salvation arrived in 1997 with NV3. Outperforming rivals, it landed Nvidia deals with PC titans Dell Technologies and Gateway. They'd been granted a seat at the 3D graphics table. 🔎 1999's GeForce 256 GPU was Nvidia's big bang. Bringing massive parallel processing to PCs, it revolutionized gaming visuals overnight. The world took notice of these graphics gurus.  🔎 Nvidia charted its own path, unconventionally championing gaming as the next cultural wave before it crested. Jensen Huang's foresight cemented their iconic brand. 🔎 Restlessly innovating, they expanded GPUs to mobile, automotive, and kickstarted the AI/deep learning era. Nvidia emerged as a multidimensional computing force. 🔎 Today's climax? Nvidia's stock skyrocketed past $AAPL to become the 2nd largest $SPX component at $1164 - an all-time high with a 1:10 split looming. The former underdog has outgrown the giants. 🔎 From near-bankruptcy to industry dominance, NVIDIA's rise is a GOAT tale of perseverance, vision, and constant reinvention. They dreamed of changing graphics computing forever, and delivered that in spades.

11

Meta 's relentless pursuit of "compute" ... by "the end of 2024, we’re aiming to continue to grow our infrastructure ... 350,000 NVIDIA H100 GPUs". Total Power Consumption = 350,000 * 700W = 245 MW For every 2W, you need 1W for HVAC so est total 370 MW. That is equivalent energy to power the whole city of San Francisco. src: https://lnkd.in/gUP49eFZ Supplemental: The average electricity customer in San Francisco, CA uses 729 kWh of electricity per month, and 8,748 kWh over the course of the year. Given that there are approximately 361,222 households in San Francisco, we can calculate the total power consumption for the city 361 Megawatts of power.

9

3 Comments

I had a great discussion with Abi Noda on his Engineering Enablement podcast! Among other things, we dove deep into how our Platform Engineering team at SiriusXM prioritizes projects to boost developer speed and align with business objectives. Listen to the full episode here: https://lnkd.in/eN57Hync

34

What the heck is an LLMOS? Andrej Karpathy, co-founder of OpenAI and former AI head at Tesla, has been talking about an LLMOS -- Large Language Model OS. A new kind of operating system in this AI world. The RAM is replaced with a context window. The CPU is the LLM itself. And the instructions are the prompt. Interesting, but I don't see it that way. An operating system classically manages software/hardware interactions and allocates hardware resources. An LLM OS doesn't really do any of that. It is a nice analogy however. But I think the more accurate description would be it's a new development paradigm and platform. It allows you to program new interactions with a new model (context window, tools/providers, prompts, and the execution engine is the LLM). Thoughts?

17

2 Comments

A fantastic use of AI. Practical and seamless.

19

3 Comments

AI giant NVIDIA founder & CEO Jensen Huang on 60 Minutes last week: “If you want to do extraordinary things, it shouldn't be easy”. It sure wasn't easy for Jensen. His story is a masterclass in how bold bets, technical brilliance, and relentless drive can take a company from underdog to industry titan. As a 30-year-old immigrant working as an electrical engineer, Huang and two friends hatched the idea for a new kind of graphics chip in a San Jose Denny's in 1993. It nearly bankrupted them. But they pivoted, invented the GPU, and revolutionized video game graphics. And gaming was just the beginning. Huang envisioned using GPUs - which can process a myriad of complex calculations simultaneously - to power the next frontier: artificial intelligence. It was a risky bet, but researchers soon discovered Nvidia's architecture was perfect for the deep learning algorithms underlying modern AI. Huang then went all-in, developing software and hardware for a new class of GPU-driven AI supercomputers. Tech giants and startups alike began using them for everything from drug discovery to robot design to weather prediction. Today, Nvidia's chips power ChatGPT, self-driving cars, and countless other AI breakthroughs. As AI exploded, so did Nvidia's value - shooting from $1 trillion to $2 trillion in just 8 months 😳 Despite being a celebrity CEO now, Huang remains laser-focused on pushing the boundaries of what's possible. Under his intense (some say demanding) leadership, Nvidia is developing ever-faster and smarter AI chips to power use cases we can hardly imagine today. And it all started at Denny's. Phenomenal.

6

In the realm of RAG, there is no one-size-fits-all solution. Building a robust Retrieval-Augmented Generation pipeline is inherently complex, requiring expertise across various domains. The challenge amplifies when using open-source components, as deploying a RAG system in production demands stringent data protection while ensuring scalability and maintenance. Nuclia • AI Search & RAG as a Service's customisable RAG aims to address these challenges by allowing you to fine-tune every aspect of the RAG process—from data ingestion to chunking and retrieval strategies—without the hassle of piecing everything together. This seamless integration facilitates bringing your tailored RAG pipeline to production swiftly and securely. Try yourself at https://nuclia.com

16

Meta Introduces Chameleon: A State-of-the-Art Multimodal Model 🦎 Meta has unveiled Chameleon, a groundbreaking multimodal model that pushes the boundaries of AI capabilities. Chameleon seamlessly integrates vision and language, enabling it to understand and generate images and text with remarkable accuracy. Key Takeaways: 🎨 Chameleon excels in both image captioning and text-to-image generation, showcasing its versatility across multiple tasks. 🧠 The model leverages a novel architecture called Mixture-of-Attention (MoA), which efficiently processes and combines visual and textual information. 🌐 Chameleon outperforms existing state-of-the-art models in zero-shot image captioning and text-to-image generation on various benchmarks. 🔍 The model's ability to handle complex, compositional prompts sets it apart from other multimodal models. 🔒 Meta has open-sourced Chameleon's code and trained weights, fostering transparency and enabling further research in the field. Chameleon represents a significant milestone in the development of multimodal AI systems. Its ability to understand and generate both images and text opens up exciting possibilities for applications in fields such as creative content generation, assistive technologies, and more. Read the full article to learn more about Chameleon and its implications for the future of AI: https://lnkd.in/gk_ZTZdR #chameleon #multimodalai #artificialintelligence #ai #machinelearning #ml

5

Live2Diff: Real-Time AI Video Translation 🚀 Live2Diff, a pioneering model that transforms the way we process live video streams. With its unique uni-directional attention mechanism and highly efficient processing pipeline, Live2Diff achieves unprecedented temporal consistency and performance. Key Highlights: 1. Uni-directional Temporal Attention: Ensuring smooth and consistent video streams. 2. High Efficiency: Achieves 16FPS for 512 × 512 videos on an RTX 4090 GPU. 3. Structural Consistency: Maintains the integrity of video streams with a lightweight depth prior. Why this matters? 1. Enhanced Efficiency: Build on an optimised pipeline for faster processing. 2. Collaborative Improvement: Innovate further with community contributions. 3. Broader Accessibility: Implement cutting-edge technology without high costs. Read the full paper here: https://lnkd.in/eySY8PZM #RealTimeAI #VideoProcessing #OpenSourceAI #Innovation #AIResearch

3

MRjs v0.6.3 is out !!! 🎉 **The Highlights:** animation, animation, animation 🎦 - Added a lot of fine tuning for animation on models including looping, end state, and replayability.  - Speed improvements and optimization fixes for a missed memory leak and infinite loop - Some fixes for entities and models related to physics, selection, and the collider interactions Check out the rest of the update on github (https://lnkd.in/g-h3BqsX) and find the release on npm (https://lnkd.in/ghvXvtk3)

9

2 Comments

Which is the best #GenerativeAI model to use? Adobe just answered this question! In their recent video, Adobe showcased how you can generate and extend your videos in PremierPro, using the built-in models and other models like OpenAI #Sora, Pika or Runway. This seems to be a plugin based way to extend and choose the right models for your video. And this definitely is the best way forward when it comes to GenerativeAI models across text, image, video etc. It's too early to decide the best models or platforms to invest in. But at the same time, its important to experiment and implement #GenAI projects to stay competitive. Keep your workflow, data, prompts and business logic separate from your models by building abstraction layers. This way, it will be cheaper to switch to newer and better models as they arise. What strategies do you employ to #futureproof your #AI implementations?

14

The Memory Stick Story That Unlocked Innovation Remember those big chunky USB memory sticks back in the day? The ones that looked more like a brick than a portable storage device? 64MB or 128MB of storage packed into some clunky plastic casing that made zero sense for actual portability. Until some innovative folks had an "aha!" moment - why the hell does a memory stick need to be so damn big? Cue the era of sleek little USB drives you could dangle off a keychain while still cramming in tons of storage. Smaller form, same core function - genius! That pivotal rethink on the basic memory stick design is a blueprint for the innovative mindset private label brands desperately need today. True innovation isn't just making an existing product incrementally better. It's fundamentally rethinking the core utility and reimagining how to deliver it in a fresh, defensible way. With USB drives, separating the storage from that bulky casing unlocked new IP across the design, functionality, user experience - the whole nine yards. Turning a commodity into a product line you could own and protect. So I'm challenging all of you - look at your product categories through this innovative lens. Ask yourself: What basic utility does my product provide? Can I re-architect how that utility gets delivered in a novel form? What new IP avenues could emerge from that fresh approach? I know, I know - it's not easy work. But doing the heavy lifting of true innovation sets you up to: Own defensible IP that deters copycats Capture untapped market segments Forge a unique, hard-to-replicate brand identity If something as basic as a USB memory stick could spark game-changing innovation, just imagine the opportunities across any product category. Innovation is the most sustainable edge for us private labels battling those multinational monster brands. But it all starts with rethinking the fundamentals of your product. So what's your portable USB drive waiting to be reinvented and disrupted? What core utilities are you reimagining in an innovative way? Drop some ideas or examples in the comments - I'm really keen to see everyone's innovative product thinking! ---------------‐‐ And who am I to wax poetic about innovation? Just a dude with an unhealthy obsession for reinventing the wheel...literally. My latest brilliant idea? The CircleTire™ - tires reimagined as giant circular rubber rings for a smoother, more energy-efficient ride. Doctors say the head trauma from all the facepalms is treatable, so I've got that going for me! If you want to join me in a tragically hip venture of trying to spark innovation in the stodgy old private label world, give me a follow and slide into my DMs. I'll hook you up with some hot new CircleTire prototypes...they roll real nice, I swear! Now food for the algorithm: #Innovation #ProductDesign #PrivateLabel #NewProductDevelopment #Entrepreneurship

2

This week Jam hosted 15 DevRel for an off-the-record dinner Here's 7 quick bites about what we learned about The Art of DevRel: 1. The best DevRels work closely with engineering and product. It’s not just about sharing what they're building with devs, it’s also about sharing dev feedback with product. When you get that feedback loop right, that’s how you get magical developer experiences. 2. DevRel should help the entire engineering team do DevRel Like how HR enables any manager to hire, DevRel should enable any engineer to write blogs, create tutorials, give talks, etc. Swyx’s “DevRel as a Service” post was big topic of discussion— will link it in the comments. 3. No 2 companies do devrel the same. No 2 people do devrel the same way There is a lot of art here that’s not science. It’s about finding the right fit person for your exact company Like how there is founder market fit, there is a devrel company fit too 4. Reporting up was a huge point of discussion The best devrels over-communicate internally in the company what is happening on their side Communication creates trust, and trust creates freedom for devrel to do things that don’t scale that builds real community (and results). 5. DevRel is always on; you are always representing the company Community never takes time off, devs around the world are building nights, weekends and holidays You have to love helping people build - truly - to love devrel (and it’s so inspiring to meet people who do) 6. The best DevRel build genuine relationships They become friends with community members and devs The best community and relationships aren’t measurable But can have immeasurable impact on the trajectory of a product/company 7. Last one: Are developers on LinkedIn? Consensus: yes - yes, they are.

20

3 Comments