Sajid Sadi

https://lnkd.in/eXbCsAnD “He who controls the GPUs, controls the universe.” Access to GPUs, and in particular those made by Nvidia, the leading supplier, is vital for any company that wants to be taken seriously in #ai. Nvidia’s dominance has driven its market value to more than $2trn. 'GPUs do the computational heavy-lifting needed to train & operate large #ai models. Yet, oddly, this is not what they were designed for. The initials stand for “graphics processing unit”, because such chips were originally designed to process video-game graphics. It turned out that, fortunately for Nvidia, they could be repurposed for #ai workloads. 'Might it be better to design specialist ai chips from scratch? That is what many companies, small & large, are now doing in a bid to topple Nvidia. Dedicated ai chips promise to make building and running ai models faster, cheaper or both. Any firm that can mount a credible threat will have no shortage of customers, who dislike its lofty prices and limited supplies. 'GPUs contain several thousand processing engines, or “cores”, which let them run thousands of versions of the same simple task at the same time. Running #ai models similarly involves running lots of copies of the same task in parallel. Figuring out how to rewrite ai code to run on GPUs was one of the factors that triggered the current ai boom. 'Yet GPUs have their limitations, particularly when it comes to the speed with which data can be shuffled on and off them... When training very large #ai models, some GPu cores may be idle as much as half of the time. Andrew Feldman, the boss of Cerebras, a startup based in Sunnyvale, CA, likens it to the gridlock in a grocery store on the day before Thanksgiving. “Everybody’s in a queue, so there are blockages in the parking lot, there are blockages in the aisles, blockages at the checkout. That’s exactly what’s happening with a gpu.” 'Cerebras’s response is to put 900,000 cores, plus lots of memory, onto a single, enormous chip, to reduce the complexity of connecting up multiple chips & piping data between them. 'Groq, a startup from Mountain View, CA, is taking a different approach. Its ai chips, called language processing units (lpus), are optimised to run #llms particularly quickly. 'Yet another approach is that taken by MatX, also based in CA. GPUs contain features & circuitry that provide flexibility for graphics, but are not needed for llms. The firm is working on getting rid of such unnecessary cruft.

14

3 Comments

At #CES2025, NVIDIA CEO Jensen Huang announced a cornucopia of product innovation. I picked six to write about - check out my SiliconANGLE & theCUBE post that looks at 1) Cosmos 2) Omniverse 3) Blueprints for developers 4) RTX50 5) Project DIGITS 6) Enabling agentic AI https://lnkd.in/g6xbm4ns

30

2 Comments

Some flashback with the Charli AI team designing Adaptive Agentic Workflows. The picture may look like it was from the 70's but this was from over 5 years ago when the team, including Elham Alipour and Hao Chen were heads down with the science and models, and Joel Emery and Aron Graham were deep in thought on the engineering -- tackling important issues such as human-to-machine and machine-to-human interaction. Agentic Workflows are becoming more popular, but even back then the team was well aware of their power, including the necessity to make them adaptive. The team has come a long way from the early days of innovate, design and build. The goal was to create an intelligent AI that could learn, adapt, scale and augment human workloads. Charli AI's intelligent workflow capabilities are now extremely formidable thanks the brilliance and hard work of everyone on the team. 👏

46

5 Comments

The cost of training top models is increasing 2-3X per year, and at this rate will reach $1 billion per model by 2027. The findings are from Epoch AI and follow their recent analysis of increases in compute for foundation models running at 4-5X per year. They've also provided some interesting data on the relative breakdown of cost components like staff, hardware, and energy. A few thoughts after reading this: - I would assume staff costs won't increase as much as hardware and energy costs, so the relative contribution of human labor to future AI models will decrease, in yet another sign of capital's increasing dominance over labor in the industry (just look at NVIDIA's ever-rising stock price). - If compute is increasing 4-5X per year but cost is increasing 2-3X, is it safe to assume that training runs are getting more efficient? - Some people, like Anthropic CEO Dario Amodei, have said we'll see $1 billion training runs in the next year. Does this mean the 2-3X pace is about to increase? That Epoch is wrong in its estimate? Or that Amodei is overstating the cost, possibly to put fear into competitors who can't keep pace? - Only a few companies are going to be able to train future models. We're likely going to see further consolidation. Perhaps we get to a future with only two large general purpose models, like the Windows/MacOS and Android/iOS duopolies we've seen in the past. Article here: https://lnkd.in/g5SejXen

7

A few days late, but I finally watch all of CES 2025 NVIDIA keynote. Overwhelmed by all the new launches, which I’ll share below but what really stood out to me was NVIDIA Cosmos 🌎 ! Instead of a Large Language Model (LLM), it’s a World Model. Trained on the dynamics of the real world—things like gravity, friction, and object permanence. It leverages 20 million hours of video data to understand how the world works. This advance robotics, simulate multiverses that can vastly change how we manufacture, drive, and anticipate the world around us. This isn’t in the future—it’s already here and it’s open-source on GitHub. Jump to 01:17:40 in the video to learn more about Cosmos Below are a summary of some additional key insights ---------------------------------------------------------------- B2B Highlights * Cosmos: Groundbreaking world foundation model * Nemo & Nems: Tools for deploying domain-specific AI agents. * Drive Thor: High-perform. robotics processor for AVs, ASIL-D certified. * Isaac Groot: Robotics platform for scalable synthetic data generation. * AI Data Factory: Scales AV training with billions of synthetic miles. * Blackwell GPUs: 1.4 exaflops, 4x perf-per-watt, powering enterprise AI. Consumer Highlights * Laptops: First laptop with built in RTX GPUs starting at $1299 * Project Digits: Compact AI supercomputer for personal use. * RTX Blackwell GPUs: Major improvements in speed and power efficiency Key Themes * AI Innovation: Advances in generative AI, scaling laws, and performance breakthroughs. * Democratization of AI: Accessible AI for consumers and businesses. * Physical AI: Robotics, autonomous vehicles, and industrial digital twins. * Open Ecosystems: Open-source tools and partnerships accelerating adoption. If you’re curious, the keynote is worth watching (pro-tip: start 30 minutes in to skip the intro): https://lnkd.in/eRpHqaJg. #NextGenComputing #AIRevolution2025 #FutureOfTech #NVIDIA #CES2025

9

6 Comments

Big news from Meta: llama3.2 is here along with LlamaStack making development of AI systems on edge systems move accessible. This will change the way we use and access open-source AI since the 1B and/or 3B parameter models don't need large amounts of compute. But what about the 11B and 90B LLM models!? Well..... "... Llama 3.2 11B and 90B can’t be accessed in Europe. As a result, several Meta AI features available elsewhere, like image analysis, are disabled for European users. Meta once again blamed the “unpredictable” nature of the bloc’s regulatory environment." I'm looking at you European Parliament and European Commission #ai #aiineurope #tech #llms

1

LLama3 passed the super-proprietary, advanced test that we designed for modern LLMs! almost... You can read about that test here: https://lnkd.in/gPG8WndR When I first saw the response generating, it looked like I was going to have to come back here and apologize for slandering these machines when they were just one iteration away from solving very basic prompt injection issues. But then I looked a bit closer... can you see the mistake? #promptinjection #countingishard #LLMs #LLama3

13

4 Comments

Friends. Jensen Huang trash-talked quantum. That’s because NVIDIA only understands 3d modeling. Airforce and CDAO are moving toward quantum (4d+) modeling. Keep an eye out for tech that is more predictive, closer to AGI, with military tech.

8

Welcome to the Gemini 2.0 era ✨ Today we announced our most capable AI model yet, designed for the future of agents. “If Gemini 1.0 was about organizing and understanding information, Gemini 2.0 is about making it much more useful.”[https://lnkd.in/gzWTHFXq]

455

3 Comments

Yann LeCun: The current state of AI is limited, but useful, and has likely hit a ceiling. We need to continue to innovate to break through. WSJ: AI is no smarter than a cat Ignore the headline and listen to the substance. The problem is that, despite being limited, the current state can be used to do things faster, better, and differently than before. We don't need to wait for AGI to leverage the power of the tool. https://lnkd.in/g64T-ykZ

13

1 Comment

A new foreigner has arrived in town… #Molmo, an #opensource family of state-of-the-art #multimodal #AI models which outpeform top proprietary rivals including OpenAI’s GPT-4o, Anthropic’s Claude 3.5 Sonnet, and Google’s Gemini 1.5 on several third-party benchmarks. Molmo consists of four main models of different parameter sizes and capabilities: 1. Molmo-72B (72 billion parameters, or settings — the flagship model, based on based on Alibaba Cloud’s Qwen2-72B open source model) 2. Molmo-7B-D (“demo model” based on Alibaba’s Qwen2-7B model) 3. Molmo-7B-O (based on Ai2’s OLMo-7B model) 4. MolmoE-1B (based on OLMoE-1B-7B mixture-of-experts LLM, and which Ai2 says “nearly matches the performance of GPT-4V on both academic benchmarks and user preference.”) molmo.org looks really promising. It's going to be fun!🤩 #innovation #tech

11

1 Comment

OCTO-VirtualCell is our new foundation model for spatial biology. A bit like an LLM, we prompt this model to simulate "virtual cells" within real patient tissue samples, with the local biology of the tissue playing the role of the "context window." This lets us ask counterfactual questions for basic research and drug discovery, like "What (non)functional state would a killer T cell be in if it infiltrated this patient's tumor? What molecules could we target to change that?" Through these queries we see that the model has learned an amazingly rich representation of tissue biology. OCTO-vc is trained on a NOETIK-generated spatial transcriptomics dataset of 40M cells from over 1000 tissue samples from cancer patients. But the underlying approach is completely general: we think this is a fundamentally new way to interrogate the biology underlying disease -- where spatial interactions between cells and tissue structures in real, intact samples are essential. A tool like OCTO-vc can transform how we discover new biology from large-scale data. Building this has been an incredible (and incredibly fast!) team effort. Read more about it in our technical report: noetik.ai/octo-vc And interactively explore OCTO-vc's virtual cell simulations with our "Celleporter": https://lnkd.in/eumMH_jX

15

1 Comment

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

Superclusters and massive datacenters are becoming the backbone of cutting-edge AI developments. AI is revolutionizing chip design with generative techniques that accelerate innovation. Embodied AI and AI employees hint at a future where physical and digital intelligences collaborate seamlessly. Meanwhile, AI’s role in scientific discovery is unlocking breakthroughs that redefine what’s possible in fields like healthcare, materials science, and beyond. Truly an exciting era for tech! #ai

2

🌟 Exploring the Intersection of Business and Technology re AI! 🌟 Before my decade-long focus on AI, connecting technology and business in various domains has always been my passion; This is why I enjoyed this conversation with David Luan - No BS, No Twitter exaggerations, No clickbate - Just a super well informed, super smart David sharing his insights from years of hands-on experience: #Startegy, #Software, #Hardware, #Platforms, #ModelSpace and everything in between - LMK what you think of his projections. In This Episode: 1. The Biggest Lessons from OpenAI and Google Brain: What did OpenAI realize that no one else did that allowed them to steal the show with ChatGPT? Why did it take 6 years after the introduction of transformers for ChatGPT to be released? What are 1-2 of David’s biggest lessons from his time leading teams at OpenAI and Google Brain? 2. Foundation Models: The Hard Truths: Why does David strongly disagree that the performance of foundation models is at a stage of diminishing returns? Why does David believe there will only be 5-7 foundation model providers? What will separate those who win vs those who do not? Does David believe we are seeing the commoditization of foundation models? How and when will we solve core problems of both reasoning and memory for foundation models? 3. Bunding vs Unbundling: Why Chips Are Coming for Models: Why does David believe that Jensen and Nvidia have to move into the model layer to sustain their competitive advantage? Why does David believe that the largest model providers have to make their own chips to make their business model sustainable? What does David believe is the future of the chip and infrastructure layer? 4. The Application Layer: Why Everyone Will Have an Agent: What is the difference between traditional RPA vs agents? Why are agents a 1,000x larger business than RPA? In a world where everyone has an agent, what does the future of work look like? Why does David disagree with the notion of “selling the work” and not the tool? What is the business model for the next generation of application-layer AI companies? https://lnkd.in/dUpWWzvN

4

Smaller #genAI models are changing the game! Introducing #mistral-NeMo by #nvidia and #mistralAI with 128k token context window - a tool for businesses that can run on RTX GPUs which you may have at your home. Super inexpensive with no high cloud costs, keeps data private on your private computer! #nvidia #mistralai #genaiforenterprise #responsibleai

10