7 OCTOBER | LONDON 2024
SEPTEMBER 12TH - 14TH
The O2, LONDON
The World's Largest AI Supercluster: xAI Colossus
xAI's Colossus, the world's largest AI supercomputer, is getting even bigger. Just over a month after its initial launch, the company announced it's doubling its GPU count to 200,000.
By the CogX R&I team
October 31, 2024
On October 4, 2024, xAI's Colossus - the world's largest AI supercomputer - got the green light from the Greater Memphis Chamber to double in size. Built just a month earlier in Memphis, Tennessee, Colossus will boast an unprecedented 200,000 NVIDIA Hopper GPUs, up from its already impressive initial deployment of 100,000 units.
Source: Midjourney
In a statement released earlier this week, NVIDIA revealed that rather than using its InfiniBand for networking interconnectivity, the cluster instead relies on the NVIDIA Spectrum-X ethernet networking platform for its Remote Direct Memory Access (RDMA) network.
What's really turning heads is how quickly they put it all together. The entire facility and supercomputer were constructed in just 122 days, with only 19 days needed from initial rack installation to the start of training AI models. According to NVIDIA, the system's performance has exceeded expectations, achieving 95% data throughput thanks to Spectrum-X's traffic control, compared to the typical 60% with standard Ethernet systems.
“xAI has built the world’s largest, most powerful supercomputer,” said a spokesperson for xAI. “NVIDIA’s Hopper GPUs and Spectrum-X allow us to push the boundaries of training AI models at a massive scale, creating a super-accelerated and optimised AI factory based on the Ethernet standard.”
Now read the rest of the CogX Newsletter
The Future of AI Cannot Be a Race to the Bottom
Sama CEO Wendy Gonzalez shares invaluable insights on building an ethical foundation for AI development in our latest thought leadership piece.
Guest contributor: Wendy Gonzalez, CEO, Sama
It’s no secret that AI is already a major part of the global economy—the question isn’t if AI will grow, it’s how much AI will grow. One projection has the market reaching over $800 billion by 2030. With so much money to be made, it can be tempting for companies to pursue the lowest-cost options for their AI supply chains—maximising their own profits.
But AI itself is a world-changing technology that presents an opportunity. It can completely reshape our world economy and allow capital to flow to places that have been traditionally left behind by tech booms. My company, Sama, was founded on the principle that talent is equally distributed around the world, but opportunity isn’t, and we’ve focused our efforts in Kenya and Uganda.
AI is still relatively young, and companies have more power than they realise to shape this technology’s future for the better. The time is now to build a healthy, thriving industry that values every worker in the supply chain.
… want to keep reading? Check out the full OpEd here on the CogX Blog
Also in the news
A new AI search engine to rival Google: Meta is reportedly developing its own AI-powered search engine to challenge Google's dominance. The company aims to provide AI-generated summaries of current events directly within its chatbot, according to a report from the Information. However, the road to success won't be easy. After all, Google has been the king of search for decades.
Google Deep Mind open-sources SynthID: Google DeepMind has taken a significant step towards enhancing AI transparency by open-sourcing SynthID, a tool designed to watermark AI-generated text. While the technology is still in its early stages, it offers a promising solution to identify AI-generated text and promote trust in digital information.
Google is reportedly developing a new AI system: Google's cooking up a new AI assistant, codenamed "Project Jarvis," to automate web tasks. Powered by Gemini, Jarvis could take over web browsers, performing actions like gathering research, making purchases, or booking flights. It seems like every tech giant is racing to create super-powered AI assistants. Recall Microsoft's Copilot Vision or Anthropic's new "computer use" feature.
Grok gets eyes: X's AI assistant, Grok, is now getting smarter. With its new vision capabilities, Grok can analyse images, helping you understand complex visuals, identify objects, and even generate creative ideas based on what it sees.
Apple's AI is here, but not for everyone. Apple Intelligence, a suite of AI tools, is now rolling out to select iPhone users. It offers features like smarter notifications, improved messaging, and a revamped Siri. However, access is limited to the latest iPhone models and requires joining a waitlist.
Let Anthropic take control of your computer
Want to automate your tedious computer tasks? Anthropic's new Claude 3.5 Sonnet might be the answer. This AI assistant can handle tasks like form-filling, file management, and web searches. With just a week into its beta, it's already making waves in the AI world. Right now, this is only in beta testing, and only available to developers with access to the Claude API.
How to set up Claude's ‘Computer Use’ in minutes:
Install Docker: Download and install Docker from docker.io. Launch Docker Desktop and accept the recommended settings.
Get Your API Key: Visit console.anthropic.com, log in, and create a new API key under "Get API Keys." Copy the key for the next step.
Configure the System: Open Docker's terminal and paste the configuration code with your API key. Wait for the system to initialise.
Access the Interface: Open localhost:8080 in your browser to see the virtual desktop. You'll find tools like Firefox, Excel, and Terminal ready to use.
Start Commanding: Type your instructions in the left panel. For example, "Find the weather for the next 3 days in San Diego and create a spreadsheet with the information." Watch as Claude navigates the web and creates files autonomously.
In case you missed it
Chinese company EngineAI just unveiled their life-size general-purpose humanoid SE01. A humanoid robot with a natural, human-like walking gait:
Popular Articles
Get the CogX Newsletter
Get the latest tech news in your inbox each week
Getting Machine Learning Projects from Idea to Execution
Issue 43
Eric Siegel, Ph.D., former Columbia University professor and CEO of Gooder AI, outlines practical strategies discussed in his new book, The AI Playbook: Mastering the Rare Art of Machine Learning Deployment, to help organisations turn machine learning projects into real-world successes.