NVIDIA's Most Powerful AI Chip

PLUS: Elon Musk Open Sources Grok

Author

AI For All
March 22, 2024

Sponsored by

Hello readers,

Welcome to another edition of This Week in the Future! NVIDIA returns to deliver us from slow AI infrastructure, unveiling the GB200 Grace Blackwell Superchip and the DGX SuperPOD at NVIDIA GTC 2024. Plus, Elon Musk open sourced Grok.

Let’s get into it!

NVIDIA’s Most Powerful AI Chip Yet

NVIDIA Blackwell architecture

So, NVIDIA GTC 2024 happened, and there were several announcements that should put a smile on the faces of technology executives the world over (except China).

NVIDIA GB200 Grace Blackwell Superchip

Behold, NVIDIA’s greatest achievement, the NVIDIA GB200 Grace Blackwell Superchip. This innovative superchip connects two NVIDIA B200 Tensor Core GPUs to the NVIDIA Grace CPU via a 900GB/s ultra-low-power NVLink chip-to-chip interconnect, marking a significant advancement in computing efficiency and performance for AI applications. With these new chips comes a new supercomputer …

NVIDIA DGX SuperPOD

The DGX SuperPOD, powered by the Blackwell chips, is said to be capable of processing trillion-parameter models while using less energy. Training a 1.8 trillion parameter model, which previously required 8,000 Hopper GPUs and 15 megawatts of power, can now be accomplished with just 2,000 GPUs from the new Blackwell architecture, using only four megawatts.

Project GR00T

NVIDIA also showed off their robotics prowess with Project GR00T. GR00T (Generalist Robot 00 Technology) is a general-purpose foundation model designed for humanoid robots, enabling them to learn from minimal human demonstrations. It utilizes NVIDIA's advanced GPU-accelerated simulations for training, incorporating imitation learning and reinforcement learning through NVIDIA Isaac Lab, as well as generating robot movements from video data​.

Our Take

NVIDIA’s efforts depend on the scaling hypothesis being true (as well as the inefficiencies of generative transformers). It remains to be seen how much more performance can be squeezed from these LLMs just by improving hardware. Should a fundamental breakthrough make AI as compute efficient as the human brain, then it’s goodbye NVIDIA. Regardless, NVIDIA deserves credit for perfectly responding to the current demands of AI and pushing the limits of computation.

Grok-1

True to his word, Elon Musk open sourced Grok. xAI released the model weights and network architecture of Grok-1 on GitHub. This comes after Elon Musk sued OpenAI, criticizing it for keeping its models private (and serving them up to Microsoft).

While Grok isn’t GPT-4 level, it’s a capable 314 billion parameter Mixture-of-Experts model with 25% of the weights active on a given token.

Why This Matters

This is another huge win for open source AI, which is said to be necessary if AI is to mediate the information diet of the future (since the internet is also built on open source technologies). The future is looking like a plurality of personalized AIs.

🔥 Rapid Fire

🎙️ The AI For All Podcast

This week’s episode featured Nikola Mrkšić, CEO of PolyAI, who discussed AI-powered solutions and voice assistants in unlocking efficiency and improving the customer experience. Our conversation explores the growing adoption of AI across various industries, the pivotal role of generative AI in deploying scalable solutions, and the risks associated with implementing generative AI.

📖 What We’re Reading

When AI Outperforms Humans: A Lesson from Medical Text Summarization (link)

“AI-produced summaries can be a huge time-saver. We’ve seen this serve as an aid to draft emails and papers and research in-depth topics and more. The results, despite needing to be verified, often provide a baseline to get started at the very least. In healthcare, this is of significant value to busy medical professionals.”

Source: AI For All
To Drive Innovation with GenAI, Start by Questioning Your Assumptions (link)

“A company’s efforts to innovate can fall short for many reasons. Critical assumptions about customers, technologies, rivals, and innovation domains may not reflect current trends and circumstances. Whether the focus is on new products, services, processes, or business models, generative AI can enhance and challenge the work of teams across all phases of the innovation cycle.”

Source: Boston Consulting Group

💻️ AI Tools and Platforms

  • Phospho → Open source text analytics for LLM apps

  • Nanonets → Intelligent automation for business processes

  • Airtrain → No-code LLM fine-tuning and evaluation

  • Clarifai → Full stack AI developer platform

  • Adaptive ML → Build singular generative AI experiences

AI: Balancing Risk and Return

Innovative technologies are revolutionizing business as we know it, and they’re more accessible than ever. But to truly harness the transformative potential of AI, you need to know how and when to use it. And which pitfalls to avoid.

The six-week Artificial Intelligence: Implications for Business Strategy online short course from MIT Sloan School of Management and MIT Computer Science and Artificial Intelligence Laboratory explores AI’s business applications and challenges. 

Choose this program to:

  • Optimize your business: Leverage AI, ML, and robotics to drive efficiencies, improve productivity, and support your growth.

  • Develop a strategic roadmap: Apply your knowledge to effectively integrate AI into your business.

  • Gain a dual perspective: Benefit from a course designed by two prestigious schools — the MIT Sloan School of Management and the MIT CSAIL.

  • Conveniently build career-critical skills: Follow a program that fits your schedule and benefit from 24/7 support and various payment options.