The Pivotal Role of Nvidia in The Future of AI
NVIDIA has evolved beyond selling graphics cards.
Hi!
Just a quick reminder that there’s a special 25% Substack discount for the annual plan. Click here to get an annual subscription to my Substack (yes, annual subscribers can redeem up to 2 of my Udemy courses for free)
In 1848 in California, gold was discovered, sparking what historians would later call the Gold Rush. Hundreds and thousands of individuals flocked to the San Francisco area, and within just two years (1848 and 1849), the population surged to 25 times its original size. Many people struck it rich in this period of upheaval, but there were also those who lost everything. You might be curious about who managed to amass such fortunes and what they had in common.
Contrary to initial assumptions, the biggest beneficiaries of this frenzy were, surprisingly, the manufacturers of pans used for gold extraction. Additionally, clothing manufacturers like Levi’s also saw considerable success. Essentially, the real wealth during the Gold Rush was accumulated by those who supported the gold seekers.
This pattern has recurred throughout every technological or industrial boom humanity has faced.
Turning to the present, we're in the midst of what could be termed the AI fever. You might quickly wonder, who are the early winners in this AI boom?
Companies like Meta, Google, OpenAI, Tesla, and others are all on the hunt for gold. They are currently exploring, as demonstrated by the launch of their AI models, representing a significant investment. This approach reflects a broader pattern of seeking foundational tools and support systems as the most reliable means to capitalize on new frontiers.
But specifically, the company that is enabling the full potential of AI to be unleashed and become a reality is NVIDIA, led by Jensen Huang. On March 18, at an event called “GTC March 2024 Keynote” he gave us a glimpse into what AI is working on regarding advancements that will impact human life in the not-so-distant future.
GPU Blackwell
"Blackwell is not just a chip; it's the name of a platform," Jensen Huang emphasized while introducing his new GPU, set to succeed the Hopper model. This new platform marks a significant opportunity for the tech giants to deploy new AI models that can handle an immense number (about 10 trillion) of parameters, with an energy consumption cost reduced by up to 25 times compared to the norm. Although it may be challenging to comprehend the scale of this progress in straightforward terms, NVIDIA's breakthrough signifies for us, the users, the possibility of accessing models that will undoubtedly be trained with enhanced multimodality and a higher capacity for generative response processing. Simply put, Blackwell is poised to propel the next industrial revolution.
Though it may seem surprising at first, major tech giants like Google, Meta AI, Microsoft, OpenAI, and AWS are already queuing up to acquire this groundbreaking technology. The AI race involves algorithms, but there's an equally crucial component where GPUs play a leading role.
Robots: Physical AI
NVIDIA is developing a digital ecosystem known as Omniverse Cloud, which is set to operate on a computer designated as OVX. The intention behind this is to provide robots with a simulated environment, allowing them to train across a variety of contexts that closely resemble real-world conditions.
Aligning with this initiative, NVIDIA has unveiled its project Gr00t, which is promoted as a general-purpose model for humanoid robots. This project integrates both software and hardware components, embodying the concept of a versatile robot. These robots are designed with the capability to understand human language and interact effectively with their environment.
At a glance, autonomous or AI-driven robotics finds application in a wide array of practical areas, including vehicles, industrial equipment, and sequential tasks or processes, such as in a warehouse setting.
NVIDIA's Impact on AI
From the early stages of Deep Learning, which we might now view with a touch of nostalgia, the journey wouldn't have been possible without the simultaneous evolution of the graphics card market. As AI models grew in complexity and size, NVIDIA's expansion became inevitable.
But why is this the case?
This can be explained by Moore's Law, which predicts that the number of transistors on microprocessors used in our devices would double roughly every two years, enhancing their capabilities and efficiency. This trend is exactly what we've observed with GPUs. We saw this progression with the Q8000 in 2019, the A100 in 2020, the Hopper H100 in 2022, and now with Blackwell.
In essence, NVIDIA's current role is to provide the powerhouse necessary for AI to reach its full potential.
NVIDIA's business strategy might seem to be currently geared towards providing AI-related technologies primarily to corporations. However, many of the open-source models that we download and run on our personal computers are increasingly demanding more memory. Jensen Huang is aware of this trend, as leading companies like Meta, OpenAI, and the newly introduced Grok project are all striving for AI models with enhanced capabilities. This indicates that to access these advanced models, we will need a graphic card robust enough to handle them.
Jensen Huang has established a near-monopoly in this space, with numerous companies attempting to ensure that their progression in AI model development doesn't lead to dependency. Yet, the lack of competitors in the semiconductor industry currently makes it challenging not to link NVIDIA directly with the advancement of AI.
This dependence is set to expand further through Omniverse Cloud, NVIDIA's initiative to interconnect various departments within a company to enable AI-driven operations, thus minimizing errors and the need for data conversion or transfer.
NVIDIA has evolved beyond merely selling graphics cards. With little noise, it has become a pivotal player in enabling access to many of the models we use today.