This isn’t the first time I have talked about Nvidia and it certainly won’t be my last. Nvidia had some really exciting announcements at Computex Taipei 2023. I’m not going to talk about all of them, I am just going to give a brief overview and then talk about what I find the most interesting.
Nvidia Loves Gamers
CEO of Nvidia Jensen Huang certainly isn’t catering to the gaming community, as Jensen started the presentation excitedly holding up an RTX 4060ti for their “core gamers”, arguably the biggest failure of a card in several generations, using it as a prop to transition to talking about ray tracing. Later in the presentation Huang did get in a joke about the DGX GH200 GPU server stack with 144 terabytes of connect ram being able to run Crysis. Or can it? He said “only gamers would understand it”. From a business standpoint, obviously gamers are not the priority, but I don’t really think Nvidia should be rubbing it in our faces. When gaming is really what built their company it really does feel like a slap in the face especially after the last two generations of graphics cards.
Although he did introduce an interesting thing within the gaming industry: Nvidia’s (ACE) Avatar Cloud Engine, part of Nvidia Omniverse, using Generative AI to simulate characters in real time. Not writing dialog for the simulated character, only writing it a background is enough for it to interact with the inputs of the player or another AI. This is a staggering development for gaming, and game developers. Imagine the NPCs acting in a realistic manner when you speak to them. Jenson declared “this is the future of video games”. I agree with that sentiment. The ACE is a suite of real-time AI solutions for end-end development and deployment of interactive avatars and digital human application at-scale according to their website. It supports multiple input types that can then be turned into a 2d or 3d avatar by the AI.
Not Just for Gamers
Just in case you thought your customer service job was safe, Nvidia is also leveraging this technology with their (UCF) Unified Compute Framework, which allows for low code implementation of cloud native, multimodal AI applications. This uses Avatars to assist in the retail sector making use of AI Avatar Workflows. It’s being touted as the next generation of customer engagement, marketing and branding content generation, and live communications, intelligent AI-service agents, but not limited to just that.
Although he did introduce an interesting thing within the gaming industry: Nvidia’s (ACE) Avatar Cloud Engine, part of Nvidia Omniverse, using Generative AI to simulate characters in real time. Not writing dialog for the simulated character, only writing it a background is enough for it to interact with the inputs of the player or another AI. This is a staggering development for gaming, and game developers. Imagine the NPCs acting in a realistic manner when you speak to them. Jenson declared “this is the future of video games”. I agree with that sentiment. The ACE is a suite of real-time AI solutions for end-end development and deployment of interactive avatars and digital human application at-scale according to their website. It supports multiple input types that can then be turned into a 2d or 3d avatar by the AI.
The Nvidia Omniverse has other capabilities as well. Creating digitized factories will allow manufactures to retool in the digital space, to optimize workflow and minimize downtime. This can be done through simulation with no art required. It can start from photos of the inside of your factory or a 2d floor plan. It has realistic physics and robots you program within the space will notice no difference between the real-world application. In the keynote Jensen gave a demo of digitized smart factory. He stated “The worlds largest industries make physical things. Building them digitally first can save enormous costs,” This gives manufactures the power to build and operate virtual factories, manufacturing and inspection workflows. It will also help prevent last minute delays and surprises.
World Domination is Eminent
The main theme of the presentation as a whole is that computing is experiencing a paradigm shift with two new types of computing. Accelerated computing and Generative AI. Its domain specific, meaning, it needs a new full stack development for each language it represents, and a data center scale. Medical researchers need a full stack, engineers need a full stack, manufacturers need a full stack, each domain of science needs its own stack.
Accelerated computing is created through specialized hardware to speed up the execution of demanding applications. It’s generally some combination of a (GPU) graphics processing unit and or an (FPGA) field programmable gate array. The hardware is referred to as the accelerator. Essentially it’s a dedicated server stack that is a computer that is dedicated to a specific task or set tasks to accelerate or increase the compute power of those tasks.
Generative AI is technology that uses advanced algorithms to simulate human intelligence and beyond using accelerated computing to do it at lightning speed. It allows for incredible prediction for anything that has enough data and crushes the traditional methods. DALL-E2, GPT-3, GPT-4 and Chat GPT are all examples of generative AI, but they are likely just the beginning of what this technology is capable of. Imagine a large language model trained in only chemistry or only physics and what it could be capable of or capable of assisting with.
Skynet Is Going Live
Jensen explains that because each domain needs a full stack that is why it took 3 decades. The entire system needs to be retooled from the ground up. From the chip to the systems to the system software, algorithms and the applications. Jenson goes on to explain that in order for there to be a new computing model, you need developers, but to get developers you need end users. Without users you have no computers. Without those computers you have no developers and thus it’s a loop. Nvidia has solved that problem by building the computers and getting the developers and building the applications and is now rolling it out to end users like you and me or the 40,000 mega corporations that invested in the technology.
He demonstrates Nvidia’s $34 million Data Center Model that is 150x more powerful at performing 1xLLM (1 Billion parameter Large Language Model) tasks for the same power output 11 GW/H of a similar power consumption of a modern Data Center. The implications are that the businesses using these computers that are data centers are going to need the power to drive AI. Jensen goes on to say that in the future every single company will have factories, intelligence factories, artificial intelligence factories.
Be Ready: The future is Now
Nvidia’s technology along with the mega corporations and conglomerates will be changing the way we live everyday. From this moment forward AI will be inculcated into society as a whole. Prepare yourselves because many things will change forever. This technology will allow us to make massive progress in a short period of time, but it will also alter many of today’s existing markets, the way we do business and the way we live.
This discussion highlighted Nvidia’s keynote at Computex Taiwan 2023, showcasing their focus on gamers, innovative technologies like the Avatar Cloud Engine, applications in various sectors, and the paradigm shift in computing with accelerated computing and generative AI. Nvidia’s vision for the future, along with its collaborations, will undoubtedly reshape our lives. Join the discussion by subscribing and following on social media.
Best Regards,
Chris