"NVIDIA-Hosted Fireside Chat with Rod Makoske and Jensen Huang"
Feature Article by Jessica Walton
What exactly does it mean to operate at machine speed? It’s one of several futuristic topics raised by Rod Makoske, Lockheed Martin’s chief engineer and senior vice president, Corporate Engineering, Technology & Operations and Jensen Huang, the founder and CEO of NVIDIA during a special virtual fireside chat held on April 22. These two visionaries spoke to an audience of nearly 2,000 employees from their companies about “the art of the possible,” next-generation computing concepts and game-changing technologies in their respective industries.
At the Rate of Machine Speed
The fireside chat came on the heels of NVIDIA’s GPU Technology Conference (GTC) that took place April 12-16 and featured several panelists from Lockheed Martin. As part of an ongoing collaboration, NVIDIA supplies Lockheed Martin with AI-ready enterprise hardware platforms to accelerate AI and support virtualized AI applications with scale-out performance.
“This is all very relevant for our customers across many legacy platforms and systems,” explained Makoske in his opening remarks during the fireside chat. “We have to bring our customers the most advanced technologies available today to deter adversaries and protect the country. That means being able to operate ‘at machine speed.’”
“And we really have to leverage AI and machine learning across everything that we do,” Makoske continued. “Whether you're talking about reconnaissance or surveillance systems, resilient communications or the cyber environment, we must manage these complex missions that we refer to as Joint All-Domain Operations (JADO), where you have to simultaneously operate across air, land, sea, space and cyber in an integrated fashion.”
(Omniverse is NVIDIA’s open graphic platform utilized by Lockheed Martin for real-time interchange, collaboration and shared virtual worlds and experiences.)
NVIDIA’s Jensen Huang discussed how AI professionals are pioneering solutions to complex challenges across several domains via computers that augment reality, collecting data from millions of sensors across land, sea, air and space. “The rate of progress is not just in scale, complexity and precision,” said Huang. “But also the velocity of innovating at machine speed with the ability to rapidly update our systems in real time through machine learning and a constant stream of data.”
“Basically,” he concluded, “we’re moving faster than Moore’s Law ever predicted.”
The Art and the Science Behind the Digital Twin
Huang primarily attributed the emergence of the “digital twin” to enabling Lockheed Martin to apply AI-driven data to reduce manufacturing costs while continuing to advance design in the defense industry. “It’s about improving or updating the asset, for the entire life of the asset. We discussed it at some of the recent GTC panels, basically the digital twin idea where we have all of these sensors connected to an asset and by monitoring it carefully we learn about the characteristics of the machine and make adjustments for improvements – first in simulation, then in the real world.”
“Challenges that used to take three to five years to solve for a customer are now being resolved in a fraction of the time,” added Makoske. “Let’s say we’re looking to make an adjustment to a rotor or the exhaust gas produced by a helicopter. We can use fluid dynamic-, thermodynamic- and structural models, along with high-performance computing, to look at the existing model in high fidelity. It just really opens the door to the art of the possible.”
Huang was quick to point out that the rate of speed in AI/ML improvements has been developing so rapidly that our own preconceived notions about what is possible in both the defense and commercial industries will have to change, too: “When you think about Moore’s Law in the context of scale, we’re talking about being able to successfully approach problems that are a thousand times bigger tomorrow – things that we would have had trouble imagining being able to analyze and improve just a few years ago. That’s where our imagination has to live.”
“Imagine taking the digital twins and these physics-based models,” responded Makoske, “and everything is integrated and influencing the physical world, challenging the art of the possible. You populate data from a factory, for example, from a model and then those models are going to start impacting what we do in our systems, what we do in our operations. Having that interplay in what you would call a really smart factory – as opposed to just a factory of sensors – well, that’s pretty amazing. That’s a game changer.”
Real World AI Applications
On the topic of applying AI to humanitarian needs, Makoske discussed Lockheed Martin’s ongoing efforts in enhancing firefighting operations through a cognitive mission manager in collaboration with NVIDIA.
“When we look at the wildfires that took place in countries like the U.S. and Australia in recent years, the sheer number of acres that were burned or damaged is hard to comprehend,” said Makoske. “Lockheed Martin and NVIDIA are working together with AI to assist in prediction.”
Expanding on this concept, Makoske provided suggestions on how AI will assist in such a prediction and response model with potentially powerful results:
Assisting fire line detection and prediction by aggregating disparate sensor data with the environmental conditions.
Aiding in response planning by pairing platform capability/availability against the location and other indicators.
Enabling integration of this data across multiple users and the ability to communicate with one another.
“We essentially want to predict the future together, better. AI is going to assist us with sharper prediction. Communicating with each other with these new methods will have huge implications for firefighting, so this is something really powerful that we’re working on together with NVIDIA.”
Success in this area will be contingent on cutting across data from different domains that normally aren’t connected. As Huang pointed out, the solution includes sensor fusion cuts across time on an ongoing basis and rapid response in order to stay ahead of a problem. In other words, the time has come for engineers to tackle modern, geospatial planetary-scale problems.
Imagining the Future of AI/ML
Makoske concluded the fireside chat with a question for Huang: What should Lockheed Martin employees be looking out for in the near future in AI/ML advancement?
“Basically, edge applications,” responded Huang. “Imagine a planet connected by trillions of sensors. Some of these sensors are vision-based, others might be sonic- or temperature-based. The big question is when you’re monitoring every single telemetry, you’re sensing it all in real time, what kind of decision do you make in the heartbeat of a situation? How would you respond differently with this multitude of data in your hands now?”
As Huang and Makoske closed the fireside chat with remarks on recent inventions and new ideas for future applications, the theme of pushing the boundaries of imagination was characterized as equally as important as technology advancement itself. Their view was reminiscent of the words of Albert Einstein: “Imagination is more important than knowledge. For knowledge is limited, whereas imagination embraces the entire world, stimulating progress, giving birth to evolution.”
The speed of improvement in the context of scale might be hard to imagine today, but the payoff is exponential.