From monitoring and mitigating the impact of climate change on the environment and economy, to advancing disparate domains such as robotics, precision agriculture, education and quantum computing, the Commonwealth Scientific and Industrial Research Organisation (CSIRO) is working across a broad range of areas to apply AI to Australia’s national challenges.
At NVIDIA GTC, CSIRO CIO Brendan Dalton discussed this effort with Keith Strier, vice president of Worldwide AI Initiatives at NVIDIA, during a 45-minute fireside chat.
Touching on CSIRO’s history of innovation, Dalton described how the organisation, tasked by the Australian government to lead the national scientific and industrial research agenda, invented wireless LAN in the 1990s, which led to a global boom in wireless devices.
Today CSIRO, is leading Australia’s national AI ambitions, which includes a collaboration with NVIDIA to help expand the country’s domestic AI capabilities and accelerate adoption of the technology.
This includes teaming up on a broad range of accelerated computing initiatives. One of the timeliest — physically accurate digital twins — will harness NVIDIA’s unique capabilities in graphics, scientific and high-performance computing.
“CSIRO has been using NVIDIA’s accelerated computing platform for over a decade, and I envisage that this new collaboration will expand our efforts around AI research, startups and industrial ventures, grow a more robust local AI ecosystem, and support the launch of our new National AI Centre,” said Dalton.
As part of the Governments AI Action Plan, Australia’s new National Artificial Intelligence Centre is hosted by CSIRO and will work with partners across Australia to weave the national fabric for responsible and inclusive AI, creating opportunities for every person, every business and the country.
Supporting AI Infrastructure
NVIDIA DGX systems have been at the heart of the AI infrastructure that supports these initiatives.
Dalton pointed out that NVIDIA has a history of providing the full stack — hardware and software — to enable an AI revolution.
“Machine learning has been around for a while, but people just didn't have the compute power to be able to make it real. And it was a translation of GPUs from gaming to research applications that provided the impetus,” he said.
Faster GPUs, Better Predictions
Researchers working on the industrial application of new AI solutions will have a safe place to test and iterate their models. They can access optimally configured NVIDIA GPU clusters at scale and a library of software development kits and containers to accelerate data science workflows. They can also trial new technologies on sand pits and get needed technical support.
“We will have people to support scientists and engineers to make sure their codes are running the best way they possibly can, and to make sure they have the right training to develop the kind of algorithms they want,” said Dalton. “I think getting these things right for innovators at a national scale is the best way to grow the AI economy in Australia.”
Australian researchers have developed or are working on a host of AI applications, one of which is bushfire prediction. Bushfire, or wildfire, is a regular occurrence during the hot summers in Australia and the U.S. It can cause widespread damage to people, property and the environment.
CSIRO scientists have designed an AI tool that helps first responders better predict what the bushfire around them will do next so that they can adopt the appropriate firefighting tactics. Using NVIDIA GPU technologies helps train the models faster. The goal of using AI for predicting weather and different kinds of natural disasters is to save lives and property.
Dalton also shared his intrigue about a recent article on how AI can control superheated plasma inside a fusion reactor.
“To me, that’s the holy grail. It’s moving from models that predict technologies developed within models to pushing the frontiers of engineering, controlling forces and shaping solutions in ways that humans just can’t,” he said.