When America agreed to stop underground testing of nuclear weapons, supercomputers from the U.S. Department of Energy provided the precise and sophisticated simulations necessary to ensure the arsenal’s ongoing effectiveness.
When scientists sought to model the universe’s evolution, they brought their data to the experts and supercomputers at DOE.
When studying cancer, interpreting mammograms, and improving vaccination campaigns and diagnoses for Americans from children to veterans, DOE computers have been crucial to success of researchers seeking to deliver precision care to patients.
The quiet contributions of these incredible machines have solved some of the world’s toughest problems and kept the U.S. at the forefront of high-performance computing for decades. They are developed and operated by some of the most gifted scientists in the world. These computers will also fuel America’s efforts to maintain the Nation’s global leadership in artificial intelligence.
While the United States leads the world in AI capabilities today, China is gaining ground. The government in Beijing has invested in and backed AI efforts in China with one goal in mind: to overtake the United States and dominate the AI world.
Thankfully, we have a president who is committed to ensuring that doesn’t happen. That is why he recently issued an executive order creating the government-wide American AI Initiative. The president wants to ensure this innovation occurs in the United States.
The executive order calls for increased access to federal data, models and high performance computing resources for America’s AI researchers.
I am proud that some of those resources and computers live at our DOE National Labs, where we currently possess five of the 10 fastest supercomputers in the world, including the top two fastest and the world’s top AI supercomputer.
Through this executive order, the president has given DOE and other agencies a mandate to further utilize our world-class high performance computing infrastructure to build upon and accelerate America’s existing AI research. We are ready for the challenge.
DOE computing capabilities will be critical to securing America’s dominance in artificial intelligence. That is why we must take these already high-performance capabilities to the next level.
I am excited to announce today that along with Cray and AMD, DOE will be building Frontier, a new exascale system at Oak Ridge National Lab, or ORNL.
Exascale systems are capable of at least a billion-billion calculations per second, which means they are at least 10 to 20 times faster than today’s fastest supercomputers, and 50 to 100 times more powerful – the equivalent of going from a flip phone to a smart phone or from a dialup to a 4G internet connection.
DOE has committed to building not just one, but three exascale machines at its National Labs: Frontier at Oak Ridge in Tennessee, Aurora at Argonne in Illinois, and El Capitan at Lawrence Livermore in California, all of which we expect to deploy around 2021-22.
With exascale capability, researchers will be able to look deeper and see further than they ever have before. They will develop better answers to more complex questions and meet more complex challenges with greater precision and speed.
Expected applications include identifying next-generation materials like exotic alloys for the defense, aerospace and energy sectors; determining optimal cancer treatments based on a patient’s unique genetics; refining earth system models to more reliably assess water cycles and agricultural impacts; deciphering high-energy physics data; and accelerating industrial product design.
DOE-fueled AI work is already being used to optimize grid security and increase energy efficiency – among a host of other energy-related applications. It is also being used to reduce traffic jams, strengthen cybersecurity, predict extreme weather, improve emergency response and even develop better treatments for cancer and traumatic brain injury.
DOE’s systems have used machine learning to search through a billion combinations of drugs to find those predicted to maximally inhibit a given tumor, to search through a billion hypothetical compounds to identify those with potential as new drug development candidates, to design new materials, to develop new algorithms to manage electric grid operations.
We must use our supercomputing capabilities to fully develop the potential of AI. That’s why we are coordinating those efforts across our National Labs, and have stood up an AI program at DOE.
Those are the stakes in AI and exascale computing.
The competition from China is great, but America is the world’s leader, and has never shied away from a challenge … and it always thrived when exploring new frontiers.
Rick Perry is the secretary of Energy and a former governor of Texas.
