LEMONT, Ill. (WGN) — The term artificial intelligence dates back 70 years. Today, it’s part of our daily lives in ways unfathomable in the 1950s. And just this year, in Chicago, the technology made a giant leap forward.
The WGN News team, led by Tom Skilling, recently took a trip to Argonne National Laboratory to meet “Aurora,” a supercomputer that’s been on the drawing board since 2015, took eight years to construct, and is a collaboration of Hewlett Packard, Intel, and Argonne.
Among the scientific tasks to which it will be put is mapping the human brain, to model and understand how cancer and other diseases spread—and develop new medicines and treatments, lead the way to cleaner, nuclear power production, and enhance food production. And, Aurora will help forecast the weather and climate faster—and farther out in time—than we ever imagined.
The Aurora Exascale Computer is huge. It’s the size of two NBA basketball courts, weighs 600 tons, and takes 44,000 gallons of recycled water to cool the processors. The engineering gives the machine its distinctive roar.
Katherine Riley, the Director of Science at Argonne Leadership Computing Facility, explained, “What you’re hearing primarily is forced air. So, most of Aurora is water-cooled. So, we’ve got cold water going in, warm water coming out. But still, there are huge numbers of other parts to this system that need to be air-cooled.”
The Aurora is the fastest artificial intelligence computer in the world. It is one million times faster than supercomputers at the turn of this century 25 years ago and 6-billion times faster than Seymour Cray’s 1964 cdc 6600 machine in Minneapolis—widely regarded as the first supercomputer.
The behemoth machine cannot just solve complex equations, but can also deduce things much as the human mind does through “generative machine learning”—meaning, it has the ability to review available information and observations to solve new problems.
“I’ve heard it described as the dawn of a new age in science introducing that. Is that true?” Skilling asked.
“I would think that’s a very fair description, absolutely,” Riley replied.
One of the supercomputer’s superpowers is looking for patterns hidden within immense data sets. It’s a skill that holds promise when it comes to identifying specific compounds that may prove to be more effective medicines for cancer and other disease states. Fusion energy is another area of exploration. And, perhaps more relatable to our daily lives, is what it can do for our phones and cars.
“Other places this is huge is your phone. Do you want a better battery? Do you want a better car?” Riley said. “If you want better materials for those things, this is where artificial intelligence is contributing a huge amount.”
Aurora performs two billion-billion mathematical operations per second – that’s 18 zeros! If every human on earth were issued a computer and knew how to use one, the number of calculations the planet’s entire population would do in four years —is done by Argonne’s Aurora supercomputer in one second – time it takes to say “one-Mississippi.”
Dr. Rao Kotamarthi, a senior scientist at Argonne’s environmental science division, plans to harness Aurora’s computing power to build even faster and longer-range weather models.
“At this point, most of the kind of weather models from AI that you are seeing in newspapers and media are actually not using any physics. You are deriving the models from data itself, there is no physics involved, not much,” he said. “Just looking at the data, learning from the data, it figured out what the next steps of where the weather should go next. That’s the most amazing thing about this.”
Instead of relying on complex physics equations which take hours to run, Argonne’s rapid AI stormer forecasts are trained using the past 50 years of weather observations.
“So, you would feed in the initial state of the atmosphere in a given situation, and it would essentially go back through 50 years of data and look for analogs?” Skilling asked.
“Yes,” Kotamarthi replied.
The past patterns help Aurora predict the weather in a fraction of the typical computer time it takes and with stunning accuracy.
“If you remember, to go from three to five days, it took 15 years of hard work to improve the 3-to 5-day forecast. Now we are saying that I’m going from 5 days to 12, 12 to 15 days in like a year, less than two years!” Kotamarthi said.
While its applications from rapid weather modeling to finding targeted medications are no joke, AI is a technology that’s been sensationalized for decades.
The 1970s science-fiction movie “Colossus: The Forbin Process” portrayed a runaway, impenetrable government supercomputer designed to prevent war. Instead, it threatened to launch nuclear weapons and blackmail its creators.
But this machine is no work of fiction, and according to its keepers, not to be feared.
Dr. Ian Foster, the director of Argonne’s Data Science and Learning Division, explained, “Artificial intelligence is a term that’s used to describe a computer that can perform tasks that you might think a human would be needed for, and why would we want to do that? In science, there is so much information, so many problems to think about that a single human can’t do it all.”
“The focus here is really on how we can use the capabilities to get to really impactful, potentially world-changing science results,” Riley said. “But the point is that it’s really what we choose to do with it. We are the ones making that choice. It’s not making the choice.”
Dr. Foster added, “It’s about intent. The scientists at Argonne don’t want the supercomputer to think on its own.”
In real life, guardrails, including transparency of the uses to which the Aurora Exascale Computer applications at Argonne are in place. More important will be careful observation of private sector AI applications and those originating in unfriendly governments to avoid misuse of a technology which otherwise holds the promise of advances assuring better health, a cleaner environment, better and faster weather forecasts, and technological advances we can only begin to imagine.
We’re in a race with other countries for leadership in high-speed computing. China, Japan, India, and the European Union are among those working on computational speed like the Aurora. National defense is an inseparable part of the equation—maintaining defense against armaments which might be deployed against us.
FAQs:
Q: What is the purpose of the Aurora Exascale Computer?
A: The purpose of the Aurora Exascale Computer is to aid in scientific research, particularly in the fields of medicine, energy, and the environment, by providing a powerful tool for data analysis and processing.
Q: How fast is the Aurora Exascale Computer?
A: The Aurora Exascale Computer is capable of performing two billion billion mathematical operations per second, making it the fastest artificial intelligence computer in the world.
Q: What are some of the potential applications of the Aurora Exascale Computer?
A: Some potential applications of the Aurora Exascale Computer include mapping the human brain, developing new medicines and treatments, leading the way to cleaner and more efficient energy production, and enhancing food production.
Q: What is the potential for misuse of the Aurora Exascale Computer?
A: The potential for misuse of the Aurora Exascale Computer is a concern, as with any powerful technology. However, the scientists at Argonne National Laboratory are working to ensure that the technology is used responsibly and for the betterment of society.
Q: How does the Aurora Exascale Computer work?
A: The Aurora Exascale Computer works by using “generative machine learning” to review available information and observations, and to solve new problems. This allows it to learn and improve its performance over time.
