WASHINGTON, D.C.—The U.S. Department of Energy (DOE) is planning a major initiative to use artificial intelligence (AI) to speed up scientific discoveries. At a meeting here last week, DOE officials said they will likely ask Congress for between $3 billion and $4 billion over 10 years, roughly the amount the agency is spending to build next-generation “exascale” supercomputers.
“That’s a good starting point,” says Earl Joseph, CEO of Hyperion Research, a high-performance computing analysis firm in St. Paul that tracks AI research funding. He notes, though, that DOE’s planned spending is modest compared with the feverish investment in AI by China and industry.
But DOE has a big asset: torrents of data. The agency funds atom smashers, surveys of the universe, and the sequencing of thousands of genomes. “We generate almost unimaginable amounts of data, petabytes per day,” Chris Fall, who directs DOE’s Office of Science, said at the last of four town halls DOE has held here to build support for the AI initiative. Algorithms trained with these data could help discover new materials or spot signs of new physics. “It’s going to impact everything we do,” Fall says.
DOE is joining a global rush to fund AI. Worldwide corporate AI funding is expected to hit $35.8 billion this year, up 44% from 2018, according to IDC, a market analysis firm. Companies see commercial advantages in AI. It helps banks detect and prevent credit card fraud, and oil and gas companies use it to pinpoint productive drilling sites in mounds of geological data.
Governments are also jumping in, with goals as diverse as improving traffic flow and detecting early-stage cancers. In February, President Donald Trump signed an executive order launching the American AI Initiative, and the administration requested nearly $1 billion for AI and machine learning research in fiscal year 2020 across all civilian agencies, according to the U.S. Office of Science and Technology Policy. The U.S. Department of Defense is seeking a similar level of funding for unclassified military AI programs.
Definitive numbers are harder to come by for other countries. But in 2017, China announced a national AI plan that aims for global leadership, and a projected commercial AI market worth 1 trillion yuan ($140 billion) by 2030. And the European Union has committed to spending €20 billion through 2020. “Just as with the race to create the first exascale computer, the AI world is getting really competitive,” Joseph says.
Just who is leading the race depends on what you measure. According to Hyperion Research, China accounted for 60% of all investments in AI from 2013 to 2018. U.S. investments were about 30% of the global total. China dominates the number of AI publications, whereas the European Union has the most AI researchers, Joseph says. But U.S. researchers in AI get the most citations per paper, he says, suggesting their research has the most impact.
DOE officials say the AI initiative will help keep U.S. researchers at the forefront. Though DOE has yet to detail its program, it’s likely to include funding for national labs to optimize existing supercomputers for AI, and external funding for academic research into AI computer architectures, says Rick Stevens, associate laboratory director for computing, environment, and life sciences at Argonne National Laboratory in Lemont, Illinois.
Not only is the initiative likely to speed up data analysis, but it could also boost the pace of data collection, by using AI to come up with hypotheses and design new experiments. “AI won’t replace scientists,” says Jeff Nichols, associate laboratory director for computing and computational science at Oak Ridge National Laboratory in Tennessee. “But scientists who use AI will replace scientists who don’t.” Battery researchers, for example, could use these tools to test the properties of thousands of materials in days.
Nichols acknowledges that DOE’s push for AI lags efforts at the U.S. National Science Foundation (NSF), which has spent roughly $4.5 billion over the past decade on research to improve AI algorithms and software. Nevertheless, Erwin Gianchandani, NSF’s acting assistant director for computer and information science and engineering, says that given DOE’s role in funding major user facilities such as supercomputers, the agency’s AI push “would dovetail well” with NSF’s programs.