Communications of the ACM
Machine-Learning Systems: Your New Co-Workers
Domain scientists and computation scientists "are far apart as collaborators," says Nirav Merchant of UA's Data Science Institute.
Credit: Bob Demers / University of Arizona
A dystopian world ruled by machines, one bereft of human input or interaction, seems like a far-fetched scenario. And yet, such sophisticated technology is fast becoming a differentiator in the workplace, changing the way we do our jobs—and the very jobs themselves.
"Machine-learning-based systems are becoming your new co-workers," said Nirav Merchant, director of the University of Arizona's Data Science Institute, at the University of Arizona College of Science lecture series on "Humans, Data and Machines."
"This technology will push us toward excellence, and that's what we have to look forward to," Merchant said, addressing a capacity crowd at Centennial Hall. The lecture series is designed to delve into various aspects of the revolutionary social change now underway with the convergence of the physical, digital, and biological worlds.
Statistics Merchant cited in his talk, "Working Alongside Thinking Machines," can be startling. In one minute's time on the Internet, 87,000 hours of programming are streamed on Netflix, 4 million searches are done on Google, and 13,000 apps are downloaded. The U.S. alone generates 2.6 petabytes of Internet data every minute; a petabyte is the equivalent of 13.3 years of HDTV video.
Too much to wrap your head around? Merchant said technology's evolution can be observed in something as simple as the cruise-control feature on a car. What began as a way to give the driver's right foot a rest—and still maintain speed while traveling—can now adapt to distance as well, keeping a vehicle safely separated from the ones in front of it. Clearly, today's cruise control is doing more "thinking" than the older version did.
Using the illustration of a pyramid, Merchant showed that it's no easy matter transforming data (the pyramid's base) into wisdom (its apex). Through processing, data becomes information; through cognition, information becomes knowledge; and through judgment, knowledge becomes wisdom. Even now, most of our understanding is limited to the data and information levels, he said.
He said a tech tool such as Google Project Sunroof might work fine for determining the amount of solar radiation on your house. But give it a larger target—say, Arizona's Santa Catalina Mountains—and the assignment becomes much more complex. The answer, in part, will require bridging the gap between domain scientists, who are experts at gathering data, and computation scientists, who specialize in analysis methods and infrastructure.
"These two groups are far apart as collaborators," Merchant said, "and we need them to have a much more active dialogue. They need to work together."
The UA's CyVerse represents a step in the right direction, he said. Launched 10 years ago as iPlant, it has become a cyberinfrastructure for 1,600 non-academic institutions, 3,700 academic institutions, and 51,000 active users, converting data to information. Analysis that formerly took months can now take days, in some cases.
"CyVerse has democratized cyberinfrastructure," Merchant said. "What we're trying to scale is the sharing."
Another way of explaining that, he said, is through an African proverb: "If you want to go fast, go alone. If you want to go far, go together."
No entries found