[tt] CHE: What's So Super About Supercomputers, Anyway?

[Premise Checker]

What's So Super About Supercomputers, Anyway?
The Chronicle of Higher Education 7.9.21
http://chronicle.com/weekly/v54/i04/04a02402.htm

By DAN CARNEVALE

At last week's National Science Foundation symposium on "Cyber-Enabled 
Discovery and Innovation," the speed of new supercomputers was not what 
people were talking about. How those computers can solve the world's ills 
— and why they have been slow to take those problems on — was the big 
topic. The event was held at Rensselaer Polytechnic Institute, in Troy, 
N.Y.

Clas A. Jacobson, chief scientist of controls at United Technologies 
Corporation, said that even with the growth of computing power recently, 
computers can only tackle so much information at one time.

Plus, he said, supercomputers are not available to everyone who needs 
them.

That leads to researchers taking short cuts, perhaps by focusing on small 
samples of data instead of looking at the big picture. For combing through 
data on, say, human behavior on a wide scale, supercomputers could 
eventually become overwhelmed with information.

"Computing is a means to an end," Mr. Jacobson said. "It's not the answer 
by itself."

That doesn't mean institutions will be putting superfast computers on the 
back burner. To the contrary, at this very meeting, Rensselaer held a 
ribbon-cutting ceremony for its new supercomputer, which officials said is 
the seventh fastest in the world and the fastest of any supercomputer on a 
college campus. It's able to perform 100 trillion calculations per second, 
and cost $100-million.

Supercomputers are often limited in their application because of computer 
scientists' ignorance of subjects other than computer science, presenters 
at the supercomputer symposium said.

Russ Miller, a professor of computer science and engineering at the State 
University of New York at Buffalo, said university computer-science 
departments need to recruit graduate students who have undergraduate 
degrees in diverse fields, such as biology, physics, and humanities. 
Currently, most computer-science students know a lot about computer 
science, he said, but not much else.

"They don't have a broad enough worldview of what they're doing computer 
science for," he said. "So they're just interested in building the next 
best computer-science widget and hope somebody is interested in that."

John Sasso, who works for the New York State Office of Technology, said 
computer-science departments needed to take students with diverse skills 
and use them to think creatively when solving problems. It's too often, he 
said, that universities teach students how to create a computer program to 
solve a problem, but not how to identify the problem in the first place.

"Students are being taught to solve things in more of a recipe kind of 
fashion," he said.