CORVALLIS — A little-known military agency that spends billions on futuristic projects expects little success from most of its work.

But the small percentage of radical ideas that emerge from researchers, including those at Oregon State University, could dramatically change your life.

Need proof? Perhaps you’ve heard of the Internet.

Projects to construct an entire computer system on a single chip or create artificial intelligence put graduate students through school and earn them jobs in the high-tech market, garner patents and royalties for organizations like OSU, and often lead to successful start-up companies. But the funding behind much of it happens to come from the U.S. Department of Defense.

The Defense Advanced Research Projects Agency, or DARPA, is a group of federal offices in Arlington, Va., which doles out close to $2 billion each year to fund high-risk but potentially high-yield experiments around the country.

The idea is to keep the U.S. military one step ahead of its enemies, but the groundbreaking research affects people at all levels of civilian life as well.

In a computer lab in OSU’s Kelley Engineering Center, Professor Tom Dietterich opens a laptop to display the results of a three-year project to create artificial intelligence, or a machine that learns.

The lines of text and symbols running down the screen are a computer program which Dietterich and others hope will some day learn to teach itself.

Dietterich, director of Intelligent Systems Research in OSU’s School of Electrical Engineering and Computer Science, is in charge of three projects supporting 16 graduate students and several undergraduates.

“In the artificial intelligence field, people always ask how smart a computer can be,” he said. “We don’t know. We think we can make them smarter.”

One project attempts to build a computer than can learn to play video games, like “War Craft 2,” then transfer what it learned to other games.

The challenge digs at the question of how strategies are learned, Dietterich said.

“Computers traditionally learn things by rote memorization, but in order to transfer skills from one task to another, you have to get at the deep, underlying knowledge,” he said. “We won’t completely solve this problem in three years.”

Another project, called TaskTracer, is a desktop program intended to coordinate the various applications run on a computer, remember where the user left off and predict where the user will resume.

TaskTracer is part of a larger research project to invent a personalized assistant that learns (PAL).

PAL is hoped to be an artificial intelligence secretary which does certain tasks, such as prepare purchase orders or documents, and tries to understand the needs of its owner.

The experiments are subcontracted from larger projects awarded to groups like Lockheed Martin and SRI International.

While groups such as these conduct the fundamental research, Dietterich said commercial spin-off companies usually form to turn the research into marketable, military products, delivering benefits such as greater satellite surveillance imagery or increased detection of improvised roadside bombs.

Other spin-off companies, like MusicStrands of Corvallis, market to a civilian audience.

However, DARPA technology is usually so forward-looking that it takes years before it filters into common use.

At first glance, the agency resembles something from a spy novel.

According to its Web site, DARPA’s only charter is radical innovation.

After the Soviets beat the U.S. into space with the 1957 launch of Sputnik, President Eisenhower started the research organization to bridge the gap between short-term budget constraints and futuristic research.

The agency initially innovated space technology until the formation of NASA. It moved to missile defense until an Army agency took over.

DARPA focused on many areas over the years, creating Internet technology in the 1960s, stealth technology, infrared sensors for night vision, Tomahawk cruise missiles, unmanned aerial vehicles and GPS technology.

The list goes on.

The overall goal is advancing military capabilities by funding ideas so experimental that private industry would likely not finance them.

The hypothetical nature of the research means that most projects do not lead directly to a product.

Conversely, any ideas successfully researched and turned to practical use usually pay off exponentially.

“Underlying all this work is military need, but what DARPA has done is create a fundamental research program beneficial to the entire country,” said professor Terri Fiez. “For years, DARPA has supported research that has helped the U.S. economy become successful.”

Fiez, director of electrical engineering and computer science at OSU, works with graduate students such as Jim Le on a project of integrated circuit design to condense an entire computer system onto a single chip.

The project attempts to reduce interference between digital and analog circuitry, and create software that can predict what problems will be encountered in a given design.

The technology has already been licensed to a California start-up company, Clear Shape Technologies.

Le, several months from completing his graduate work, said he’s looking at employment with technology powerhouse Intel Corp.

Fiez said that in a time when graduate research funding has dropped dramatically, DARPA money has not only paid the way for many new students to receive hands-on training, but has also bolstered the school’s engineering programs.

OSU is probably the number five school in the nation for analog mixed signal research, she said, and Dietterich’s artificial intelligence program is one of the top projects of its kind in the country.