You can talk to a computer, but DARPA wants a real conversation

People have gotten used to the idea of talking to their technology, whether it’s their smartphone, voice-activated search engine or a speech recognition program. Often, as in the case of Apple’s Siri, the technology even talks back, asks questions and learns things about you.

But those algorithm-driven exchanges, as far as they’ve advanced in recent years, still fall far short of the conversation you can have with another human, which can be influenced by context, familiarity, shared experiences, humor, anger or any other number of factors. A computer would have a hard time interpreting sarcasm, or a knowing reference to something said earlier or a gesture than says more than words.

Nevertheless, voice interaction is big part of the future for computing, and it’s the idea behind the Defense Advanced Research Projects Agency’s new program, simply titled Communicating with Computers, or CwC.

"Human communication feels so natural that we don't notice how much mental work it requires," Paul Cohen, DARPA program manager, said in a release. "But try to communicate while you're doing something else—the high accident rate among people who text while driving says it all—and you'll quickly realize how demanding it is."

As demanding as it is for humans, though, it’s a lot more difficult for machines. So with CwC, DARPA wants to take human-computer interaction to the next level, with the eventual goal of getting to where computers can work with humans on problem-solving in areas ranging from medical research to national security.

The program will start by setting humans and computers to work on two tasks that require them to work together. The first will focus on storytelling with a kind of “Naked Came the Stranger” approach, as a human and a computer take turns writing the sentences of what will add up to a short story. If nothing else, the result could be entertaining.

The second task is more in line with what DARPA hopes to accomplish, with a computer and biologists building models of the processes that cause cells to become cancerous. Computers are building those models now as part of DARPA’s Big Mechanism program, but the process isn’t collaborative with biologists. CwC wants to find a way to combine the strengths of humans and machines—computers’ speedy reading and numbers-crunching with humans’ knowledge and judgment—to make the process more effective.

“Because humans and machines have different abilities, collaborations between them might be very productive. But today we view computers as tools to be activated by a few clicks or keywords, in large part because we are separated by a language barrier,” Cohen said. “The goal of CwC is to bridge that barrier, and in the process encourage the development of new problem-solving technologies."

This isn’t the first time DARPA has tackled the  a speech recognition system whose 1,011-word vocabulary represented a great leap forward at the time. Verbal interactions with computers have come a long way since then. In DARPA’s eyes, though, it still has a long way to go.