Dan Kaufman, director of the Information Innovation Office at DARPA, thinks that everyone needs to be much more concerned about cyber hacking — and that Silicon Valley’s corps of 25-year-old app developers should dream a little bigger.
Kaufman, a Silicon Valley lawyer who was previously chief operating officer at DreamWorks Interactive, spoke today at the Code Conference at the Terranea Resort in Rancho Palos Verdes, Calif. He outlined his four main projects and shed light on DARPA, the government institution that invented the Internet and has since developed and sold the technology for such innovations as self-driving cars and Siri.
As our homes and cars get connected to the Internet, our risk for cyber attacks only grows, he said in an interview before the conference. “The first is HACMS.”
“Oh, we’re the worst marketing people in the world — my office did Siri, but when we had it, we called it ‘programming assistant that learns’ — PAL,” Kaufman said. “To make it worse, because we don’t even understand our own acronyms, we called it ‘AL,’ because the ‘P’ is silent, I guess.”
DARPA employees only remain with the unit for four years — their badges have start and end dates printed on them. This, Kaufman said, creates a sense of creativity and brutal efficiency. There’s also a huge amount of money for smart projects: At any point, there are 100 employees at DARPA — and it has $3 billion a year.
“Everyone talks about, ‘Oh, I wish I had the money to actually do something great,'” he said. “And that’s what’s awesome about DARPA. You do a pitch to us — 30 minutes — and you could walk away with $50 million. After I pitched, I went from absolute excitement to abject terror in about 15 seconds.”
Kaufman said he wanted to speak at Code because not enough people know what DARPA is.
“Maybe somebody out there is bored of writing the next app, or maybe a random 25-year-old wants to come and build something actually amazing,” he said.
He outlined four programs:
High-Assurance Cyber Military Systems (HACMS): A program to build super-secure computer chips.
“We, like you, are excited about the Internet of Things,” Kaufman said. “But without a fundamentally new model, we won’t get to enjoy it. A computer is the only expensive appliance that comes broken. At the store, they tell you to go home and patch it, and it’s still broken, and you have to patch it again on Tuesday,” he said, referring to the constant need to update our computers to battle the latest iteration of cyber attackers.
“With the Internet of Things, I can’t imagine what it will look like with your refrigerator or your car — or in my case, can you imagine telling the Navy, ‘Hey, I know we’re going to war, but if you could come back on Tuesday … ‘”
So, what’s the plan?
“Inside all these devices everywhere are these little embedded chips running operating systems. We could make these so they’re provably secure.”
Why don’t these already exist?
“You mean, why is DARPA so goddamn slow? I don’t know,” he said. “We just made this up two years ago. No one knew these would be possible at all before that.”
Cyber Grand Challenge (CGC): DARPA’s fully automated computer-security challenge.
“Even if HACMS succeeds, creating software that can beat the best hackers in the world in real time is hard. Before you laugh, remember everyone laughed when we said a computer could beat a human at chess,” Kaufman said. “When we get good enough, a computer will beat all the hackers one day.”
And this year?
“If we finish not-last, I’m taking a victory lap.”
Memex: To extend the reach of current search capabilities, and quickly and thoroughly organize subsets of information based on individual interests.
“With most search engines, whether you use Bing or Yahoo, you search about one percent to five percent of the Internet. You can imagine many cases where I might want to search the deeper Web. We’re applying it against the scourge that is human trafficking. There will be huge commercial benefits, but I leave that to others.”
Big Mechanism: A project to develop technology to read research abstracts and papers to extract pieces of causal mechanisms, assemble these pieces into more complex causal models, and reason over these models to produce explanations.
“Imagine the following — the cure for cancer exists,” he said. “The problem is that it’s spread out in hundreds of thousands of different articles in different languages. What if I could create software that could read all of those articles, put together a causal article, and bring it back to me?”
What does Big Mechanism stand for?
“It’s not an acronym!” Kaufman said. “Which you must understand is a massive government win.”