"I figured out long ago what I'm good at and what I'm not," she says. "I feel I have skills in running companies and the mechanics of that, but I'm never going to think up the next big idea. Jeff is good at that, but he's not good at running companies. So my career is building companies around Jeff's big ideas."
Together, Dubinsky and Hawkins made Palm into the first really successful PDA. At Handspring, with the Treo they pioneered the first smart phone. Now they are trying to make a breakthrough in artificial intelligence by teaching computers to recognise patterns, the thing humans are best at.
"A two-year-old can reliably point at a dog, but a computer can't. Lots of applications could be built on basic image recognition." The key, is creating a computer that can generalise from a single dog to the infinite variations of dogs.
Numenta is privately funded by Hawkins, Dubinsky, and a few colleagues. They are deliberately keeping the team small and the burn rate low. "Then we can fund it for quite a while." They may look for venture capital funding later on; the company has some limited revenue from support services.
The software the company is building - and providing to others to work on - is based on algorithms that Dubinsky describes as "a cross between computer science and neuroscience". They are, she says, patterned after the human brain in order to create learning architecture that is robust and scalable. Numenta wants to be a platform company, providing the tools others can use to solve the really hard problems that today's computers can't cope with. "There are so many people coming to us with interesting problems," she says. "For example, pharmaceutical companies, trying to see patterns in drug testing data. Transport companies - can they make cars smarter? A couple of game companies, trying to inject some reality and real-world stuff into games."
Dubinsky and Hawkins met when Dubinsky was freshly back in the US after a year's sabbatical in France.
"I came back wanting to be CEO of a company," she says, "and I was looking for a product partner." Someone who was interviewing for vice-president of engineering at Palm told her he was uncertain about the job because they didn't know who the CEO was going to be. Dubinsky went out networking.
By that time, she already had substantial experience near the top. A Harvard MBA, she had worked at Apple and was a founder of Claris.
"I felt prepared for doing a start-up as CEO. The only thing I didn't have was running engineering and product development." That was timing - Dubinsky was born in 1955. "I think if I were growing up now, or if I'd been a boy then, I'd would have been a much more technical person, but I was never offered the opportunity. In high school, the girls took sewing and the boys did wood shop."
She learned technology on the fly in her various jobs. The brain science behind Numenta was not, she says, an interest of her own. "But the first day I met Jeff he explained that this was a huge interest of his, and one of his goals in Palm was to make it big and successful so he could go work on his brain stuff. Through the years, I've followed his work in this area very closely - I helped with the drafts of his book and helped set up Redwood."
Hawkins created Redwood Neuroscience Institute in 2002. "Numenta has the world-changing potential that everything I've been involved in has had. That's why I'm doing it. I feel like, wow, it may take some time, but I think that what we're doing here is going to have a significant impact on the world of computing." µ