It was a typical Silicon Valley gathering. On a sunny Saturday everyone was indoors discussing arcane technical subjects. The event was the Artificial Intelligence (AI) Unconference, sponsored by Intel at its Mission College Blvd. headquarters.
The “un” part of the conference is the “pop up” nature of the sessions. It is, to borrow a term from AI, a self-organizing system. The schedule is decided on the spot in a process that’s surprising when you consider that this is a crowd of people, any one of whom in high school probably wrote a program for optimal scheduling.
Anyone can suggest session topics by writing them on a card and people vote for them if they’re interested. The schedule is then decided by laying out all the cards on the floor and organizing them in tracks. Then plain old human intelligence decides the best order.
One of the presenters was Geeta Chauhan, CTO Silicon Valley Software Group, a company that provides “CTO as a service,” as Chauhan describes it.
In a sea of Sheldon Coopers, Chauhan would stand out as one of the few women in software engineering. But she would stand out on her own for her enthusiasm for making the arcane simple, and her “anyone can do it” approach to subjects most people think are only for the most brainy.
Her non-techy approach to technology—her presentation was titled “How to Build a Deep Learning Network in Less Than a Week”—was illustrated by her talk on how to build a cat door with off-the-shelf components that will recognize your cat and only your cat.
Born and educated in India, Chauhan has a BS in software engineering from Delhi University and an MS in Software Applications from the University of Pune. Chauhan has 25 years of experience in software development, including as Alcatel-Lucent’s Director of Advanced Technology and Advanced R&D.
Alcatel-Lucent (now Nokia) was where Chauhan first became interested in artificial intelligence. “We had lots of data, but we wanted to find insight,” she said. She managed a project to develop software to provide this analytical insight and this started her on the road of software that can learn.
As arcane as AI can sound—Hidden Markov models, multi-layer perceptrons, restricted Boltzmann machines—Chauhan keeps her focus on the goal, not the journey. “We need to be solving user problems, solving the pain points. The technology should be hidden.”
When it comes to the absence of women in technology—outside the marketing and HR departments, that is—Chauhan says that the problem starts in school. “What I see is a social stigma. You’re a ‘geek.'” Social media, she said, makes the anti-engineering peer pressure worse.
She also points to pressure on young people to focus on resume-building rather than experimentation. “If you’re interested in something you should be able to try it. It’s OK to fail. There is so much early pressure and expectation. And many people are afraid to try things because they might fail.”
Education should be about increasing, not stifling, experimentation, Chauhan said. And that gets back to the cat door.
“It’s a simple, interesting problem,” she said. “People see, ‘Oh I can do this.’ I break it down into simple steps.” And if people can see how it’s put together, she said, they can see themselves doing it.
“Some schools are encouraging students to do more hands-on,” said Chauhan. “Things like the ‘maker movement’ are encouraging youngsters to do things by themselves instead of ordering it from Amazon. How else can we expect people to solve problems unless they feel comfortable experimenting?”
