Trailblazers 2019: Sanctuary AI's Suzanne Gildert aims to develop human-like intelligence in robots

    1 of 1 2 of 1

      Scientist Suzanne Gildert’s dream is to create robots that are indistinguishable from humans. She’s already taken the first steps towards achieving it.

      After completing her PhD in quantum physics, Gildert took a job at Burnaby’s D-Wave Systems, one of the world’s premier quantum-computing companies. Much of her work was in machine learning and AI—a field that spurred her long-standing fascination with untangling the mysteries of the human mind.

      Four years later, Gildert—who had spent much of her free time building her own robots—decided to try something previously relegated to the pages of science fiction. The result was Kindred.

      “I founded the company with the idea that we were going to explore humanlike intelligence in a machine,” Gildert tells the Georgia Straight, on the line from her Vancouver lab. “We wanted to put intelligence into robots in the same way that we think about humans having intelligence—so being able to make smart decisions, and learn from experience.”

      The venture was a huge success. Gildert hand-built over 30 robots to demonstrate how it was possible for humans to first control the machines remotely and then watch as they learned to do tasks themselves, including complex assignments like choosing and sorting items in a warehouse. The company grew to over 50 employees and had offices in Vancouver, Toronto, and San Mateo, and boasted $50 million in venture funding. But for Gildert, it wasn’t enough.

      Realizing that the robots she had built deviated from her vision of creating true human intelligence, she started her second company, Sanctuary, in January 2018 to build machines that are akin to people in every way—a goal she believes will become achievable over coming decades.

      “This time it’s not just figuring out how we can put humanlike intelligence in robots, but actually building the robots to look like humans, too,” she says. “We decided that if we’re going to try and build a robot and it looks nothing like a human, it’s always going to end up doing a specific task that isn’t a human-intelligence-level task. So we have to do this properly.

      “I’m of the opinion that within 10 years we’ll be starting to build AI systems that begin to exhibit these humanlike effects,” she continues. “I don’t think they’ll be these full Westworld-like robots, but we’ll start to see little sparks fly. They’ll begin to do things that we didn’t program in, and just decide to do them on their own.”

      Kate Wilson is the Technology Editor at the Georgia Straight. Follow her on Twitter @KateWilsonSays