Lots of CEOs in the AI space show off examples of robots that look and sound a bit like human beings. Suzanne Gildert, CEO of Sanctuary AI, may be the first to show off one that is designed to look exactly like herself.
Speaking at the Canadian Innovation Exchange (CIX) conference in Toronto on Monday, Gildert’s robotic doppleganger was displayed on a screen at the end of her keynote where she thanked the audience for their time, right next to where the Sanctuary AI founder stood on stage. The robot wore the same glasses, the same long hair, and had a trace of the same Australian accent.
Gildert made clear that the robot was not intended as an exercise in narcissism but a combination of art project (she called it “the ultimate self portrait”) and a way to start thinking about the kind of “robot rights” superintelligent machines may need once they start to make their way into offices and society as a whole.
“It’s a way of seeing that these entities might go through when they start coming to life in the world,” she told the CIX audience. “I want people to treat this robot well, and to see that it inherits my values.”
Whereas some firms may downplay the similarities between robots and humans, Gildert said Sanctuary AI’s mission is to make them closer to what has traditionally been depicted in science fiction films, similar to the work of firms like Hanson Robotics. In other words, robots that look and behave almost exactly like their creators. Doing so will not only offer new use cases for AI and robotics, Gildert said — it may mean that “onboarding” them into their jobs will be a lot smoother.
“Our world is designed for humans, whether it’s opening a doorknob to sitting on an airplane,” she pointed out. “We can change existing infrastructure, which may require trillions of dollars to match the form factors of robots and AI currently, or make them like we are now so they can take advantages of all those complementary resources.”
Besides eliminating the need to design working environments around a robot, Gildert said the Sanctuary AI approach will offer a platform of sorts to advance research in areas such as sequence learning, computer vision, cognitive architectures and sensor fusion. Creating more realistic androids will also help them learn better themselves and communicate more effectively, she suggested.
“If you are a human, you understand the world through experiential learning,” she said. “If you hold a cup, it’s not just a word that an AI program would understand. It’s a physical object that you interact with, you feel it, you hold it, you understand that it can spill. All those experiences give us meaning behind the words we use in everyday language.”
Such robots, or “synths” as Sanctuary AI calls them, need not be cost-prohibitive thanks to other recent advancements in technology, Gildert said. The rise of 3D printing, for instance, means a life-sized form can be generated in a day and will be lighter than aluminum. Generative design in CAD systems, meanwhile, mean engineers can simply indicate certain parameters (like rounded corners) on what they need created to have parts developed quickly. Over time, Sanctuary AI expects such robots to cost about the same as a car, she said.
Although Gildert said her firm is still exploring its product-market fit, some potential areas include areas such as wellness and elder care, where the synths Sanctuary AI builds can do more than repetitive tasks but offer social and creative value in a variety of human interactions. In the meantime, one of its robots, dubbed Nadine, is already being used at Science World B.C. in Vancouver in educational sessions with children.
“We’re trying to strike that balance between not wanting to take way people’s jobs but also looking at what’s socially good,” she told B2B News Network. “We want to focus on places where there’s a shortage of people to work and where the interaction with humans is very caring and compassionate to work on these reward functions.”
Latest posts by Shane Schick (see all)
- Unbounce product director says Smart Traffic will assess five factors to boost conversions - November 20, 2019
- Why are COOs the forgotten B2B marketing target persona? - November 19, 2019
- ActiveCampaign VP predicts its machine learning features will offer an alternative to A/B testing in e-mail marketing - November 15, 2019