HUB CO. TAKES AI TO EARTH
Adapts Mars Rover software
A Boston startup has raised $14 million to take artificial intelligence developed for the Mars Rover down to earth to make terrestrial products autonomous.
“We envision a future where most of the physical world surrounding us could be fitted with sensors and processors to make it aware and useful,” said Max Versace, chief executive of Neurala. “All the machines today that are stupid machines will be aware and have brains.”
Boston-based Neurala said yesterday it has raised $14 million in venture funding to expand its artificial intelligence software to be used in everything from drones to self-driving cars to toys. Neurala is already working with a toy maker on a doll that will get to know its owner.
“When your daughter walks into the room, and the doll sees your daughter, the doll can get really happy and do really nice things,” said Roger Matus, VP of products and markets.
The $14 million in new funding comes from Sherpa Capital, Motorola Ventures and others.
The company’s technology was originally created for NASA to help the Mars Rover autonomously navigate unfamiliar terrain. Because Mars is too far away for the software to be boosted by servers on earth, the company had to develop an AI system that would work entirely on the rover. That same principle will help consumer products not only work better, but be more secure, the company says.
Unlike most virtual assistants, such as Siri or the Amazon Echo, products using Neurala wouldn’t need to be connected to the internet or send information to remote servers, so private information or images of a young child’s face stay on the device and nowhere else.
Along with toys — the company is already working with a Fortune 500 company on a home robot that can do things like recognize a beer — Neurala is working with two of the top consumer drone companies so a drone can follow a skier or bicyclist or hover in the right place for a selfie; one of the top five automakers on self- driving car technology; and with Motorola on software to make security cameras understand what they are seeing better.
The software is what’s called a neural network, a type of artificial intelligence that is designed to mimic how a brain works. The system is able to learn on its own, recognizing and adapting.
“You take an ordinary camera, see the world, understand what’s in front of it, and enable something to act, whether it be a drone for inspection, a security camera, a self-driving car, or a doll that is happy to see you,” Matus said.