Drive.ai, a start-up born out of Stanford’s artificial intelligence lab, has officially entered the driverless car arms race.
But Drive.ai has no interest in building an actual, self-driving car. Rather, as co-founder and president Carol Reiley told Business Insider, the start-up plans to sell “the brains of the car.”
That car brain comes in the form of sensors, LiDAR, and Radar like many other companies are using, but with one important addition many aren’t using: deep learning software.
“We definitely think the car is a computer on wheels, and would love to really build the brains behind it and figure out what the sensory inputs are,” Reiley said.
Drive.ai’s product is a roof kit with “the brains of the car” that can be retrofitted to any vehicle, from a truck to a golf cart. Reiley said the start-up has official partners with auto companies, but declined to say how many companies or which ones.
Drive.ai will release fleets of its cars on select routes for package delivery, ridesharing, and public transit, as part of its unnamed partnerships within the next few months, Reiley said.
The start-up has raised $12 million from early stage venture capital fund Oriza Ventures. Steve Girsky, who sat on General Motors’ Board of Directors for seven years, is sitting on the board of directors for Drive.ai.
Building the car’s brain
Deep learning is a branch of artificial intelligence that allow computers to learn on their own.
Because there are so many different scenarios that driverless cars face, it’s impossible to code rules that will safely govern how a car operates.
“What deep learning does is not micromanage what the rules are, but gives you the context of different examples and you start learning,” Reiley explained.
Most driverless car projects, like Google’s, rely on a combination of sensors, LiDAR, Radar, high-resolution maps, and GPS that allow autonomous cars to locate where they are, sense their environment, and then drive safely.
But Drive.Ai wants to add deep learning to the mix so cars can better deduce what’s around them and make reactive decisions faster. “You don’t want the car to sit there trying to think ‘what is this’ as opposed to ‘what can you safely do?’ Reiley explained.
Reiley explained that Drive.ai’s focus on deep learning software is what really sets it apart from other companies in the driverless car race. But there are others interested in applying advanced forms of AI to autonomous vehicles.
Drive.ai was the 13th company in California to test its driverless technology on public roads, Reiley said, adding that the start-up is “already in the top 5 [for] cars in most miles driven.” But Reiley declined to disclose how many miles its cars have driven in autonomous mode. She did say Drive.ai has an existing fleet of “under a dozen” vehicles that have been used for testing.
Communicating with pedestrians
Another way Drive.ai plans to set itself apart is in its communication system for pedestrians. The roof kit will also feature an LED sign that allows the car to talk to pedestrians using sound, text, and even pictures.
“Most people’s interactions with a self-driving car will not be that they own the self-driving car, but they’re going to be one of the pedestrians walking by and they need to understand what is happening with this car,” Reiley said.
But Drive.ai isn’t the only one looking at ways to improve the communication gap between driverless cars and pedestrians. Google filed a patent for a driverless car system that will display messages to pedestrians at crosswalks, and a BMW self-driving concept car features a system for visually communicating messages to pedestrians.
What will most likely determine if Drive.ai can get its footing in the competitive space of driverless cars is its timeline for rolling out the product. Reiley said the start-up is definitely planning on getting its tech on the market before 2020, a timeline most companies are operating on.
Although Drive.ai plans to roll this out through its partners, it’s not out of the question that it could hit the consumer market down the road.
“Obviously, long term, especially with the automakers, we would love, love to get this in your hands and that you could buy this type of technology,” Reiley said.