Drive.ai, a startup focused on self-driving car technology, is looking to teach driverless vehicles to communicate with humans using deep learning, which is an advanced form of artificial intelligence.
The startup, which was founded by former graduate students under the Artificial Intelligence Lab of Stanford University, only recently revealed itself and its plans, along with the addition of former General Motors director Steve Girsky as a board member.
Drive.ai has been able to raise funding worth $12 million from a venture capital company and strategic investors. It was forced out of stealth mode back in April when the California government awarded the startup with a testing license for self-driving vehicles in the state, but it was not clear what exactly Drive.ai was working on.
It has now been revealed that Drive.ai is developing an autonomous driving system that will utilize deep learning in every aspect of the technology. The self-driving vehicles will be given a large number of sample situations and objects, with the system then extrapolating the learnings that it acquires from these samples to decide on what to do when it encounters new experiences.
The technology that Drive.ai is developing will still require a heavy amount of testing to acquire basic information, hence the application for a license in California, but Drive.ai president and co-founder Carol Reiley believes that such testing will help the development of the startup's system as it will be able to encounter more unique cases.
According to Reiley, Drive.ai is using deep learning in its self-driving technology not just for functions such as detecting objects, but for the vehicles to make decisions on what would be the safest for their passengers and the rest of the cars and pedestrians on the road.
"We think that deep learning is the definitely the key to driving because there are so many different edge cases," Reiley said, adding that the startup has seen millions of cases. Such a large number of situations would make it impossible to write a rulebook that a self-driving car could follow, so using deep learning to allow the vehicle to make its own decisions based on what its sensors pick up would be the best way to address any situation.
"Deep learning is the best enabling technology for self-driving cars," Drive.ai CEO and co-founder Sameep Tandon reiterated, adding that while sensor input is important for the vehicles, what is really needed is a brain that will allow the self-driving car to travel safely by understanding the environment.
The platform that Drive.ai is developing will also allow the self-driving cars to communicate with humans through an electronic billboard strapped to the vehicle's roof. The billboard will communicate through text and emoji for messages such as "safe to cross" when the vehicle stops to let pedestrians cross the street.
This not only gives the people and other vehicles around the self-driving car the reassurance that they are detected by the vehicle, but also gives the self-driving car a personality that could make others feel safe and welcome.
The technology that Drive.ai is developing might not find its way into private vehicles, so the startup is aiming to deliver its product to public transportation, including ride-sharing vehicles, along with fleets of delivery trucks.
Drive.ai enters an increasingly crowded industry that is looking to release self-driving cars and the technology that powers them. Google, Tesla and Uber are just a few of the big names that are involved in the race, with startups such as nuTonomy, which recently started testing a self-driving taxi service in Singapore, also growing in number.