The Center for Data Innovation spoke with Eran Shir, founder of Nexar, an AI dashcam company based in New York City. Shir discussed how artificial intelligence can help dashcams make driving safer, and the potential benefits of vehicle-to-vehicle communication networks.
Joshua New: Dashcams are already pretty popular. What kind of value does AI actually add here?
Eran Shir: The key value of bringing AI techniques such as deep learning into dashcam applications is the introduction of perception. Once you can actually understand what’s happening outside and inside a car, you can start reacting to it. You can provide the driver with early collision warnings to prevent collisions. You can detect free parking spots or road hazards and notify everyone else on the network. You can coach the driver on how to become better. In essence, you can digitize the public space, which results in a boundless set of new capabilities.
New: What can Nexar recognize? Is it only obvious visual cues, such as a car swerving, or can it learn to recognize more subtle changes over time?
Shir: Nexar can recognize what we’re teaching it, which is growing every week. At the moment, we are focused on everything that has to do with driving, and in particular, safety: vehicles and their behavior, such as cutting off and braking, traffic signs, traffic lights, lanes, and pedestrians.
New: What are the vehicle-to-vehicle (V2V) component of Nexar, and what are the potential benefits of this technology if it were more widely used?
Shir: In a V2V network such as Nexar’s, every Nexar vehicle doesn’t keep to itself what it does and which cars it sees, but rather shares that information in real time (~100ms round trip latency) with other Nexar vehicles in its neighborhood. With that, Nexar can warn drivers about a car doing a hard brake five cars ahead, even though there’s no line of sight to that car, and no sensor onboard the car can detect the car ahead. It can warn a driver about a car coming into the intersection behind the corner on a collision course. The National Highway Traffic Safety Administration estimated in 2014 that as much as 80 percent of collisions could be prevented had there was a mass adoption of V2V.
The concept of V2V is not new and has been around for almost two decades, but always in lab and testing, rather than commercially, because there was no path to create enough network density to justify the deployment. At Nexar we managed in a span of two years to build and deploy the first commercially viable V2V network in places like New York City and San Francisco and started providing this extra layer of safety.
New: I imagine this technology has a lot of valuable insurance applications, such as verifying a claim or identifying good driving behavior. Are you exploring any of these applications?
Shir: Indeed we are, and we are already in partnerships with several insurance companies across the world. We’ve already saved millions of dollars to drivers who receive insurance discount because of them being a Nexar user.
New: As cars gradually adopt more advanced sensor technology and there’s less need to use a separate device as a dashcam, what’s the plan for the future of Nexar?
Shir: Nexar’s plan is to be the “road’s air traffic control”. The mesh network that synchronizes all vehicles, predicts dangerous or inefficient road situations, and prevents them.