Nexar Co-Founder and CEO
By: Joe Gomez
Locate | Pier 27, San Francisco, CA | May 30–31, 2018
At Locate, we’re bringing together thinkers and makers in the mapping space — people who are building new things and solving tough problems. We’re excited to have Eran Shir, Co-Founder and CEO of Nexar, to speak at the event. Nexar is using computer-vision and inter-vehicle networking (V2V) to make driving safer.
Reality is rich, complex, and ambiguous. There are so many nuances in navigating the world that we process instinctively as humans… Humans can interpret maps that are 90% accurate. For vehicles, 99.999% accuracy may not be enough. -Eran Shir
One of the themes at Locate we’ll be exploring with Eran is creating Maps for Machines. Humans have always been in between the map and the real world (or the“join” between the visual map and reality). But now, humans are increasingly being removed from the equation which changes the requirements for maps in terms of latency, accuracy, and connectivity.
As we gear up for a two-day party with all of the most interesting people and companies in the mapping space, Eran gave us a preview of what’s on his mind:
As AI-based systems progress, how do we make driving smarter and safer?
Driving will be safe when all drivers have advanced warnings of dangers on the road; a 10 second heads up about every potential collision. To get there, we need to know in real time the trajectory and state of every vehicle, every pedestrian, every traffic light, and generally all objects on the road.
There are three key pillars required to enable this: 1. AI-powered perception that allows us to actually detect all of these objects and understand their state. 2. Sensor fusion that allows us to localize all of these objects in the world with high accuracy, and 3. A massive live network that can bring all of this data from all of the different vehicles and nodes together and unify them to create a holistic view of the road.
What challenges are top of mind as we talk about engineering “maps for machines”?
The biggest issue is interpretation. Reality is rich, complex, and ambiguous. There are so many nuances in navigating the world that we process instinctively as humans. For machines to be able to replicate our abilities and navigate the world in a trusted way, we need to create maps with a deep level of detail that will remove the ambiguities. We also need to train our machines to develop intuition, instincts, and reflexes that we naturally possess. Humans can interpret maps that are 90% accurate. For vehicles, 99.999% accuracy may not be enough.
What are you personally most excited about exploring at Locate?
I’d love to explore the convergence of AI and live use cases. I’m asking myself, what kind of new products can we build in a world where we understand everything on the road in real time?
We’ll be announcing more speakers, demos, and code labs at Locate in the weeks to come. Register for early bird tickets now to save your spot. You can also sign up to race autonomous Robocars at the event. More details here.
Speaking @ Locate: Eran Shir was originally published in Points of interest on Medium, where people are continuing the conversation by highlighting and responding to this story.