DeepMap’s Richard Lucquet in Conversation with Frank Chen of a16z
Richard Lucquet of DeepMap and Frank Chen of Andreessen Horowitz held a fireside chat on Feb. 28 at the Autonomous Vehicles Silicon Valley conference in Redwood City, Calif.
The pair discussed a wide range of topics, from “Will the infrastructure for autonomous vehicles (AVs) evolve from inside out or outside in?” to “In the future, will high-performance race cars still be manufactured for human drivers?”
In this first post in a series, we highlight Frank’s response to Richard about the economics of autonomous driving, and Richard’s answer to Frank about the importance of high-definition (HD) maps.
* * *
On the Economics of Autonomy
Frank Chen: Most of you probably got here in a car and at home you probably own and operate a car. The question is — how much does it cost for you to own and operate that car? And the answer is — depending on how fancy a car you have — it’s going to be anywhere between a buck to two bucks a mile for you to own and operate it.
Then the question is — at what point will AVs offer a service that’s cheaper than that? What if I gave you the option that for two bucks a mile, you could drive yourself anywhere versus for 50 cents you could be chauffeured everywhere? I think most of us would say chauffeuring sounds pretty good, if it’s a quarter of the cost.
The operating costs of getting from point A to point B are literally either cut in half or a quarter. And as your user experience gets better, we’re going to have a mass market shift towards what people are calling Transportation-as- a-Service.
If that’s the case, then what happens is that the fleet operators who are operating that service, such as Lyft or Uber, are the big beneficiaries. And you will see a giant sucking sound of revenue moving from the car manufacturers, the car insurers, the used car market, etc. to the fleet operators.
That would be one of the biggest transfers of a revenue pool and a profit pool that we’ve seen in many, many years.
On the Importance of HD Maps
Richard Lucquet: Despite what some high-profile experts might proclaim, the industry definitely needs HD maps! In order to have accurate self-driving, you have to get the car into a position where the driving decisions are informed by more than real-time systems.
Not too long ago, we were introduced to “foot-free” driving, with automatic cruise control. More recently, with autopilot, we’ve moved to “hands-free” driving, where you can ride in the car and have the car do what your hands would be doing.
But today, we still need to be very alert, with our eyes looking around and taking in inputs. We think of the next step in autonomy as “eyes-free” driving.
In order to achieve “eyes-free,” the automated system that drives the car needs to be very aware of the environment. It’s difficult today for a perception system to have a real-time understanding of the environment and make real-time decisions very quickly. It cannot wait for even one second before it makes a decision.
So, bringing in an HD map into the decision-making loop is fantastic because now you have a map that can “see” through a turn or around a corner, which the camera or Lidar cannot. The map effectively serves as the “memory” of the car.
Picture yourself in a new city, where you’ve never been. You’re not the safest driver out there if this is new territory for you. You know the rules of the road, but you are looking for the signs and the traffic lights, etc. But what if you have the ability to have an input — the HD map — that already has a very detailed description of the environment. It will dramatically increase safety.
And remember, an HD map is something that can be created offline with lots of computer power. If the computer has any doubt about what to describe, a human can step in to compensate for the deficiencies of the computer because it’s all done offline.
Once this highly-precise map is completed and loaded onto the car, then the car has an additional, powerful input to help make the right decision. We believe that HD maps are very, very critical to the next stage of autonomy.
(Thank you to Asmita Wankhede for her contribution to this post.)