Here is John Krafick interview from December 2020. It is 35 mn long. He spoke with Patrick McGee, SF Correspondent for The Financial Times, where they reflected on developments in 2020, and what’s to come in 2021.
Here is my summary for those who don't want to watch a 35 mn video. Enjoy!
Can you talk about background of Waymo?
- In 2009-10, Google achieved some FSD "demos". 10 routes of 100 miles each with no disengagements.
- When looking at potential commercial applications, the original plan was a hands-free driver assist system called "autopilot" that could do on-ramp to off-ramp highway driving. In 2013, Google got beta testers to test it. Cameras monitored driver attention. Project was shut down after just 1 month because volunteers were misbehaving and not paying attention to the road. Conundrum: as driver assist gets better, humans get worse at supervising. So Google decided to focus on solving full autonomy rather than driver assist.
Are you concerned with Tesla's approach to FSD that it could harm autonomous driving in general if there are accidents?
- Waymo is not a car company. Waymo's product is a driver. Ride-hailing in Phoenix is open to everybody in the geofenced area. Just download app and fully self-driving car will pick you up and take you anywhere in the geofenced area.
- Tesla and other car companies, doing driver assist, are not competitors to Waymo. Driver assist technology can be a good thing and save lives but consumers need to understand limitations. As driver assist gets better, humans pay less attention. There is increased risk that humans "check out" just at the wrong moment when driver assist needs help.
- If a licensed driver is required, tech should be required to as a driver assist, not full autonomy. If a licensed driver is not required, then tech is fully autonomous.
What is taking Waymo so long in deploying fully autonomous cars in large numbers?
- Full autonomy is the most complex challenge that humans have ever tackled.
- Yes, it has taken time. First, driverless ride was in 2015. It was limited to less than 25 mph. It took another couple of years to go from that low speed FSD to autonomous at full speeds on 4th Generation hardware.
- Current standards do not allow cars with no steering wheels to operate at speeds above 25 mph. So "firefly" was not scalable.
- Waymo Driver is designed to work on pretty much any vehicle type.
- Developing Waymo Driver is 99.9% of the problem.
How does Waymo Driver compare to your driving skills?
- Waymo Driver is most experienced FSD. 20M autonomous miles and billions of simulation miles. Simulation is primary way that Waymo is refining the Driver.
- Waymo Driver is probably better driver than me. Humans can be great drivers but we can easily lose focus, get distracted, get tired. Autonomous driving like Waymo Driver does not have these failures.
- Waymo has tested in 25 cities over the years. Waymo is currently in 6 cities (Phoenix, San Francisco, Los Angeles, Seattle, Detroit, Miami)
Hypothetically, what would happen if Waymo One did launch tomorrow in those 6 cities? Sounds like tech could probably do it but you are not doing it yet. What would happen?
- Each city has different driving challenges. Waymo Driver does best when we have time to learn and master the environment. It takes time to make sure we are driving well in those locations.
- In Phoenix, in order to demonstrate that we were capable, instead of releasing demo video, we shared safety data and methodology for all driving. 10M autonomous kilometers in Phoenix from Jan 2019 to Sept 2020. 47 contact cases. Incidents were low speed and low damage. High confidence in safety so Waymo decided to roll out driverless to public in Phoenix area.
- Before we move to other cities, we want to have that same level of comfort in safety. We would do that before we just deployed anywhere.
What about profitability? What is cost per mile to operate vehicles? Will you undercut Uber and other taxi services on price?
- Unit economics for autonomous driving is very attractive.
- Ride-hailing revenue is currently about $2/mile. Easy to imagine path to strong margins.
- Technology cost is "company secret" but cost of Waymo Driver is significantly less than expectations. Roughly in the range of cost of cars we are using. Not extraordinary expensive technology when integrated with electric vehicle. Goal is to have 100% EV fleet.
Do you envision future where you are competing with other driverless ride-hailing services in a city? Will it be a race to the bottom in terms of prices? Will the differentiator be something else like comfort?
- Second hardest challenge is figuring out where competitors are in terms of their FSD. Need more transparency from other companies on their FSD.
- Waymo has gained a lot of experience. In 2017, we had 3 fully driverless cars. In 2018, we had 100 fully driverless cars. It took another year to go driverless for early testers and another year to open driverless up to general public.
- FSD is a long road, extraordinary grind. It is extremely expensive to do well, requires massive computing power, huge team of talented software engineers. Waymo has 2100 employees.
Who is your biggest competitor?
- Can't really say.
Have you tried say an Argo or Zoox?
- Have not been invited yet. But anyone is welcome to come down to Phoenix and ride in a Waymo.
How will you commercialize your product? License to OEMs? Are you working on "firefly 2.0"?
- Two commercial paths: Waymo One is ride-hailing, moving people in Phoenix. Waymo Via is transportation of goods. Using Pacificas to move goods too. Also will start deploying autonomous trucks in SW.
- Working with OEM's on personal car ownership model but not #1 priority. Not as much benefit since personally owned cars are only used 5% of the time. But we will do it. It will be subscription model, 6 months or 1 year. After subscriptions ends, car will go into ride-hailing network. We imagine getting 300k, 500k, up to 1M miles out of cars which will help drive down unit costs.
If you starting now what would you do differently?
- Our journey was pretty efficient. I would think about doing "firefly" or not. It did serve a purpose as an avatar for the work we were doing. No significant changes. Moving people and goods was always the mission, even during covid. Approach of flexibly deploying Waymo Driver to different vehicles is a robust approach.
Congrats on being first to launch true driverless service for public but do people in Phoenix see it as transformational? People envision future where entire cities are transformed by robotaxis? Is that decades away?
- Well over 1000 riders in Phoenix, 10,000's of rides. For them, it feels normal. We focused on regular people, not tech enthusiasts.
- That revolutionary change will start to happen more as we scale up in places like San Francisco.
Will my 3 week old need a driver license when they turn 16?
- 100% confidence that they will not need a driver's license. There will be so many different modes of transportation. She will be able to use a Waymo just about anywhere. There will be other companies offering ride-hailing by then as well. Of course, personal car ownership won't go away. There will be cars that can convert into true L4 experience. There will be subscription model.