Tesla’s head of self-driving admits ‘lagging a couple years’ behind Waymo

Tesla’s head of self-driving has admitted that the automaker’s autonomous program is lagging “a couple years” behind Waymo, but he believes the cost advantage will enable it to scale faster.

In a rare candid interview, Tesla’s head of AI and self-driving, Ashok Elluswamy, has admitted that Tesla is a couple of years behind Waymo on the autonomous driving front.

The interview can be hard to follow for English speakers as both Elluswamy and the host switch from English to Tamil frequently, but you can clearly hear the Tesla VP says that Tesla is lagging behind Waymo when talking about Waymo’s different approach:

When asked about the difference between Tesla and Waymo on self-driving, Elluswamy says that Tesla’s approach is much cheaper. The host asked if he means it is less expensive but “equal quality” and the Tesla VP answers:

Equal quality. Technically, Waymo is already performing. We are maybe lagging by a couple years.
This should be obvious to anyone following closely since Tesla has yet to be able to do what Waymo has been doing for years: provide customers with level 4 autonomous driving rides.

Tesla has been limited to a level 2 advanced driver assist system (ADAS), which requires constant supervision from the driver.

Nonetheless, it is a rare admission from Tesla as its CEO, Elon Musk, has been minimizing Waymo’s achievements for years and claimed that he doesn’t see anyone close to Tesla on autonomy.

That’s even though Tesla only plans to finally start offering level 4 autonomous rides to customers next month in Austin, while Waymo has been doing that for years, including in Austin specifically, since earlier this year.

It’s true that Tesla’s vehicles are much cheaper than Waymo’s, but there are many reasons for that.

The cost of lidar sensors has been one of the top suspects. Costs have come down quite a bit, and it is not really a problem anymore, but they are more power hungry than Tesla’s sensors, which are just cameras.

The real difference in the cost of the vehicles is the fact that Tesla produces over a million cars a year, versus Waymo producing a few hundred units now and a few thousand units soon. Waymo also buys the vehicles from other manufacturers and simply integrates its sensor suite and hardware.

Tesla benefits from economies of scale, but that’s because it sells those vehicles to customers who, in the vast majority, do not buy Tesla’s Full Self-Driving package since it doesn’t do what the name implies.

In the upcoming pilot program in Austin, Tesla plans to use the same vehicles it delivers to customers. It will use different software that has been optimized to work in a geo-fenced area of Austin and it will also be supported by teleoperation, but the hardware is going to be the same, which does reduce costs.

Electrek’s Take
Right now, I think the cost of operating limited autonomous ride-hailing fleets like Waymo’s has little to do with the vehicles’ cost.

I think it is more related to the training and the support, specifically the level of teleoperation. If you have a 1:10 ratio of one teleoperator to 10 cars, it is going to be much cheaper than a 1:1 ratio of teleoperator to car.
We know that Waymo uses some levels of teleoperation and that Tesla plans to use a “high level” of teleoperation, but we don’t know the specifics of each program.

These, along with the training of specific regions and regulatory approvals in some jurisdictions, will be the main limiting factors.

Considering Waymo has a system that already works, it is currently completing over 250,000 paid rides per week, it already is operating in 5 markets, and it is both expanding the geo-fencing areas of those markets and expanding into other markets with more vehicles, I think it’s clear that it is ahead of Tesla in autonomous driving.

Tesla is now going to start catching up to Waymo next month with its first market and its first 10-12 vehicles.

For now, I haven’t seen serious evidence that Tesla can scale faster than Waymo. The only real advantage is the availability of the vehicles to deploy in the fleet. Tesla has plenty of those lying around, but that’s hardly a major bottleneck for Waymo.

The only way Tesla could leapfrog Waymo is by deploying level 4 autonomy in its customer fleet as promised for years, but I don’ see that happening anytime soon.

I think that the only way Tesla can safely deploy level 4 in an internal fleet in Austin next month is through mapping, geofencing, and high level of teleoperations, maybe even 1:1 teleoperation. I’d be happy to be proven wrong though.