Did you know that Waymo was trying to turn itself into a sensor OEM, selling its custom-designed sensor sets to other autonomous vehicle companies? No, neither did I. It appears almost no one noticed, actually, hence it was unsurprising when the company accepted its failure and pulled the plug on that bad idea.
Waymo is a software company that uses sensors, not an electronics manufacturing company. It has been deeply confused about what it is since it was formed as a Google X self-driving car initiative in 2009. Basically, a purely virtual company was trying to do something very, very physical and real, with real human split-second safety requirements.
The Google X guys were well out of their corporate field of comfort when they started, and thought it was all a mental problem. I wrote about this in 2015, having enough of a background in robotics and AI to understand the broad strokes of what they and Tesla were doing, and to have an informed opinion that Tesla’s approach was better.
Their cute little bubble cars were the epitome of the software-will-do-everything mindset, and the nipple on top was lidar, initially the $80,000 variety.
Over time, as Tesla racked up billions of real-world miles with its highly desireable, indestructible, test platforms that test drivers spent up to $140,000 to drive for them, Waymo was falling further and further behind. Tesla got it right on lidar too, something else I assessed and published on in 2016.
Meanwhile, Waymo fell further and further behind, with each mile its cars were driven requiring massively detailed, accurate, and up-to-date computer models in order for them to drive down a street.
Then the company got into the sensor manufacturing business, maybe because it never had anywhere near enough volume for OEMs to be interested in cutting it deals. Waymo started bike-shedding, just as BMW did with its never-to-be-used again brand new body materials for the i3 and i8, and designed its own sensors from scratch. Yeah, that’s the ticket.
And then they started thinking that their sensors were really awesome, and they could compete with massive OEMs that make sensors cheaply and reliably. More non-real-world thinking.
The company started trying to sell its road-worthy sensors to off-road farming applications and the like, but not being a manufacturing company, and not having anything particularly better than what was available much more cheaply, couldn’t get anyone to buy its non-real-world product.
Waymo is still limping along, but Google X pulling the plug on it was the end of large cash infusions from Brin and Page. Like another Google X head-scratcher, Makani airborne wind power, it’s been heading downhill to its eventual dissolution and fire sale of assets since.
Meanwhile, Tesla’s solution continues to be so much better than any other company’s that it gets used a lot every day everywhere there are Teslas. And they’ve recently made it even simpler, cutting it down to just visual cameras, albeit ones that can see in every direction, don’t blink, and don’t get distracted, making them far superior to human eyes. In talking with my visual machine learning expert friends, sensor fusion is a real hassle, so there’s a strong argument for this in terms of ‘simplifying’ probably the second most widely used neural net technology in the world. (Apple’s iPhone face recognition security has the edge in sheer numbers, and I can’t think of anything else in the range of millions, but inform me in the comments if I’ve forgotten something).
As the not-so-humorous meme goes these days:
With the dramatically higher death rate in 2020, overall fatalities increased to 38,680 from 36,096 in 2019, even though miles driven decreased by 13.2% due to the pandemichttps://t.co/Ic8E6Vn3vf
— Not_an_Analyst (@facts_tesla) August 19, 2021
Autopilot is vastly safer than the alternative. Sandy Munro — automotive journalist, test driver, trusted source of reality, and someone who opted for a Tesla with self-driving — made it really clear recently how far behind everyone else was.
Every time there’s a bake-off comparison between Tesla and any other self-driving product, Tesla always wins, typically by miles.
So yeah, Waymo was formed on incorrect premises, used the wrong approaches, flailed, and is failing. Selling sensors that were too expensive so no one bought any was par for the course.