History & approaches
The Google Approach: All or Nothing
Google's self-driving car project, now Waymo, pioneered a strategy of skipping the partial, human-supervised steps of the SAE ladder. After early tests showed that people cannot reliably supervise a car that mostly drives itself, the team committed to a fully automated driver with no human fallback. The stance is often summed up as all or nothing.
Where it came from: the DARPA Challenge
Google's effort grew directly out of the DARPA Grand Challenge, the desert and city robot races that many consider the big bang of the modern self-driving car. The founding engineers came straight from those teams. Sebastian Thrun led Stanford's winning car, Stanley. Chris Urmson was technical director of Carnegie Mellon's winning car, Boss. Google launched its self-driving car project in 2009, and it became Waymo in 2016.
Read: The DARPA Grand Challenge: the big bang of self-driving →
The realization: people cannot babysit a robot driver
In its early years the team tested a highway system that drove itself while a person sat ready to take over. The footage showed the problem in plain terms. Trusting the car, people stopped watching the road. They turned to rummage in the back seat, looked at their phones, and in at least one case dozed off, all at highway speed.
The conclusion was blunt. Asking a human to monitor software that rarely needs help, then snap back to full control in an instant, is not something people are good at. The handoff itself was the danger.
The decision: build a complete automated driver
Rather than ship a system that leans on human supervision, Google chose to take the human out of the loop. In May 2014 it revealed Firefly, a purpose-built prototype with no steering wheel and no pedals. It was a physical statement that the car, not the passenger, is the driver.
This is the all or nothing stance. Skip the supervised middle of the SAE ladder and go straight to a vehicle that needs no fallback driver, within the conditions it is designed for. The project became Waymo in December 2016 and went on to carry passengers on public roads with no one in the driver's seat.
Why this still divides the industry
All or nothing trades a fast path to market for a higher safety bar and a much harder technical problem. Many carmakers take the opposite bet, shipping Level 2 driver-assist features and wagering that incremental automation will build toward autonomy over time. Which view is right is still one of the central debates in the field. It is also why the SAE levels are best read as a description of capability rather than a plan for how to get there.
Frequently asked
- Did Google invent the self-driving car?
- No, but its project, now Waymo, was the most influential pioneer of the modern era. It grew out of the DARPA Grand Challenge teams and was among the first to pursue fully automated driving on public roads.
- Why did Google remove the steering wheel?
- Early tests convinced the team that people cannot safely supervise a car that drives itself most of the time. Removing the wheel and pedals committed the design to full automation with no human fallback.
- What is the all or nothing approach?
- Skipping partial, human-supervised automation and building straight for a fully automated driver. It is the opposite of the incremental Level 2 driver-assist path taken by most carmakers.
- Is the Google self-driving car project the same as Waymo?
- Yes. The Google self-driving car project, started in 2009, became Waymo, an Alphabet company, in December 2016.
