The News
Waymo has tried to prove its autonomous drivers are statistically safer than human drivers, but a few recent deadly incidents involving pets have prompted it to attempt another trust-building strategy. In a blog post Tuesday, Waymo revealed a little more of its secret sauce to safety by describing its model architecture, including how it uses both fast- and slow-thinking AI to predict behaviors from “other road users” (ahem). It also details how various components of Waymo’s AI work together to make decisions it describes as “demonstrably safe.”
Know More
A Waymo spokesperson said the release wasn’t a response to bad press and instead, the company wanted to provide a deep dive for the AI community into how its technology differs from that of its competitors. The disclosure also serves to convince the public that safety is engineered into the core of Waymo’s product, not just its marketing.
Unlike other AI technologies, Waymo is in the business of both predicting the next best move for its user and anticipating the next move of everything surrounding it — an incredible technical feat. There are acceptable mistakes and unacceptable ones — Waymo has largely figured out how to tell them apart.