This episode from Derek Thompson's Plain English podcast was a pleasant distraction from news podcasts I've been wallowing in. Here in the snowy, densely populated northeastern U.S., we are far removed from the latest exploits of self-driving companies such as Waymo One (a subsidiary of Alphabet). That company has worked its way up to around 150,000 weekly autonomous rides in locales such as San Francisco and Phoenix. In Derek's lengthy interview with Timothy Lee, I learned about the cautious roll-out of these technologies and some of the technology that is slowly lifting self-driving out of the trough of disillusionment. I especially liked listening to the bits about how AI "transformers" are now being used to teach software how to respond to the rigors of the open road.
Lee discussed how various companies have approached the self-driving problem and speculated that the reason for their failures is the scope of the challenge. Some companies wrongly reasoned that they could tackle a sector of the self-driving space (e.g., urban deliveries, highway travel) only to concede that a system capable of operating in every environment is the only practical model.
About 40,000 Americans lose their lives in motor vehicle crashes each year. But despite over 6 million accidents, fatalities are relatively rare. Actually, the U.S. suffers just a bit more than 1 annual fatality for every 100 million miles driven. Waymo has not even logged enough autonomous miles to make a straight annual comparison (no fatalities yet). However, where comparisons are possible, Waymo's record is impressive when compared to human drivers.
Cutting-edge safety features seem to be coming online just as the scourge of distracted driving is cresting. I'm fascinated to learn how self-driving cars will be changing our lives in the next few decades.
Kommentare