Traffic cones confused a Waymo self-driving car. Then things got worse
cnn.com
cnn.com
The safety standard for robots cannot be that they never make a mistake.
The safety standard should be that they're 10% or 100% or 300% less likely to make a fatal mistake than a typical human driver.
That's it. That's all i want.
I don’t even need A to B, just give me A to A* where that distance is 90% of the trip.
Tesla's Full self driving package can do that in production today. I've personally driven 90% of my ~25k/km on AP.
Where I used to live I was about 1km to the highway, but about 40km to anything I cared about so AP was my best friend.
And this tells you everything you need to know about the overall viability of self-driving cars. There's a human. There's always a human.
I've seen this in other "High quality AI" powered products too. One, for example, parsed the Twitter firehose for events that folks would pay to get notified about. But it had to be backed up by an entire call center worth of people - some of whom also monitored live feeds and others monitoring AI feeds - to make sure that the quality was acceptable.