Comments by "Joe Qi" (@i6power30) on "California Suspends GM Permit after Robotaxi Ran Over u0026 Dragged Pedestrian" video.

The problem is that AI only has vision it doesn't feel human under the car. It has no sense of hearing, balance or bump under the car. The current state of AI is not yet practical enough
5
@rogergeyer9851 yeah but it's not just simple statistics like fatality per 100000 miles. It's also the types of mistakes. If AI makes weird mistakes that humans would rarely make it's still a problem. It makes AI cars behaviour unpredictable. Difficult for humans to avoid potential accidents with them.
2
@rsmonge We don't make decisions only based on numbers alone, it's too simple minded. If nuclear power plants are so safe on paper causes less fatalities than natural gas plants, why aren't we building nuclear plants like crazy? Not in most democratic countries at least. It's because when it does fail, it can be catastrophic, many magnitude worse than when a coal or natural gas power plants fail or leaks polution. Back to Self-Driving cars. Any statistics about fatalities are based on past history, it doesn't tell you about any future black swan events. Because all cars of a given brand use the same software. If there were to be a bug or malicious hacker causes all cars to misbehave deadly ways all at same time, it will be 10 million accidents and 1 million deaths on that day alone, causing a spike anomaly in your statistics chart. Humans are more resilient as we are a distributed system, everyone has a different brain. Though humans are not perfect on individual basis, but it is nearly impossible for all humans to make deadly mistakes all at once, at same time.
1