Comment by seanmcdirmid

You think? Checking for significant differences between before and after images doesn’t even need AI. Have a triaging system, something like: image diff -> AI assessment -> human assessment -> car drives to human cleaner. In time you can streamline significantly.

I think simple solutions exist. For example the waymo could just ask the rider at the beginning of the ride “Is your car smelly? Any trash in the car?”

If they reply YES to any of those questions, then google keeps the recording of the previous ride for later review. Simple, no?

If the car actually was dirty, give the person that said “car was dirty” a refund of 20% on their ride. Then charge the party that made the car dirty a few. If a person is found to consistently say cars are dirty when they are not, give them a warning and consider kicking them off the platform (abuse).

Alternatively google can switch from recording the first minute of the ride to the last minute of the ride. In other words, only look at the “delta” to compare the car state. That would require a lot less video storage and less effort to review.

No it won’t. The problem is already solved, event detection is AI from 10 years ago, let alone today. Rider puked in the car isn’t that hard, they can add smell sensors also.

it's actually pretty easy to automate the state of the vehicle. if anything is left in the car the video is sent to an operator. The service instructs a cleaner to check the car and the passenger is sent a warning and eventually is banned from using the service.

The next rider will alert if something outrageous has happened.

This doesn’t even happen for cheaper mass transit, why would it happen for more expensive automated taxis? Cleaning a car isn’t hard, they can centralize that fairly easy, Waymo has already talked about adding e-nose sensors to the interior eventually, then we don’t even need AI to flag a strong odor.