> Compared to, say, a large language model that writes news or fiction... an autonomous-driving AI is less subjective and easier to evaluate, so there’s little room for error.
That's brave. What about all the edge cases?
> Compared to, say, a large language model that writes news or fiction... an autonomous-driving AI is less subjective and easier to evaluate, so there’s little room for error.
That's brave. What about all the edge cases?
They're talking about verification. Was there a collision or not compared to is this story good.
I can think of scenarios that avoid collision in probably 95% of the cases while being bad choices
- Driving through an empty prairie
- Driving slowly through a crowded square, giving people enough time to jump out of the way
There's less gray area with driving wss the point. Legal or illegal, efficient or inefficient. And whatever gray area remains is generally "good enough" compared to human drivers on the road.
It's easy evaluate if there's no victims. Stats say there's more than 6000 injuries per day and more than 16000 deaths per year on the road.
If you choose between hitting a pedestrian and swerving potentially injuring the driver and totaling the car, what would you like the car to pick? Or any given trolley problem variation, etc