[ActorNightly]

The issue with FSD systems as they are implemented today is that they aren't AI as much as just complex control algorithms. You can only go so far with mapping sequences of world snapshots to control actions.

I do think that once we start to investigate ML/AI structure in the direction of figuring out the correct solution rather than trying to just find functions for control algorithms based on input->output mappings, then a lot of these problems are going to disappear.

[adrianN]

> You can only go so far with mapping sequences of world snapshots to control actions.

Mapping some complex input state to control actions is literally the definition of driving a vehicle.

[ActorNightly]

No, thats the definition of closed loop vehicle control.

Driving, at least in the way humans do it, is more then that. We have internal sim running in our head that allows us to deal with conditions that we have never seen before.

[adrianN]

The internal state is simply part of the input. Your brain holds a finite amount of information, your sensors add a finite amount of information, your brain decides on which muscles to move in which way.

[ActorNightly]

Yes, but that decision process is much more than a one way compute graph. Muscle memory for actions (like throttle, steering, brakes) is probably closer to one way compute graph. Higher level strategy planning definitely has recursion to it.

[minwcnt5]

This is not true. The most advanced systems in the industry today like Waymo, Cruise, Baidu, Pony.AI, and Tesla, are all primarily AI.

[ActorNightly]

There is nothing "intelligent" about these systems, they are just complex forward directional compute maps.