How do folks think Tesla will address basic Autopilot once they have end-to-end FSD working well enough? Will they leave Autopilot as a heuristic system because it's good enough for what it is? Will they modify FSD to provide Autopilot functionality so that it's more versatile and reliable? Will they train a new neural network with Autopilot functionality in an attempt to provide a competitive edge at that level of product?
As part of this, will they attempt to create neural networks for other assist features? Traffic Aware Cruise Control (TACC). Autosteer.
I'd say that parameterizing FSD to provide lesser functionality is a fool's errand, but creating a new neural network from all the data they have could produce some really nice basic assists. For example, knowing how to negotiate various situations better than a heuristic system, even if the only thing the system is doing is keeping a lane. Just keeping a lane would involve some thinking, as lanes are added and removed from the road. Perhaps the adaptive speed control would be a big safety win.
At a larger scale, will things like TACC and Autosteer simply go away once FSD is the norm? Or will FSD remain an expensive option on expensive cars?
As part of this, will they attempt to create neural networks for other assist features? Traffic Aware Cruise Control (TACC). Autosteer.
I'd say that parameterizing FSD to provide lesser functionality is a fool's errand, but creating a new neural network from all the data they have could produce some really nice basic assists. For example, knowing how to negotiate various situations better than a heuristic system, even if the only thing the system is doing is keeping a lane. Just keeping a lane would involve some thinking, as lanes are added and removed from the road. Perhaps the adaptive speed control would be a big safety win.
At a larger scale, will things like TACC and Autosteer simply go away once FSD is the norm? Or will FSD remain an expensive option on expensive cars?