AlanSubie4Life
Efficiency Obsessed Member
Without training, my raw reaction time is more like 200ms-250ms. These are just pedestrian times.Not better than Alan. He has .1 ms reaction time and super human driving skill. It’s all easy and simple to do.
This is far faster than anything yet demonstrated by FSD v12. And that's without any training, and it neglects the ability to anticipate, which helps humans even more.
Certainly in future, some version of FSD will be superhuman and completely unbeatable, but there is zero evidence of that when looking at pure reaction time, for now. Response time is something that will clearly improve in future even if there are other major limitations of AI which render autonomy impossible for the time being.
There are plenty of arguments to be made about how it's always aware, won't screw up, has eyes in all directions, etc., and those all have validity, but are not relevant to the discussion. Competent human still wins out; that is something we can all agree on! And a competent human with FSD Background Assist (which will always be a free feature!) may be completely unbeatable.
It's definitely going to be a noticeable step in improvement, otherwise they would not release it.being a definite noticeable step in improvement.
The question is will it be 5x-10x better as would be expected based on the theory? That's the bar the end-to-end AI proponents have set for themselves. They won't be saying "well they have to learn how to get the big improvements in performance with retraining, that will take time, we'll have to wait for 12.5 or 12.6" - they won't engage in that sort of moving of the goal posts - end-to-end AI is the answer.
Last edited: