EAP and cars that cut in front of you

[voip-ninja]

I think it might be more appropriate to say that AP is not programmed to pre-react to what might happen.

Humans tend to pre-react to what they perceive will happen in the future. You may see a vehicle changing lanes towards you from two lanes over and thinking that they may continue into your path you may pre-adjust your speed and position to allow for the possibility. AP is only going to react AFTER the offending vehicle/object breaks it's "safety barrier".

That's a very good point that I eluded to in a follow-on post when I mentioned predictive strategy and how computers are barely able to beat humans at the game of chess, which is exponentially easier than driving a car in all situations we find ourselves in on a daily basis.

Now I'll say that if you can teach a computer to think ahead 30 moves in chess than you can eventually teach it to think 30 moves ahead in all driving situations, but it is going to take a lot of time, and major advances in machine learning, sensor technology and processing capacity.

People who think that their Tesla is going to do the things Elon waxes poetic about anytime soon, well, let's just say that I'm pretty sure they stopped offering FSD for a reason, which is that they know it's quite a ways out from living up to those promises of doing things like being an autonomus uber part of a ride sharing fleet while you spend the day at work.

 

[StellarRat]

I think it might be more appropriate to say that AP is not programmed to pre-react to what might happen.

Humans tend to pre-react to what they perceive will happen in the future. You may see a vehicle changing lanes towards you from two lanes over and thinking that they may continue into your path you may pre-adjust your speed and position to allow for the possibility. AP is only going to react AFTER the offending vehicle/object breaks it's "safety barrier".

Yes, that's what I meant as well.

 

[derotam]

That's a very good point that I eluded to in a follow-on post when I mentioned predictive strategy and how computers are barely able to beat humans at the game of chess, which is exponentially easier than driving a car in all situations we find ourselves in on a daily basis.

Now I'll say that if you can teach a computer to think ahead 30 moves in chess than you can eventually teach it to think 30 moves ahead in all driving situations, but it is going to take a lot of time, and major advances in machine learning, sensor technology and processing capacity.

People who think that their Tesla is going to do the things Elon waxes poetic about anytime soon, well, let's just say that I'm pretty sure they stopped offering FSD for a reason, which is that they know it's quite a ways out from living up to those promises of doing things like being an autonomus uber part of a ride sharing fleet while you spend the day at work.

Yeah but you can have the car think 30 moves ahead, its just when is it going to make the decision to do something. If you look at this from a success/fail standpoint, I would bet that humans make a lot more "mistakes" than the Tesla does. From my scenario of another car moving over towards you...if you let off the gas at all or move over in your lane in anticipation and the other car doesn't actually break into an actual "threat zone", then you have made a mistake. Yes you were more cautious but you still made a mistake just like when AP sometimes brakes hard when it didn't need to. If people's AP started slowing down every time a car from two lanes over moved one lane toward you(between your car and the TACC tracked vehicle) and slightly moved past center line of the lane next to you, then people would complain about that.

I think my point here is just that you can't win, as Tesla, or as the consumer. Until all cars on the road are autonomous and/or can talk to each other about their lane/speed intentions, it's all variable.