So after 6 years of “beta”, now it’s “alpha” and you’re impressed. I have a bridge to sell. Not running red light is possibly the simplest possible thing a self driving system should do. If it fails at that after 12 versions (which also did that), and with supposedly a new end to end AI architecture and massive amount of data to train on, that should fix all the things, I’d actually say that’s pretty bad. Alarmingly bad.
And what does it mean “video was 50 min long it only needed one intervention”? Say you ride a cab. How many times in 50 minutes you need to have intervention with the driver so they don’t try to kill you?
“Oh it’s only failing catastrophically once an hour” is not the win you think it is. It means this thing can’t be trusted alone even for a second. It fails once an hour but you don’t know WHEN EXACTLY. So you need to sweat it every moment of the drive.
Tesla self driving is literally more stressful and dangerous than just driving yourself. It’s like giving the car to a mute autistic teenager who can swerve into a biker, speed past a stop sign or go in the wrong lane direction at any point.
And what does it mean “video was 50 min long it only needed one intervention”? Say you ride a cab. How many times in 50 minutes you need to have intervention with the driver so they don’t try to kill you?
“Oh it’s only failing catastrophically once an hour” is not the win you think it is. It means this thing can’t be trusted alone even for a second. It fails once an hour but you don’t know WHEN EXACTLY. So you need to sweat it every moment of the drive.
Tesla self driving is literally more stressful and dangerous than just driving yourself. It’s like giving the car to a mute autistic teenager who can swerve into a biker, speed past a stop sign or go in the wrong lane direction at any point.