E.W. Niedermeyer points out an important paradox with autonomous systems -- if Tesla's FSD didn't suck so much, it would be dangerous

Those who follow Tesla Twitter, particularly from the critical side, have been seeing quite a bit of this.

The last four seconds of this vid show an egregious and utterly terrifying error for a system that calls itself “Full Self-Driving” https://t.co/B5wJV3dMrU

— Faiz Siddiqui (@faizsays) March 17, 2021

Here's the entire video.



Road and Track has a painful play-by-play.

But E.W. Niedermeyer explains how a pretty good autonomous driving system would actually be more dangerous.

The fact that "Full Self-Driving" is so laughably bad is actually the main reason we haven't had crashes yet. If it improves to the point where it only makes a potentially fatal screwup every 100 miles or more, that's when people will become inattentive and over-trusting. https://t.co/dxh43iI0jX

— E.W. Niedermeyer (@Tweetermeyer) March 17, 2021

Which brings us to the main point from the panel linked above with @missy_cummings, @MikeANees and @mclamann: waiting for mediocre automation to screw up every 10, 50, 100 or 1,000 miles and then making a split-second life-and-death decision is something humans are really bad at!

— E.W. Niedermeyer (@Tweetermeyer) March 19, 2021

For an even more disturbing example, check out this from Jalopnik.