Interventions Elon, interventions.

Tesla really trying to engineer good will, and unsurprisingly only using their own data, which paves over things like "driver intervention prevented FSD from crashing in this instance" or "FSD disengaged 2s before crash, therefore driver error".

Rather than play statistics games with self-reported dressed up supervised driving data to try and trick people into rolling the mortality dice with a robotaxi, just let the cars drive around empty. But he can't do that, because these cars are not FSD.

Even worse, the government could easily mandate LIDAR for autonomous cars, and that would basically kill Tesla overnight.

▲

kentonv 9 hours ago | parent [-]

How do you meaningfully count interventions, though?

I intervene on FSD all the time, but only due to differences of opinion, like I want to take a different route, or I think it's being too cautious, or I want to change lanes earlier / later than it decided to. In a year of use I can't remember a time when I intervened to prevent an accident.

So if you just looked at number of interventions in my data, it doesn't really tell you anything except that I'm too much of a control freak to just let it do its thing...

	▲	Workaccount2 8 hours ago \| parent [-]
		Tesla has the data from Austin, they closely track when the supervisor pulls the e-stop. But that data, the most critical data, is unsurprisingly absent here. The dude has got six weeks to get those supervisors out of the car, and he isn't even giving the most critical data around them. Not a good sign. Elon will probably find a near closed loop around an empty development in Austin, put two robotaxis there without a driver, and let people do a loop. Tesla contractors will walk the area an stop people from crossing when a robotaxi is near. Then he will proclaim there are driverless taxis in Austin and that they will expand to the rest of the US by mid 2026, and then talk about how actually optimus robots in 2027 are the real thing to focus on.