Not saying you’re wrong (because I’ve always found it suspicious how Tesla always seems to report that autopilot is disengaged for fatal accidents) but there’s probably some people asking themselves “how could it detect the wall to disengage itself?”.

The image on the wall has a perspective baked into it so it will look right from a certain position. A distance from which the lines of the real road match perfectly with the lines of the road on the wall. As you get closer than this distance the illusion will start to break down. The object tracking software will say “There are things moving in ways I can’t predict. Something is wrong here. I give up. Hand control to driver”.

Autopilot disengaged.

(And it only noticed a fraction of a second before hitting it, yet Mark is very conscious of it. He’s screaming. )

Sidenote: the same is true as you move further from the wall than the ideal distance. The illusion will break down in that way too. However, the effect is far more subtle when you’re too far away. After all, the wall is just a tiny bit of your view when you’re a long way away, but it’s your whole view when you’re just about to hit it.

He said he tapped the brakes on one of the earlier non-wall runs.

Whilst I agree on the wall, fog and rain are not extreme weather conditions. I’d rather he’d used a level of rain that was less intense. However, the fact is lidar still worked even though it was not clear it was going to.

For those that didn’t see it…

Sadly this comes down to OpenAI petitioning Trump, and expecting trump to do anything that could stop a scam like AI is pointless.

🤞🤞 🙏

That 20% though…you want that to be physical.

Same as BMW iDrive. I’m sure some are touch screen, but you don’t have to use it.

Physical controls don’t change this.

More coding?! Are you serious? Over a touch screen!

Also, extremely easy to test.

Saying KHTML = WebKit is like saying a sponge is a killer whale.

…except they don’t for their own browser.

That’s not a flaw. That’s a right to repair requirement.