Here's this new interview of Elon.
Again complete fluff answer and flowery statements to anything Lex asked that could actually birth real responses and details. Even lex who is one of the biggest fan out there was taken back by it.
Lex: when asked "how hard is the remaining steps in software"
Elon Musk: fluff non answers
"Require detecting hands on wheel for at-least another 6 months"
"Driver monitoring/vigilance is mute"
"mute"
"mute"
"having humans intervene by end of this year will decrease safety"
"elevators"
"mute"
"Game. Set. match."
I don't understand how
@diplomat33 expects actual details from April 22 other than fluff answers.
I guess it is a matter of perspective. I found the interview plenty interesting although a lot of it was stuff I'd already heard. He talked about different types of disengagements that Tesla analyzes to improve AP. He also mentioned that his dev car stops and goes at traffic lights. He gave a little explanation of how Tesla is optimizes the software for handling intersections. I appreciated those comments. To the question you allude to of the remaining obstacles to FSD, Elon gave a good answer in my opinion. He said the following is critical to getting to FSD: 1) Having the FSD computer (AP3) that has enough base computational power, and 2) Refining the neural net and the control software. That's not a fluff answer.
Elon was also asked what needs to be done for people to really enjoy AP where AP is really useful for people. Musk again gave a good answer: 1) People will already like AP on highways because it really improves quality of life. 2) They need to extend that highway AP experience to city streets including traffic light recognition, navigating complex intersections and being able to navigate complex parking lots and the car being able to drop you off and find a parking space. I don't consider that a fluff answer.
Elon was also asked how long Tesla will require driver supervision. Musk said hands on wheel will be required for at least 6 months. Depends on how much safer the system is compared to human driver. You need large enough amount of data to have big sample to be able to statistically say that car is safer than human driver. Car needs to be 200-300% safer than a human driver. You measure this quantitatively by looking at incidents per mile, also probability of injury and probability of crash (fatal and non fatal). Maybe you consider that a fluff answer. I don't.
And Musk makes the excellent point that as the system improves, eventually vigilance is needed less and less. Eventually, when you get to L4 autonomy, the driver does not need to be vigilant at all and in fact, it would actually be counter productive for the driver to intervene. Elevator analogy is a good one.
Musk was asked about driver monitoring system. His answer was not fluff. He said that if your system is below human reliability then yes, you need a camera based driver monitoring. If you self-driving is dramatically better, you don't.
Musk was asked about the design decision for where AP could be used (in contract to GM's Supercruise which is super limited to only certain roads). Musk did give a fluff answer on this IMO. He said that when we have FSD, we will look back and think we were crazy to ever allow manual driving. I do agree with him on that though even if it is a fluff answer to the question.
But to your last point. No, I don't expect super secret details at the April 22. Yes, I expect a lot of the presentation part to be repetition of what we've already heard over and over again (FSD capable hardware, AP3 10x better than AP2, FSD features rolling out etc.). Karpathy may give some more details since he is closer to the FSD development than Musk is. But the test drives should be good too in showing us what Tesla actually has. I am more interested in the test drives than in the talky presentation.
