>
I understand ‘Level 5’ and I am an active AI researcher.Then frankly your posts make even less sense, unless you've gotten so deep into AI research that you've lost sight of what humans can accomplish without technology.
>Do you understand what kind of technology is required to get to Level 5?
Utterly irrelevant to the discussion at hand, beyond that by definition L5 means zero human involvement is required.
>Do you think level 5 can be achieved without understanding how humans behave in a vehicle?
Yeah, absolutely. If how humans behave in the vehicle mattered it wouldn't be L5. Active sabotage is out of scope since there is no difference there then a human driver. If you are a passenger and suddenly start hitting your driver in the head on the interstate what could possibly happen? Oh right, accident and everyone is badly injured or dies.
>What if someone sleepwalking or rather sleep kicking and break the windows open then leave a leg or an arm hanging out of it?
Then they'll have shattered glass and bad cuts all over their leg or arm which would likely wake them up and also hurt a lot. They should probably tell their L5 car to go to the ER and also are you fucking serious? It's not as if humans sleeping in vehicles including ones with glass windows is some new thing and somehow an epidemic of people kicking through windows has never come to my attention. Feel free to share your stats on that one.
>Do you think a Level 5 car should not monitor how the passengers behave at all?
No. On the contrary, self driving cars monitoring their passengers at all times sounds terrifyingly dystopian. Exactly what country/agency do you do AI research for?
>Do you have real information/references rather than speculation that Waymo and GM have plans to launch only after achieving Level 5 rather than Level 4 which most observers agree on?
Amongst others: "GM Says Car With No Steering Wheel Or Pedals Ready For Streets In 2019", https://www.npr.org/sections/thetwo-way/2018/01/12/577688125.... A lot of mass publication stuff doesn't use "level 5" specifically because it's not terminology everyone will understand, but if it has zero manual controls and human involvement beyond high level orders ("Go here", "stop") that's what it is.
>Do you have a cost/benefit calculation of why the driver detection system is “not worth it” since it is better to wait for Level 5 before launching anyway?
Why would I need to? You're the one making the assertion that this is something Waymo should do, so presumably you have a cost/benefit analysis right? Waymo apparently doesn't though, and I'm just observing why that would be the case. There are plenty of actual for real jobs being done all the time worldwide that are of the "99% boredom, 1% HOLY SHIT WE'RE ABOUT TO DIE" sort, it's a tough but well known problem that can be worked on in a restricted setting with training, checklists, scheduling and support, no brand new tech required. You're asserting that they should develop a whole new untested complex technology, even if less complex, that will immediately become worthless (or even a negative for those worried about mass surveillance) in deployment. Why do you think this is better then just following decent professional best practices for R&D testing?
>When do you expect Level 5 can be achieved? If your answer is less than 10 years, you appear to know more than most of the top AI researchers I have talked with or read their prognostications.
Irrelevant, and you should really rethink your logic here if you think otherwise. Whether it's 1 year, 5 years, 10 years, or 30 years it's not the public's problem, it's Waymo's and their investors. They can't deploy L5 until it's ready, the timeline for "ready" is just their issue. If they're satisfied that with how their development is going just with human testers and training and the results are what the public wants then how they get from A to B is up to them.