[Robotbeat]

They won't reach proper automation until they're getting 5,000 to 8,000 Model 3s per week. But, of course, nobody really does automation to the level that Tesla is trying to do, so their current mix between labor and automation (as they continue to build up their automation line) is approximately what other car makers do.

Getting full automation to work is really hard, takes a lot of labor and time. I think they'll do it, but the task should not be under-estimated. In the meantime, they've already achieved a production rate per factory line similar to what a typical automaker like VW would do (on the order of 100,000 per year for a factory).

Like SpaceX around 2015 or so. Respectable, but not yet Earth-shattering (i.e. 2017).

[agumonkey]

I wonder how much Tesla lives on SpaceX atmospheric results...

Musk probably knows a thing or two about technical limits since that, although people say that SpaceX needs are more high grade craft than large scale throughput but he might believe that this is peanuts.

I wonder how many lines run in parallel. thousands of cars per week is a few minute per car. That's not much :)

Someone please slip a drone inside.

[vasco]

> nobody really does automation to the level that Tesla is trying to do

Have you ever looked at production line videos from Mercedes Benz on youtube? You'll see that what Tesla wants to do is actually already being done, only with dead-tree engines instead.

[Robotbeat]

Tesla wishes to do full automation of basically everything, even final assembly. According to this recent Bernstein report (by a Tesla critic), this hasn't ever been successfully done before: https://twitter.com/valleyhack/status/979434674144423936

The point of the report is that Tesla is trying to automate too much, beyond what everyone else does today and where others have tried and failed spectacularly. The report uses that history as proof it cannot be done and therefore Tesla will fail.

Are you disputing that by saying Benz does full automation of even final assembly?

[agumonkey]

bmw x2 is impressively devoid of humans

https://www.youtube.com/watch?v=TubOzvqOCmI

if you look closely in the background at 10:14 you can catch a wild homo sapiens wandering between an army of robocop arm prototypes

[Robotbeat]

Your video makes my point very nicely.

From 12:00 onward, though, humans are used for many tasks, including multiple assembly steps and especially final assembly. These are things that Tesla wishes to fully automate and which industry experts think they're crazy for trying. And honestly, looking at that video (especially the wire harness installation at 17:00), it sure looks like an incredibly hard problem to solve and the industry experts certainly have a good point.

[agumonkey]

I think it would be hard if approached as a human technique automation, but there might be other ways to connect power and signal lines in a car in more mechanical ways. Depending on how much thought and desire has been spent by experienced factory designers...

[Robotbeat]

You're absolutely right, and there's a lot of innovation that comes not from automating existing human technique but from changing how things are made so that the technique can be easily automated. For example: from through-hole components in PCBs to surface mount components in PCBs. Or even the PCB itself, which replaced wire wrapping (which, granted, can be automated) and soldering components together with free wires.

The design and the technique are critical variables that must be tweaked to make automation feasible.

[qaq]

What industry experts were saying about Falcon?