In a Bosch news conference streamed yesterday, Michael Bolle, the company’s CTO, laid down some harsh reality about the timeline of autonomous vehicle development. He suggested a timeframe that’s much more sober than what others suggest — notably, and Tesla CEO . For the record, my money is going on what the Bosch guy said. I’ll explain why.

Here’s the quote, thankfully noticed and , an editor over at :

“In L4 and L5, the industry has realized that the complexity of the automation task, for example for a robotaxi in a highly dense urban environment, is significantly more complex than we imagined even 5 years ago.

“We think fully autonomous driving in robotaxis in megacities will not be visible in the first half of this century, maybe in the second half.”

One important clarification: Bosch says Bolle (whose native tongue is German, like our own David Tracy) actually meant to say decade instead of century, so we’re talking post-2025, not post-2050.

You can see the exchange where Bolle says this

Even a post-2025 date is far later than what Elon Musk was claiming in July when he stated that:

“I’m extremely confident that level five or essentially complete autonomy will happen, and I think it will happen very quickly.

I think at Tesla, I feel like we are very close to level five autonomy. I think I remain confident that we will have the basic functionality for level five autonomy complete this year. So, yes, I think there are no fundamental challenges remaining for level five autonomy.”

As far as I can tell, Tesla did not release the “basic functionality for level five autonomy” last year, and from what I can tell they do not appear to be even remotely close to doing so.

I know that a lot of Tesla stans will point to , a Level 2 driver-assist system, as evidence that Tesla is indeed very far along on the autonomy path and Elon is of course, not wrong.

And yes, Autopilot is getting more and more advanced! Tesla is very happy to show videos like this one

...that demonstrates just how much the system can do. It’s quite capable, no question. A technological triumph! It’s also still just a Level 2 system, and always will be no matter how much they refine it, until they make one significant change.

That change is that the system has to be able to failover, or hand off control to a human driver with enough warning that the aforementioned human driver doesn’t have to be hovering, ready to take over in an instant like drivers using the current system have to.

That’s . It simply does not matter how much an autonomous system can do if it can’t handle things when it gets into trouble. If it requires a driver to remain constantly alert to take over at any moment, it’s just a driver assist system.

Bolle is right: A system operating at Levels 4 or 5 would be doing something incredibly complicated, and this was vastly underappreciated by engineers in the past decade or so.

Yes, fantastic progress has been made, but there is still so much to do, and it’s no longer the fun stuff. The fun stuff is what Tesla is doing — actually making a car capable of driving itself under certain circumstances — but they’ve done next to nothing on how that car will handle situations where a sensor or camera fails, or it needs to safely pull off a road. Or safely give up control in situations where everyone inside the car may be asleep or drunk or whatever and there’s nobody to actually hand off control to.

This is where the real hard work is, and it’s where a lot of unpleasant realities come into play. Even if they do get it figured out, will AVs need a complete set of redundant systems? That sounds expensive.

Will cars need a network and standards for communicating with one another in situations where they find themselves compromised? That sounds time consuming and requires cross-company and even government-level cooperation, which is also no fun. This is the hard part, everyone. Elon can make his cars drive around very impressively until a bird shits on a camera lens, whereupon it all goes to hell.

Bosch seems to understand that the last 10 percent of AV development could easily take as long as the first 90, because , there remain

... many small problems

...and I think those many small problems are actually the biggest things to solve.

It’s actually at a point where all I can do is hope Bolle wasn’t right in his misstatement and we don’t have to wait until the second half of the century to reach Levels 4 or 5. But, I won’t be too shocked if his misstatement proves to be correct, either.