[cfadvan]

I think that everyone in this industry is also starting to realize that far from a quick race to be first, this is actually the start of a long slog. With the realization that full autonomy isn’t a few years, but at least a few decades away, collaboration rather than throat-slitting becomes a more sensible strategy.

[gradys]

Given that Waymo has autonomous cars driving real users around at least one real city with no driver behind the wheel, why are you so confident that autonomy is "at least" decades away? Maybe you mean decades away from market saturation? Or from being able to do literally anything a human driver can do? Even then, "at least a few decades" feels overly pessimistic, and neither of those accomplishments are required to start eating the Uber/Lyft market.

[ghaff]

I'm still unsure if all the people claiming that they're going to be able to order an autonomous cab in a random US city within the next year or two are just irrationally optimistic or if they have a totally cynical interest in telling a bigger whopper than the next guy.

[euyyn]

I've never read of a single person claiming that.

[dougmwne]

I'd be very interested to hear you up unpack your prediction of "decades away." It seems possible to predict technology developments for the next few years, but I struggle to see how predicting decades away can have any evidence or meaning. I think what people mean is "this isn't theoretically impossible but no-one has any idea how to build it and it may in fact be impossible or impractical." Tech that's decades away probablly shouldn't receive any corporate r&d dollars at all.

Given how many billions are being currently spent on self driving, there's going to be a lot of dissapointed investors if they've just been sold snake oil, and anything decades out absolutely is snake oil in the corporate world.

[cfadvan]

Given how many billions are being currently spent on self driving, there's going to be a lot of dissapointed investors if they've just been sold snake oil, and anything decades out absolutely is snake oil in the corporate world.

Hell yes. A lot of hype has been sold at a dear price, driven in no small part by Uber’s desperation to find a profitable business model through automation. The reality is that critical edge cases are hard, and no one seems to know how to fix them. Two factors that I believe will dictate “decades away” or as you rightly say, someday, but not today, probably a loooong time from now. First, these crashes are not going to stop as long as SDVs are on public roads, and that includes the bad joke that is Tesla AP. That kind of thing is just too juicy for the media not to latch on to, and too scary to the average person to ignore. As a thought experiment, imagine if Uber has mowed down a well-off white kid instead of someone they could dismiss as a “formerly homeless woman.”

Second, the tech itself is hard and may be on shaky ground to begin with. The assumption that with enough training AI can adapt to the point of “better than human” is yet to be borne out in reality. And yet, the billions have already been invested, so a lot of people have a lot on the line to make some kind of MVP. So “millions of miles driven” gets bandied about, but it’s not miles in torrential rain, or terrible roads, or snow and ice. Who cares how many miles Waymo can drive on a sunny day in California? Get back to me when they can do the same on a back road in Maine, in the winter.

IMO this creates a feedback loop of hype and investment until the bottom drops out. Then it won’t just be a technological problem, but political, legal, and shrouded in “winter” as investors once burned will be reluctant to wade in again. To the people who see the billions invested as evidence of promise in and of itself, I say Theranos... Juicero...

[ghaff]

Historically, big corporate R&D labs spent sizable sums on speculative research that, in many cases, didn't amount to anything.

In the case of the auto manufacturers, subsets of autonomous driving are interesting. If a car can truly be hands and eyes-off on highways (and the tech for that is fairly close), that's a product that almost sells itself.

I'm much more skeptical about driverless cars doing arbitrary pickups and dropoffs. I do think that anyone who is investing on the assumption that an Uber will be able to offer that in a few years is going to be disappointed. It will happen but not quickly with a few caveats like maybe fixed locations can be established to simplify the problem.

[TillE]

This is such a well-defined problem with clear constraints, it's not the kind of R&D that can fail in the sense of a drug with unacceptable side effects or a high-tech fighter jet which becomes an overpriced boondoggle. It's simple navigation of paved, mapped roads with an array of sensors. Except perhaps for really harsh weather, even the edgiest edge cases are pretty limited in scope.

The absolute worst case scenario, within a couple years, is that you have limited deployment in favorable conditions, with fail-safes and human operators ready to take over remotely.

[robotresearcher]

You understated the constraints and missed the most important one: almost never, ever kill anyone. Driving around is relatively easy. Driving around and almost never killing anyone is much harder.

[mrguyorama]

You have to do better than not killing anyone. Every single accident, including ones not your fault, will be used against you to argue that your systems are death traps. You can't injure people, or cause property damage, or even be rear ended. You have to produce cars that are able to drive so defensively, that insurance is no longer necessary

[visarga]

I very much doubt your claim about "decades away". There would be zero interest in this field from companies if the prospect of medium-term profits is zero.