Is it fair to say they are "beta testing" on public roads with the public paying the price after they've done four years with a supervised driver in the cab?
Waymo, which appears to be doing an excellent job of handling it. Their robotaxis consistently seem to have a better track record than humans (although it's possible the limits put on them taint the comparison.)
Cruise, which had a whole bunch of problems with edge cases and last I knew wasn't allowed to operate anymore.
I'd share the road with a Waymo truck (not that their system can do that yet), I wouldn't want to share it with a Cruise truck.
This might be the first foray onto a subset of open public roads but it's not the first serious use of semi autonomous trucks mixed in with human traffic.
They usually map out the site and show you which roads you can take in your vehicle and they are completely separate from the operations roads. There are only intersections and it's always the responsibility of passenger cars to come to a complete stop and wait for any cross traffic to pass before proceeding.
I remember this well because we had a job servicing specific equipment on a particular mining companies sites in northern Minnesota. One time we drove on the wrong road. A security vehicle saw us and drove up on us at an extremely high rate of speed to clear us off the truck road and onto the proper passenger road. He spent a good few minutes yelling at us once we were clear. People have been killed this way.
Let's look at some raw data. Mind you, we have 100x more data for the human drivers below. ie, billions of miles vs millions for automated drivers.
Pop examined Fatal-crash rate All Level‑4 automated‑driving tests in the U.S. 1.5–1.7 per 100 M mi Drivers 70‑74 yr 1.7 per 100 M mi Drivers 75‑79 yr 2.1 per 100 M mi Drivers 80‑84 yr 4.3 per 100 M mi Drivers ≥ 85 yr 7.6 per 100 M mi
So statistically, you are much more likely to be killed by a senior citizen. So to make a consistent argument using data around safety, we should maybe be revoking the licenses of older folks?
edit: Jesus, 19% of all US fatal crashes in 2023 invloved seniors. While they make up 15% of the driver population.
Speed, alcohol, lower seat-belt use, and of course, inexperience are listed as the primary contributing factors.
Heck I was in a parking lot yesterday and almost got hit by a truck, driver is a very old man.
In terms of safety of AVs and human drivers, AVs need to cause 10x less fatalities than humans do in order to be widely accepted by people. It’s an emotional thing.
Just like these humanoid robots we’re seeing companies develop. If we let them loose in society, would it be ok if they only committed murder 1% less than humans do?
People naturally hold the bar higher for robots and computers. It might be for the lack of accountability. I’m not sure of other reasons why this is
How can we tell in this case? Do we wait for accident reports?
"In four years of practice hauls the trucks’ technology has delivered over 10,000 customer loads. As of Thursday, the company’s self-driving tech has completed over 1,200 miles without a human in the truck."
It would be a start, but we don't seem to have that. Or, at least, the article does not have them. What we do have is a press release and an indication that the load was delivered. No indication of the issues that arose. Just the simple fact that no issues are mentioned suggests there is a massive amount of information that is being omitted, because, I am sure you will agree, it is close to impossible for any kind of rig to run for 4 years and not have issues.
edit:
Tangent. I am starting to feel real fatigue from seeing those gaping holes and just about all articles I read each day. Even FT had some weird omissions in something I saw the other day. And that is just one problem. It gets even worse, when you read something you have some expertise in and the author gets it very, very wrong.
I like how you didn't seem to notice that there are 1,460 days in four years.
I get the impression that ten semi-truck hauls a day is a lot. If you hit someone (or wreck the truck) you're not going to make your next haul... you're going to be talking to the cops and/or having a wrecker drag you back home.
Sure? You'd only remove the man in the cabin if you'd already run many, many, many miles with the man in the cabin not having to take control.
It's true that the company could be being irresponsible (perhaps even dangerously so). [0] We'll just have to see how the vehicle performs over the next year or five.
Do make sure to compare any reports of failure to the expected rate of failure of human operators.
[0] Given that they're running fully-autonomous commercial payloads, I find this unlikely.
A single person could turn all of these things into Decepticon in seconds. We need to know that that is intentionally made impossible, and no one will ever make it possible in secret. It could give someone the power of a global army.
We're talking about an ordinary semi truck with attached trailer and a bevy of sensors. This isn't a war machine. The worst thing that can happen here is the same thing that can happen with an inattentive or exhausted driver.
As you mention, the (infrequent, but not nonexistent) accidents caused by overworked commercial semi truck drivers are one of the reasons why I do hope this works at least as well as human long-haul drivers.
I suppose various LEOs are worried over nothing then. I will say this though. Car, like so many other things, is just a tool and like with most, if not all, tools, the question of whether it is a war machine is a rather pragmatic one. Can it be used for war purpose?
Ah, yes. That historically-well-known war machine known as sunlight and fresh air.
These are things that happen. How does the AI handle them? Do we know?