Originally Posted by
Horatio Leafblower
As for my hire car got upgrded to a Volvo C90 and its autopilot steers within lanes, accelerates, decelerates, maintains distance from the car in front, and navigates itself. I feel the difference between this and driverless is more in the software than the vehicle capability.
According to Google, any step up from this is the most dangerous combination of driver/automation possible. Some even argue that this is approaching a step too-far. You can’t incrementally increase automation from here. You either go fully autonomous, or stay where we are.
And I agree.
Just enough automation to make you think it’s doing ok, you can relax, take your hands off the wheel, feet off the pedals, drink coffee, send texts, watch Netflix, but all the while you are still the driver in charge. If you hit someone, it’s still your fault.
That is actually a very close parallel to the current level of automation in modern
airline cockpits.
So why is that bad?
It’s really bad for these reasons:
Aircraft: The pilot has gone through extensive training to use automation to relieve workload, but knows it must be monitored at all times and be ready and trained to take over and fly manually at any moment. He has to be licenced and tested to achieve this with medical standards and countless hours of regular retraining and assessment. These assessments will dedicate training time to his ability to recover the situation in the event the automation fails. He also has to demonstrate the attitude and discipline to achieve this challenging goal of man vs machine, day in, day out. Only a certain percentage of the population will ever have what it takes to accomplish this.
Cars: Drivers of a car with similar automation levels will never have anything like an order of magnitude of those standards, understanding, discipline, or training. It just can’t work.
And to top it off, a pilot considers it “a very close call” if he comes within maybe 30 seconds of a collision. And for most of his day, he remains at least several minutes away from any potential collision. A driver in a car with passing opposite direction traffic is literally 2 seconds from disaster
the entire time. Any abnormalities that occur outside of the scope of the “partial automation” will provide little chance for the distracted, untrained driver to take over and save the day. From that principle alone, risk analysis would dictate that pilotless airliners should in theory exist long before anyone puts their trust in a driverless car.
Your average semi-autonomous Volvo driver has no training, no standards, no medical requirements, no human factors training, no discipline, has screaming kids in the back, is half asleep, possibly pissed, and probably watching Netflix. And his 18yo son earned his “L” plates sitting in the driver seat for 120 hours without touching the wheel or pedals while filming himself on Instagram. How’s he supposed to takeover when the partial-automation fails? He doesn’t care because he’s preoccupied by the fact that his Instagram won’t upload because this is Australia, NBN sucks and Project Loon hasn’t launched here yet. This is how politicians confuse us. Give a 15 minute soliloquy about what they want to say and then in the last sentence mention the topic at hand so they can “stay on thread”
Google is right, semi-autonomous will never work in cars. It works in aircraft, but without aviation-level discipline, it can never work in cars. All or nothing.
A bit of extra technology thrown in such as radar autonomous braking is great. That’s just a manoeuvring characteristics augmention system, nothing wrong with that, just a software improvement to a proven safe design - in fact, odds are it should never even activate.
But the minute drivers think they can take their hands off the wheel and watch Netflix, ****’s gonna get real.
Fully autonomous or not at all.
And we’ve a long road ahead for fully autonomous. I might get slightly more interested when they can build a vacuum cleaner that doesn’t do
this.
Go Loon!