I'm not surprised insurers are nervous.

My wife's boss picked up his new S class last week and he had his first motorway trip yesterday,

The all singing all dancing systems really couldn't cope with roadworks where orange studs marked the lane edge and not the usual white lines, they were the wrong side of the cones and the S class really wanted to steer the car towards them.

<< While insurers and Thatcham support the introduction of more automation – and believe fully automated cars would be safer than human drivers – they said the current technology was a “quantum leap” away from what was needed. >>

If this was technically true no-one need worry for a moment, as a quantum leap is the tiniest jump possible at a molecular level. Unfortunately some time ago someone thought it made a trendy vogue-phrase, and it now means the exact opposite. Such is life.

The automated system would potentially need to return control to a human driver within three seconds to avert high-speed collisions – but insurers’ research found it takes 15 seconds for the driver to be sufficiently engaged to react to avoid a hazard, roughly 500 metres distance on a motorway.

I'm not an expert at all nor especially perspicacious in general but two things have been obvious to me since the topic of self-driving cars got going.

- autonomy in the short term can only work in a completely standardised environment which does not exist and is unlikely to for a long time

- the 'solution', based on an assumption that self-driving can work well enough such that a standby human can take control when it makes a mistake or says it can't cope is utterly unworkable at anything like normal motoring speeds. Humans can't sustain full attention for very long when they are not performing the task, and whatever anybody thinks nobody can truly multi-task - they can only time-slice their full attention.

I know how long it takes to get my wife's attention when she is absorbed in the crossword, and vice versa. She or I would have crashed well before the second alert. Even when the 'attendant's' attention has been obtained, in can be some time before they have understood the situation - if they ever do.

Examples involving transport are not hard to find. The fatal self-driving Uber accident, a number of aircraft incidents including AF447. SPADS on the railways are so frequent that TPWS is necessary - fortunately trains usually stay on their tracks and stopping them even after the signal is usually enough to prevent a collision. When TPWS was introduced it was understood that the limits of human performance in the constant monitoring of signals had been reached.

There is a world of difference, a massive gulf, between a human minding an automated system and an automated system that helps the human driver.

I suspect the insurers are also glad to have a proxy for another major concern, liability. Liability generally arises from negligence. If a driver who fails to stop a self-driving car from crashing is shown simply to be human, rather than actually negligent, then who is to blame?

I think it's unworkable even when the car 'knows' it can't cope and makes the warning announcement, but in some cases it won't even do that - like the S-class that wants to drive through cones.

I think this will be a short-lived experiment, if it happens at all. The fact (and it is a fact) that full autonomy will become possible at some point does not mean that it is imminent. The delivery robots in Milton Keynes are nearer to where the technology really is in practical terms.

Edited by Manatee on 23/10/2020 at 10:01