[Rolebama]

Just found an article on msn.com saying that a Tesla engineer says that back in 2016 Tesla blatantly lied about just how clever their cars were, and that some oftheir claims were fictional.
In the same link there was video of a Tesla crashing because a speeding BMW coming up behind caused it to accelerate, which caused it to crash. This was apparently quite recently in Arizona.
Tet another story, with video, is about a Tesla which braked for unknown reason causing a multi-cat collision.
Link: https://www.msn.com/en-gb/news/world...5c44434a5c56f5

[Marc]

A very interesting area of debate @Rolebama
I thought it would be cool to ask this question to the new chatbot called ChatGPT people are talking about.
(There was a slim chance that asking a question about autonomy to an AI trained robot, which I guess means it's autonomous, would break some kind of scientific or moral rule.... but hey-ho).

The first reply I felt like I had been told off.
And that's the difference I guess between humans and bots - it had no context. I had to narrow it down, get more specific.

[Rolebama]

With the best will in the world I don't believe that any form of AI can be programmed without some of the personality of the programmer/s. I cannot see the day when the equivalent of human error will not be present.

[Marc]

I certainly hope so @Rolebama !
The idea of bots programming themselves makes me a little nervous.
I've no idea how far we've gone. Are Teslas able to learn from their own AI? I imagine they still need a human too 'tell' the bot: 'this was a good response, that was a bad one'. How else would the bot be able to judge? (my head hurts)

[Santa]

A widely circulated "statistic" shows that 94% of all RTCs are down to human error. While many would argue that other factors such as road design and signage are also partly to blame it is indisputable that most collisions result from some mistake made by one or more drivers.

I have seen arguments against autonomous cars centring around a scenario where the computer is presented with a choice between two actions, both of which would be disastrous. This is nonsense as a human driver would, in a similar (and unlikely) situation simply react and would not make a rational decision.

It is the transition that will be difficult. As long as computer-controlled cars are in the minority there will always be situations where conflict will arise. Fast forward 30 or 40 years, when only die-hard enthusiasts will still be driving, and RTCs will be almost a thing of the past. It may well be that such things as traffic lights will become unnecessary as the cars will communicate with each other to negotiate crossings. There could be a drastic reduction in Fire and Rescue and far fewer trauma cases in A&E. Pedestrians will learn that when they press the button at a crossing, traffic is guaranteed to stop and, since no one will be speeding, collisions between pedestrians and cars will be fewer and less serious.

[Rolebama]

I think 100% of RTCs are human error. Either in their driving, lack of maintenance or generally avoidable mechanical failings. How long before after-market accessories to override safety features 'for off-road or show purposes only' become available? I have already heard conversations about how the speed limit functions could be overridden.

[Marc]

Some new Tesla news @Rolebama
362,000 cars have self-driving software flaws!
https://www.theguardian.com/technolo...f-driving-cars

[Rolebama]

As much as I dislike the idea of autonomous cars, what were the drivers doing? I have met a lot who have never read the manual, and run into problems which need never have arisen, so I can't say I have a lot of sympathy for some of them.

[Marc]

I spotted this article/video on the BBC yesterday, about a uni researcher who disguised himself as a car seat so it looked like the car he was driving was, in fact, autonomous.
In the video he talks about how pedestrians still waved to the car! Love that. Polite as ever 😁

[Beelzebub]

From today's Times:

"Tesla’s Autopilot system has been involved in more than 700 crashes in the US since 2019, including 17 fatalities, far more than previously reported, according to regulators.

There have been 736 crashes involving Teslas in Autopilot mode since 2019, according to a Washington Post analysis of data from the National Highway Traffic Safety Administration (NHTSA). They include 17 fatal incidents and five serious injuries."

[Santa]

Originally Posted by Beelzebub

From today's Times:

"Tesla’s Autopilot system has been involved in more than 700 crashes in the US since 2019, including 17 fatalities, far more than previously reported, according to regulators.

There have been 736 crashes involving Teslas in Autopilot mode since 2019, according to a Washington Post analysis of data from the National Highway Traffic Safety Administration (NHTSA). They include 17 fatal incidents and five serious injuries."

That's 17 deaths out of more than 120,000 on US roads in the same period.

Numbers mean nothing out of context.

[Beelzebub]

Originally Posted by Santa

That's 17 deaths out of more than 120,000 on US roads in the same period.

Numbers mean nothing out of context.

Indeed. And to put it in context, we'd also need to take into account the relatively tiny number of Teslas on US roads.
However, the average punter (or indeed journalist) won't ask those questions, and stories like this inevitably promote scepticism about autonomous vehicles. Rightly or wrongly.

[Rolebama]

I was going to put this in the Humour thread, but decided it was maybe a little too near reality to be funny.

[NMNeil]

Closer than you might think.
https://www.caranddriver.com/news/a4...-repossession/

[Rolebama]

Interesting and frightening at the same time. I think we have all heard of autonomous cars colliding, so who would be responsible if a 'repo' one collided with an innocent vehicle, and how long before the tech got into the wrong hands?

[Grimm]

It sounds like Tesla's self driving tech still has a lot of kinks to iron out, and I agree it's hard to imagine a future where human error is completely eliminated, even with AI.

[Rolebama]

I think the recent chaos shows how much of a problem can be caused by human error. I have not heard positive clarification of what went wrong. Was it human error in respect of programming, or human error in the implementation of the software? Or even human error in the original program amplified by the update?