Deeply pathetic to be admitting that you are such a poor driver that you are incapable of safely negotiating cyclists on roads. As EV owners we all are likely to have at least some environmental consciousness. we should be applauding cyclists.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Self-driving cars UK Discussion
- Thread starter randompixel
- Start date
Indeed and with the need for cyclists to give way to pedestrians they should have a pedestrian warning system both audible and visual and travel no faster than 5mph in case a pedestrian steps out. Rationally government should provide flat caps and clogs to save all that leather wear on't pavement wi' guide whippets for't disabled. As well as free humour implants.
Ah, you must mean the cyclists that drive out into the countryside for a cycle ride. As far as excercise is concerned then jogging is even better environmentally...no need for a bike.
I have no issue with cycling to work but the days of cars and cycles sharing the same tarmac need to be consigned to history. Or take cars away from the masses and back to the 50's for traffic density.
Fewer accident stats look fine but don't make allowances for how they would be historically if cars of those vintages had Mot tests, modern brakes, abs and the like . Some of us remember the days of bald canvas tyres and braking distance of 'perhaps' and no demisters or wing mirrors and trying to peer through the gloom ahead with Lucas headlights.....oh and the overtaking shared lane and even suicide doors.
Not looked at this thread for a while and some wild arguments been thrown around.
One point of pick up on is what’s acceptable as a human. I believe it’s +/- 2 prescription without requiring to have corrected vision when driving. But beyond that the state determined as that people freedom is greater than absolute safety. Somebody with one eye is impaired compared to somebody with two, there is trust in the person with one eye to drive within their limitations which may mean being extra cautious at junctions (those people we sit behind and get frustrated when the didn’t go could be a one eyed driver or for that matter someone with an adapted car because of no legs, one arm, or a host of other conditions). They’re highly unlikely to be in the upper quartile of safe drivers or able to react as well as a fully able bodied person. Age is another factor. As is deafness.
Now if we’re happy for self driving cars to be overly cautious, compromised, driving as well as a deaf, one armed, one eyed person and we actually think that’s acceptable, then I’m not going to sign up to that.
Self driving cars have the opportunity to be better, the sensor suite could massively outperform ‘just vision’, and if we want to rely on vision why not place the cameras that afford the best view down the road and maximise depth of field vision? If you’ve ever done advance driver training, then positioning yourself so you can see past the car in front is a common thing, So where would you place the cameras? Top right and top left of the windscreen rather in the middle for starters.
The video of the ex Tesla manager didn’t inspire confidence, the camera positions were ok.. hmm.. and the issue with the radar seems to be they bought the wrong one and/or couldn’t understand Bosch documentation - not understanding the sensor output is not the same as the sensor isn’t needed, it means you didn’t do your homework.
I’ve no doubt that cars will self drive at some point. I just think Tesla vision, the current sensor suite and level 5 all being in the same sentence is pure fantasy. Even level 4 is unlikely unless your definition of level 4 is just on motorways and good quality roads.
My 2p. Others will no doubt think differently. Only time will tell.
Edited to correct lots of typo's
One point of pick up on is what’s acceptable as a human. I believe it’s +/- 2 prescription without requiring to have corrected vision when driving. But beyond that the state determined as that people freedom is greater than absolute safety. Somebody with one eye is impaired compared to somebody with two, there is trust in the person with one eye to drive within their limitations which may mean being extra cautious at junctions (those people we sit behind and get frustrated when the didn’t go could be a one eyed driver or for that matter someone with an adapted car because of no legs, one arm, or a host of other conditions). They’re highly unlikely to be in the upper quartile of safe drivers or able to react as well as a fully able bodied person. Age is another factor. As is deafness.
Now if we’re happy for self driving cars to be overly cautious, compromised, driving as well as a deaf, one armed, one eyed person and we actually think that’s acceptable, then I’m not going to sign up to that.
Self driving cars have the opportunity to be better, the sensor suite could massively outperform ‘just vision’, and if we want to rely on vision why not place the cameras that afford the best view down the road and maximise depth of field vision? If you’ve ever done advance driver training, then positioning yourself so you can see past the car in front is a common thing, So where would you place the cameras? Top right and top left of the windscreen rather in the middle for starters.
The video of the ex Tesla manager didn’t inspire confidence, the camera positions were ok.. hmm.. and the issue with the radar seems to be they bought the wrong one and/or couldn’t understand Bosch documentation - not understanding the sensor output is not the same as the sensor isn’t needed, it means you didn’t do your homework.
I’ve no doubt that cars will self drive at some point. I just think Tesla vision, the current sensor suite and level 5 all being in the same sentence is pure fantasy. Even level 4 is unlikely unless your definition of level 4 is just on motorways and good quality roads.
My 2p. Others will no doubt think differently. Only time will tell.
Edited to correct lots of typo's
Last edited:
Mod comment:
There have been more than a few posts which are straying beyond what is acceptable and a number have been reported accordingly. There's healthy debate and there's getting personal. Does a discussion about self driving cars even need to debate whether cyclists should be on the road, other than to acknowledge they're a hazard that needs to be managed, as are horse riders, walkers, etc.? Either way, nothing has been deleted or changed, but please try and steer the conversation to well argued points.
There have been more than a few posts which are straying beyond what is acceptable and a number have been reported accordingly. There's healthy debate and there's getting personal. Does a discussion about self driving cars even need to debate whether cyclists should be on the road, other than to acknowledge they're a hazard that needs to be managed, as are horse riders, walkers, etc.? Either way, nothing has been deleted or changed, but please try and steer the conversation to well argued points.
I think you have hit the bullseye here. The fact that the cameras are placed where they are currently is a legacy artifact from the earliest days. I have often wondered why Tesla does not offer 360 camera views, which even economy cars now have. If they are going to persist with cameras as the solution to FSD, I think they will need to upgrade the cameras, the associated processing capability, and the placement.
But if they change the placement, it will be an admission that they are writing off the existing fleet and with it, the payments already made in advance by hundreds of thousands of "early adopters". I often wonder if they have gone down some technological cull de sacs in a rush to be first.
I don't think that's the case. From what I recall, Elon approaches it more in the sense of how capable the car is. The current version hardware might be learner driver level, or just about scrape a pass in your test level if Elon is lucky. And newer hardware (HW4, or any new camera configuration) will be like additional safety features. So reaching parity with the average driver safety maybe something achievable, being 10x less likely to have an accident is what a newer config gets you, 100x....
That's if they can reach the level promised with this current config. It sure seems strange to not pick more/better placements than a human driver would get from the start. After all, cameras are cheap right?! :/
The problem with aiming for "just as good as" what's before, such as an inexperienced driver or even an average driver... It won't become game changing, it won't sell, it could become a running joke.
The thing that changed public perception of electric cars, which Tesla were spot on about, is to make them better than ICE cars. Going for "as good as the worst" or "on par with the average" didn't sell, it wasn't game changing.
Once the public perceive something is better, it rapidly moves from early adopters into general acceptance. The last few years has seen an amazing growth of electric cars and their acceptance, all manufacturers started to succeed when they went for better. Obviously that did push the focus to a specific end of the market.
I see the same with any new technology, people need to feel it's better to accept change.
I think this is partly what's stalling self driving, even feeding the conscience of the regulators - a desire it needs to be better than a human.
And to make AI better than a human, even for a defined task, is still really really complex. We are not at the pinnacle of AI, we are right at the beginning.
Even outside of AI, Tesla started using robots for almost all their manufacturing process and soon discovered humans were better at many. These early decisions created their reputation of poor build quality and manufacturing hell and it's been difficult for Tesla to undo.
This highlights the dangers of promising too much too soon, it's a difficult place to climb out of.
Self driving will happen, but "full" self driving I personally think is at least a decade away.
I do think we'll see self driving in limited conditions like slow moving traffic or motorways within a year. But many of these scenarios aren't really an AI problem, they are very procedural and hence much easier.
The thing that changed public perception of electric cars, which Tesla were spot on about, is to make them better than ICE cars. Going for "as good as the worst" or "on par with the average" didn't sell, it wasn't game changing.
Once the public perceive something is better, it rapidly moves from early adopters into general acceptance. The last few years has seen an amazing growth of electric cars and their acceptance, all manufacturers started to succeed when they went for better. Obviously that did push the focus to a specific end of the market.
I see the same with any new technology, people need to feel it's better to accept change.
I think this is partly what's stalling self driving, even feeding the conscience of the regulators - a desire it needs to be better than a human.
And to make AI better than a human, even for a defined task, is still really really complex. We are not at the pinnacle of AI, we are right at the beginning.
Even outside of AI, Tesla started using robots for almost all their manufacturing process and soon discovered humans were better at many. These early decisions created their reputation of poor build quality and manufacturing hell and it's been difficult for Tesla to undo.
This highlights the dangers of promising too much too soon, it's a difficult place to climb out of.
Self driving will happen, but "full" self driving I personally think is at least a decade away.
I do think we'll see self driving in limited conditions like slow moving traffic or motorways within a year. But many of these scenarios aren't really an AI problem, they are very procedural and hence much easier.
Wol747
Active Member
I think that pedestrians - and even (!) - cyclists are very much a part of self driving.
We all see examples every day of dreadful driving - quite a lot of it intentionally, deliberate aggressive driving - but even more appalling cyclist behaviour. Regardless, a cyclist or pedestrian is going to be the loser in a collision, so FSD in whatever guise has to account for them.
"Even more appalling cyclist behaviour". This is a common misconception which is pushed by car-fans and the right wing press. Some cyclists break the rules, most do not. Some car drivers break the rules, most do not. We are all human beings going about our days. Just because you've seen a cyclist jump a red light does not mean another, completely unrelated cyclist has anything to do with that. The same goes with car drivers. Really, I think this attitude leads to a de-humanising of people who have the audacity to use something other than a car, and that leads to the many, many deaths that occur every year. Please be more respectful, we are all human beings with families, work, life, loves etc.
Wol747
Active Member
I've no desire to get into an us and them match on this issue - please read my post again.
Random musing.
The better Tesla gets with their FSD software, the more dangerous it becomes. Change my mind
Where I'm coming from with this seemingly silly statement, is the human factor. Currently, the FSD beta testers are very cautious, ready to take over at any given instance. They know that its not the finished article. But lets say that Tesla gets it to an amazing place, where disengagements become more and more rare. What will happen (human nature) is that people will let their guard down more and more, the better it gets. Does this translate into lots of accidents, and even fatalities, as the software improves?
Ie, Tesla FSD is at its most dangerous when its almost perfect? How could Tesla possibly manage this scenario?
The better Tesla gets with their FSD software, the more dangerous it becomes. Change my mind
Where I'm coming from with this seemingly silly statement, is the human factor. Currently, the FSD beta testers are very cautious, ready to take over at any given instance. They know that its not the finished article. But lets say that Tesla gets it to an amazing place, where disengagements become more and more rare. What will happen (human nature) is that people will let their guard down more and more, the better it gets. Does this translate into lots of accidents, and even fatalities, as the software improves?
Ie, Tesla FSD is at its most dangerous when its almost perfect? How could Tesla possibly manage this scenario?
I think you're right. The more I see videos of FSD struggling with simple tasks - like recognising a kerb, or a tramline - the more the end result seems unattainable. Every update makes tiny incremental improvements, but with increased capability comes increased complexity. I don't feel like the approach Tesla is taking is working. Maybe it's the camera-only idea, maybe its the software or the CPU capabilities, but when they can't make the windscreen wipers work well after 10 years of development you do have to wonder where this is going.Random musing.
The better Tesla gets with their FSD software, the more dangerous it becomes. Change my mind
Where I'm coming from with this seemingly silly statement, is the human factor. Currently, the FSD beta testers are very cautious, ready to take over at any given instance. They know that its not the finished article. But lets say that Tesla gets it to an amazing place, where disengagements become more and more rare. What will happen (human nature) is that people will let their guard down more and more, the better it gets. Does this translate into lots of accidents, and even fatalities, as the software improves?
Ie, Tesla FSD is at its most dangerous when its almost perfect? How could Tesla possibly manage this scenario?
Yup, I agree.
It seems logical that the less interventions that are required, the less prepared the human component will be to make them.
If it ends up in a situation where your commute is done fully autonomously, and only exceptional events cause it to require intervention, then it follows that people will naturally become more and more complacent, and "tune out" of the experience of travelling in a car entirely. The driver will effectively become a passenger in terms of attitude.
This will be compounded further still by a desire by some, possibly many, to explicitly opt out of driving and do something else instead. They might try and sit in the passenger seat and put something on the drivers seat to trick the car into thinking someone is there, or they might try and read a book, or be totally engrossed in their phone, not even watching the road. I would hope the cabin camera stops all of that, but we'll see.
So yeah, I'm in agreement that there will be an inflection point at which the tech is good enough that people get complacent, but not good enough to allow it.
It seems logical that the less interventions that are required, the less prepared the human component will be to make them.
If it ends up in a situation where your commute is done fully autonomously, and only exceptional events cause it to require intervention, then it follows that people will naturally become more and more complacent, and "tune out" of the experience of travelling in a car entirely. The driver will effectively become a passenger in terms of attitude.
This will be compounded further still by a desire by some, possibly many, to explicitly opt out of driving and do something else instead. They might try and sit in the passenger seat and put something on the drivers seat to trick the car into thinking someone is there, or they might try and read a book, or be totally engrossed in their phone, not even watching the road. I would hope the cabin camera stops all of that, but we'll see.
So yeah, I'm in agreement that there will be an inflection point at which the tech is good enough that people get complacent, but not good enough to allow it.
No, you're right.Random musing.
The better Tesla gets with their FSD software, the more dangerous it becomes. Change my mind
Ford back in 2016 decided that level 3 was a bad idea because they didn't know how to safely enforce it and as a result thought it would be dangerous. Waymo and Audi have had similar thoughts. I know we're talking level 2 at the moment with Tesla, but you can see the time when it will be limited Level 3 like Merc.
Now Ford changed their mind in 2019 under what could be commercial and reputational pressures (ie it would look like they're being left behind without level 3). Who knows the motivation, but the argument is one about fooling people into a sense of false security, and thats bad.
The "fix" seems to be internal cameras and making sure the driver is ready to take over.
Ford rethinks Level 3 autonomy
In 2016, Ford said it would largely pass over Level 3 autonomy, but now it's embracing a more gradual approach to fully autonomous vehicles.
I guess for me, whats the point anyway until its level 4? If I've got to be, more or less, paying attention and ready to take over at a moments notice, then I don't need any more new fangled features like the car turning across oncoming traffic with me on the edge of my seat ready to brake in case it gets it wrong. I'm happy with a decent lane and speed keep that doesn't break the speed limit and doesn't phantom brake. The rest, like traffic light stopping etc you can bundle up into passive safety features. Life ain't going to change until level 4.
I think it'll be a long time before the general population trust cars to completely drive themselves from A to B. FSD beta testers are not representative of the general population.
It'll be possibly just as long, or perhaps longer, before UNECE regulations relax enough to permit extended hands off autonomous driving. Given Tesla's attitude to Europe I would suspect that it might take longer still for them to adapt to the regulations as they get relaxed.
It is conceivable though that they end up in front of the regulations, with a "package" ready to deploy when they are relaxed, but that would really depend on how much testing they can do on European roads, etc without it being legal.
It'll be possibly just as long, or perhaps longer, before UNECE regulations relax enough to permit extended hands off autonomous driving. Given Tesla's attitude to Europe I would suspect that it might take longer still for them to adapt to the regulations as they get relaxed.
It is conceivable though that they end up in front of the regulations, with a "package" ready to deploy when they are relaxed, but that would really depend on how much testing they can do on European roads, etc without it being legal.
Level 2 allows physical relaxation.
Level 3 allows physical and mental relaxation.
Imagine there are 2 different, large busy restaurants. At both of them, you will have to wait for a table.
Restaurant 1 gives you a pager. It'll buzz when they have a table for you.
Restaurant 2 gives you a number. They'll display the number on a screen. They'll also call out once when they have your table.
Level 3 is the pager system.
Level 2 is the screen and call out.
Last edited:
I rarely find driving a mental challenge, and even when I'm a passenger I still concentrate as if I was driving. To me L2 and L3 would be little different, and I would need a lot of pursuading that L3 was good enough, and if it was good enough it would be knocking on the door of L4. The 7 second response time (which is what regulators in Europe are looking at) for L3 means there's very little else you can do, and I can't believe L3 will ever be on city streets. 7 seconds of accountability by the car means the car can deal with everything a city street can throw at it for the next 7 seconds - be that kids running out, turns across traffic, cyclists etc. If you can do all that, you'd not bother with L3.
Similar threads
