Self driving cars

I don’t think it matters. I do think we’ll gradually shift to a large number of vehicles not needing a human driver in the front seat to control it. And it will be a significant enough number to lead to a societal shift in various ways.

I think it’ll be similar to remote working now. Not everyone can work from home. Some who can work from home will still need to go in occasionally. But the shift is significant and transforming.

Getting hung up on whether we can get ubiquitous level 5 misses the point IMO.

Are you not building in danger with anything below level 5? It's creating an environment whereby drivers feel that they don't need to concentrate as much on the road, we've already seen deaths caused through this in the trials by Tesla and Uber.

I'm all for tech that'll improve safety, but not sure whether this is a good step forward. Can definitely see autonomous only lanes on motorways or toll roads in the next decade, you can limit pretty much all factors there. Gov are clearly keen on it too.
 


New FSD update was issued yesterday. It can drive your car from a to b, without any intervention from the driver.

guess it was here sooner than you thought.
😂

It isn't 'here'.

It is Beta, was only issued to the US and to selected customers. There is no proof it can drive from any 'a' to any 'b' without driver intervention.

Add to that the fact that the owner is legally responsible for the vehicle and must keep eyes on the road, hands on the wheel and feet on the pedals.

It is just another bread crumb to keep people onboard. All this for the low, low price of just £6200.

If I was going to buy a Tesla tomorrow (their cars are impressive) I would still not be confident that FSD would be complete within 3-5 years so I would probably sell the car before I got any real benefits from it.
 
I can’t wait for autonomous vehicles.

Imagine going to the pub then calling your car to come and take you home.

If you don’t need to drive, your car could have a bed in it. Imagine climbing into your car bed and then walking up in London or Paris.

It’s going to be mint.
Same here. My regular (in normal times) trip has been 330 miles to my mams, all but about 5 mile on dual carriageway. Usually you are just transfixed on the back of the car in front. We cut down on travel time, missing the traffic by trying to set off at 6:30, drive for 2 hours then have breakfast. I drive one bit and then we swap over and I either stare out the window of often fall asleep if I’m not concentrating on the road.

In 10 years time, it would be great to think that we could set off early and perhaps doze for the first hour before sitting around the table in the middle of the car ( no controls means more space inside) and have breakfast while on the move. Perhaps play a game together, watch a film or get a book out. Heavy traffic, we might not even notice. Tap a button when you need a piss and it will stop at the next services.

With fewer people driving stupidly and faster reactions to humans, they might add another 10 mph to the speed limit. If the car could just take me at 80 all the way, that is 4 hour 15 door to door. Perhaps even faster. If the cars can all talk to each other, then the sensors are more than just your single car. Something in the road 2 miles ahead? Drop everyone by 5 mph, create a bit more space, merge lanes one at a time, pass the issue and spread out. As a passenger you might not notice.

Compare that to what we have now. No idea about an issue up ahead so keep going fast until the traffic slows down. Find you need to change a lane, but nobody will let you in, so you almost stop. Then you get pissed off at the impatient fucker who does 100 right up to the blockage and tries to cut in at speed. Finally you merge but still you go slow because everyone wants a good gawp. Get past and you can’t get back up to speed because some tosser moves out to the outer lanes but stay at 50. You glance over and notice the traffic has built up on the other side of the road for a mile or so. There is no problem there, people just want a good look.

Driving is shit when it is like that. The more I can ignore the better.
 
Same here......

This sounds great and is honestly what the vast majority of people including me want and think when someone talks about 'Full Self Drive'.

The truth is it is decades off, if possible at all. Compare it to the humanoid robots of the '60s that everyone was going to have in their homes by the year 2000. They would cook, clean, feed the baby and answer the door along with a million other things. The truth was it was too difficult to produce a one size fits all solution. Most of these problems have still been solved but by simpler, separate methods, you can turn your oven, washing machine and heating on from your phone, even see who is at the door and you can buy a tiny robot to do your hoovering while you are out. The same thing will happen with the dream of L5 driverless vehicles. It can work in small areas / regular routes but not anywhere to anywhere, it is unlikely to be cheaper per mile than a traditional taxi but it could make motorway miles a lot easier. It could revolutionise logistics allowing driverless trucks to go from warehouse to warehouse doing the bulk of retail transport but avoiding towns, cities and problem areas. Many busses could have the driver removed, lots of them already have their own lanes.

Even if Tesla announce 'feature complete' on their 'Full Self Driving' package it is not what you outlined above and what everyone wants. Imagine it as doing your 330 mile journey but with a 17 year old learner driver at the wheel, you still need to be 100% alert for the full journey as you can't trust them to drive without assistance, you are held responsible if you crash.
 
😂

It isn't 'here'.

It is Beta, was only issued to the US and to selected customers. There is no proof it can drive from any 'a' to any 'b' without driver intervention.

Add to that the fact that the owner is legally responsible for the vehicle and must keep eyes on the road, hands on the wheel and feet on the pedals.

It is just another bread crumb to keep people onboard. All this for the low, low price of just £6200.

If I was going to buy a Tesla tomorrow (their cars are impressive) I would still not be confident that FSD would be complete within 3-5 years so I would probably sell the car before I got any real benefits from it.

keep saying it’s years away, but it doesn’t make it more true. This is beta yes, but a huge leap forward.
 
Last edited:
Are you not building in danger with anything below level 5? It's creating an environment whereby drivers feel that they don't need to concentrate as much on the road, we've already seen deaths caused through this in the trials by Tesla and Uber.

I'm all for tech that'll improve safety, but not sure whether this is a good step forward. Can definitely see autonomous only lanes on motorways or toll roads in the next decade, you can limit pretty much all factors there. Gov are clearly keen on it too.

There’ll be risk obviously. I guess it’s a comparison of risk. The majority of current accidents are human error. If near-autonomous had far lower levels of incident than human only driving why wouldn’t we go for it?
 
There’ll be risk obviously. I guess it’s a comparison of risk. The majority of current accidents are human error. If near-autonomous had far lower levels of incident than human only driving why wouldn’t we go for it?
I'm all for safety improvements, I'm just not sure on this. Unless you drag everyone up to the same ubiquitous level of autonomy at the same time (like dedicated lanes for example), you're leaving far too much open to human error and lapses in concentration.

The more you increase autonomy the less the car requires input from the driver, and it's human nature to switch off in those circumstances. If the tech gets confused then you're in bother and not likely to be ready to react to it.

Anyway, just glad Tesla are running their testing on American roads, not over here. Shows you what you can get away with if you're a billionaire and the industry is near unregulated. Interesting thread below on their testing practises.
 
This sounds great and is honestly what the vast majority of people including me want and think when someone talks about 'Full Self Drive'.

The truth is it is decades off, if possible at all. Compare it to the humanoid robots of the '60s that everyone was going to have in their homes by the year 2000. They would cook, clean, feed the baby and answer the door along with a million other things. The truth was it was too difficult to produce a one size fits all solution. Most of these problems have still been solved but by simpler, separate methods, you can turn your oven, washing machine and heating on from your phone, even see who is at the door and you can buy a tiny robot to do your hoovering while you are out. The same thing will happen with the dream of L5 driverless vehicles. It can work in small areas / regular routes but not anywhere to anywhere, it is unlikely to be cheaper per mile than a traditional taxi but it could make motorway miles a lot easier. It could revolutionise logistics allowing driverless trucks to go from warehouse to warehouse doing the bulk of retail transport but avoiding towns, cities and problem areas. Many busses could have the driver removed, lots of them already have their own lanes.

Even if Tesla announce 'feature complete' on their 'Full Self Driving' package it is not what you outlined above and what everyone wants. Imagine it as doing your 330 mile journey but with a 17 year old learner driver at the wheel, you still need to be 100% alert for the full journey as you can't trust them to drive without assistance, you are held responsible if you crash.
I don't think it is decades away. The difference between robots in the 60s is that the space race was on, science fiction was popular on TV, the microchip had just been invented, drastically shrinking computers, there was a wave of optimism about the amazing things that could be achieved in the 'near' future. But the fundamental thing was, most was hope, vision and science fiction. There were not really any prototypes or any feasible plans for how to build them, how they would work, what sensors etc. Same as there were concept drawings of moon bases.

Totally autonomous vehicles exist. There are a number of companies in a race to have the best, most successful one, and as I type there is probably one going around a test site. In addition, some of the technology used is live today in production in other things. There are self drive buggy style robots in large warehouses, doing parcel delivery out in busy streets and we have autonomous drones. We also have rovers doing the same thing on Mars and the moon. The technology is here, the hardware is built. The sensors can detect all of what they need, the automation of steering, indicating, going through the gears etc. What is lacking is the software able to cope with every eventuality.

There are loads of challenges on UK roads and after a few years of driving, you develop an ESP. A cyclist starts to slow a little , their position in the road wobbles a little, they have a quick glance back, all subtle signs that tell you they are going to want to make that right turn coming up and are trying to judge the traffic behind them. You slow down a little and sure enough the hand goes out and they start to make the turn. You were prepared. There are two things here. An autonomous car would not pick up on the subtle signs, but should be able to react faster. With more than just vision, even if it is a idiot cyclist all in black, it should be able to detect them (though it didn't help the person with a bike who went straight out from the side of one in the states). Making that algorithm 100% perfect and proving that is the longer slower challenge. However the other part is an autonomous car should never see a cyclist start to turn right and rather than wait 2 seconds it takes them to move, decide to floor it and overtake them while turning, almost wiping them out. (Happens to me every so often)

However one place where I can start to see the self drive and then totally autonomous cars is on motorways. That is generally a far easier, although faster, there are fewer variables. I can see a situation where you can switch to fully autonomous for motorway but take control or be ready to take control for the other bits. (Perhaps have fully autonomous in crawling traffic too.) For my 330 mile trip, that is 225 miles of no driving - win for me. Once there are sufficient numbers, I can see really smart motorways developing. Like I said before, a line of traffic able to communicate becomes a single unit of multiple sensors, reacting at lightspeed, able to see miles in advance. The idea of road trains has been developed where cars could be only 1m apart, lowering wind resistance and going at much faster speeds. This would not work with a meat driven car in the middle, so I can see a not too distant future where in 4 lane motorways, the outer 2 lanes are autonomous only (including lorries) where the inner two are for more traditional cars.
The more you increase autonomy the less the car requires input from the driver, and it's human nature to switch off in those circumstances. If the tech gets confused then you're in bother and not likely to be ready to react to it.
This I agree with. I have no desire to sit in a car with both hands on the wheel, constantly checking mirrors but not actually driving. I will really struggle to keep my concentration. I mostly think all or nothing, though the odd specific circumstances it would be useful. On a blocked motorway or ordinary commute when you are stop start, crawl at less than 5 mph. To have the car keep you 2m from the car in front and not move & beep if anything enters that space between, would be great and should be possible now. We have adaptive cruise control almost doing that at speed. I've had cramp in my calf a few times after long spells of crawling.
 
Last edited:


keep saying it’s years away, but it doesn’t make it more true. This is beta yes, but a huge leap forward.
I've been saying it is years away since 2010 while discussing it with people like yourself, I'll make you a deal, prove me wrong whenever you're ready.

Drive from a parking space on matchday at the Stadium of Light to your own driveway and park without a single input and I'll stop saying it.
 
Last edited:
I've been saying it is years away since 2010 while discussing it with people like yourself, I'll make you a deal, prove me wrong whenever you're ready.

Drive from a parking space on matchday at the Stadium of Light to your own driveway and park without a single input and I'll stop saying it.
That is a pretty long bet either way!
 
I've been saying it is years away since 2010 while discussing it with people like yourself, I'll make you a deal, prove me wrong whenever you're ready.

Drive from a parking space on matchday at the Stadium of Light to your own driveway and park without a single input and I'll stop saying it.

You were saying that a tech in 2010 that was years away was years away? Very insightful!

How would I do that? I dont have FSD.
I've been saying it is years away since 2010 while discussing it with people like yourself, I'll make you a deal, prove me wrong whenever you're ready.

Drive from a parking space on matchday at the Stadium of Light to your own driveway and park without a single input and I'll stop saying it.
That video is a year old by the way.
 
Last edited:
You were saying that a tech in 2010 that was years away was years away? Very insightful!

How would I do that? I dont have FSD.

That video is a year old by the way.
One of the interesting things is that these first semi-autonomous cars are learning as they go and gathering masses of data. Problems like those in the video are part of the dev process, not the final solution. It is fine to have a system which can go round a well marked test track, follow the lanes etc, but that is not reality.

It is really hard to make test conditions for everything that exists, such as missing or faded road markings, one situation early on had faded markings of the old road layout mixed with new. Roadworks or (not in the vid) something random like a wheely bin blown into the road. Also remember that road markings and challenges differ from country to country and the bulk of development has been in the US.

The video shows the car knows what it is happy with and what it is not. At the moment it looks like a crap experience because the driver was in control then handing back, then back in control all the way. It is the same as on roads that have spells of quiet and spells of busy, I get fed up of putting cruise control on and off, then eventually give up.

However the difference here is the car is building up masses of data that is going back to the manufacturer. Every time the car turns off auto-pilot or the driver has to take control, that is a big red light to the developers. They are working by looking at the most common situations where the driver is put back into control and crossing them off the list one by one. The driver is training the car, but not just that car, they are playing a part in training every car.

If that was a year ago, what will be the most interesting thing to see is the same trip in similar conditions, made in 4 years time. How many times will the auto-pilot give up? I suspect a hell of a lot less.
 
One of the interesting things is that these first semi-autonomous cars are learning as they go and gathering masses of data. Problems like those in the video are part of the dev process, not the final solution. It is fine to have a system which can go round a well marked test track, follow the lanes etc, but that is not reality.

It is really hard to make test conditions for everything that exists, such as missing or faded road markings, one situation early on had faded markings of the old road layout mixed with new. Roadworks or (not in the vid) something random like a wheely bin blown into the road. Also remember that road markings and challenges differ from country to country and the bulk of development has been in the US.

The video shows the car knows what it is happy with and what it is not. At the moment it looks like a crap experience because the driver was in control then handing back, then back in control all the way. It is the same as on roads that have spells of quiet and spells of busy, I get fed up of putting cruise control on and off, then eventually give up.

However the difference here is the car is building up masses of data that is going back to the manufacturer. Every time the car turns off auto-pilot or the driver has to take control, that is a big red light to the developers. They are working by looking at the most common situations where the driver is put back into control and crossing them off the list one by one. The driver is training the car, but not just that car, they are playing a part in training every car.

If that was a year ago, what will be the most interesting thing to see is the same trip in similar conditions, made in 4 years time. How many times will the auto-pilot give up? I suspect a hell of a lot less.
This is why Tesla are an estimated 6 years ahead of their rivals, they have been able to develop their AI using real world data from hundreds of thousands of cars in live environment. The FSD beta update released the other day is miles ahead of anything they have released before, it’s a huge step forward.

mad I’ve mentioned earlier in the thread, it would take a fair bit for me to trust it, I only have autopilot (not full self drive), and it only allows you t use it at certain times and even then I only use it when not facing oncoming traffic, but it does make driving a lot more relaxed on long journeys, you can take in a lot more around you and notbe anywhere near as tired as you are from a normal drive.
 
Last edited:
This is why Tesla are an estimated 6 years ahead of their rivals, they have been able to develop their AI using real world data from hundreds of thousands of cars in live environment. The FSD beta update released the other day is miles ahead of anything they have released before, it’s a huge step forward.

mad I’ve mentioned earlier in the thread, it would take a fair bit for me to trust it, I only have autopilot (not full self drive), and it only allows you t use it at certain times and even then I only use it when not facing oncoming traffic, but it does make driving a lot more relaxed on long journeys, you can take in a lot more around you and notbe anywhere near as tired as you are from a normal drive.
It is scary when you consider the progress made. In 2010 google had a prototype and by letting it run on quiet easy roads, it had clocked up 140,000 miles. That was really the first bit of learning and it had only been autonomous for 'some of the time' was the quote. 2013 major manufacturers said they were investing in the technology. 2016 was when they really started hitting the roads, so 2026 will be the 10 year milestone of anything above a proof concept car. Given the progress in those 4 years, or even the 10 years since google was running its very first experiment on public roads, I think there will be huge leaps by then.

Consider that in 1997, this was one of the most advanced 'smart' phones:
Logon or register to see this image

It was mainly just used by business people who had to travel a lot, such as tech sales people. In 2007, the iPhone launched:
Logon or register to see this image

it made the Blackberry of 10 years previous look prehistoric. Android followed a year later and looked pretty basic.

The ipad launched in 2010, so is only just 10 years old. It followed Android based tablets in 2009.

Developments seem to have slowed, but the 2015 models of iPad just don't compare to the 2010 ones. Once you get over that first evolution hurdle, progress is fast. Granted these are far simpler devices, not making decisions and nobody will die if it goes wrong. I feel with self-drive and AI in general, we are just going over that first hurdle or at least landing on the other side. It is going to be fascinating to see where it does go by 2026. If it follows a similar curve to other technology, we will be looking back hardly able to believe how shit it looked in that video.
 
You were saying that a tech in 2010 that was years away was years away? Very insightful!
I didn't make the point very well, the point was people with no background in software development, engineering or automotive have been telling me since 2010 that self driving cars are just a couple of years away.

A full decade later they still can't manage (in the UK) sharp turns, traffic lights, roundabouts, pedestrian crossings, heavy rain, low sunlight etc.

I'd argue Cadillac and Waymo are further forward in the areas that matter. All the big manufacturers are working on it but they are doing it without fanfare and fanboys because they know how big a project it is, it is uncertain who will actually be prepared to pay for the final product and it is uncertain what the final product will actually be. The legal implications alone could take years to green light. I'm confident we could come back to this thread in 5 years time and have the exact same arguments.
 
I didn't make the point very well, the point was people with no background in software development, engineering or automotive have been telling me since 2010 that self driving cars are just a couple of years away.

A full decade later they still can't manage (in the UK) sharp turns, traffic lights, roundabouts, pedestrian crossings, heavy rain, low sunlight etc.

I'd argue Cadillac and Waymo are further forward in the areas that matter. All the big manufacturers are working on it but they are doing it without fanfare and fanboys because they know how big a project it is, it is uncertain who will actually be prepared to pay for the final product and it is uncertain what the final product will actually be. The legal implications alone could take years to green light. I'm confident we could come back to this thread in 5 years time and have the exact same arguments.
Like I said above, I doubt anyone was seriously saying they were just years away. In 2010 google had a concept car that could do some driving on a quiet road. Things didn’t really take off until about 4 years ago.
 
I won't be in a driverless car
Stupid idea as most people actually like driving
Yup. The techy people are all for it,and I can see why, but techy jobs attract a certain type of personality and many of them won’t see the real issues.

(See anticipated replies to this post as evidence...)
 

Back
Top