It's precisely because of this kind of shenanigan that the EU stepped in to mandate compatibliity requirements - and minimum warranties for consumer equipment.
TTIP is likely to dismantle those consumer protections.
"In 10 years "programming" is going to be like "keyboarding" or "Microsoft Office" on a resume."
In the UK "programming" _IS_ using those programs.
No actual programming skills taught.
Couple that with rote-learned qualifications (especially things like CCNA and MSCE), it means we have to actually test people's knowledge during job interviews, which wastes a lot of time. (it's amazing how many people can't answer the exact same problems used in the exams, when the question is simply worded slightly differently)
More than a few have been highly offended - and mostly because they couldn't answer simple technical questions. One woman applying for a high-level network security position haughtily stated "I just get a man to do that" when faced with technical questions which were explicitly required in the job description and further prodding showed that the impressive list of qualifications she'd supplied was clearly imaginary and got steadily more offended, expecting to get the job simply because she was the only female candidate and "a personal friend" of one of the higher level muckety-mucks. (This was the most blatant one, most people who put fake qualifications on their CVs simply fold their tents and leave when it's clear we're checking their knowledge, but they all keep reapplying for advertised positions with those forgeries still there, clearly unaware that we keep records about the attempts and notify HR to prevent them being considered in other parts of the organisation)
I'm surprised that turtle isn't taught. It's a good way to get started in basic instructions whilst getting instant results.
As for spreadsheets/WP: ew, nasty horrble shite. This is "office/secretarial skills", not programming.
And yet, in the UK, high school computing qualifications ARE all about WP/spreadsheets/MS office
30 years ago one of the computing tasks we were given was a theoretical cpu with 12 instructions (which was 4 less than the original 256-byte CHIPOS system I first learned on a few years previously). We had to step through the assembler _on paper_ to show/understand what it was doing for various exercises and then perform several simple tasks, showing working at all stages.
Highly tedious but instructional.
I work in a space lab where postgrads and RAs are supposedly writing code for flight instrumentation or returned data but it's clear they don't understand computer fundamentals or really understand what they're doing. Pulling those same exercises out and asking them to try it got "What's this shit? I don't understand it, it's too hard".
Is it any wonder that in the last decade we've noticed a substantial deskilling of project staff? (Anyone with decent computer skills is promptly hired away at 3-4 times the salary. The last one ended up working in a patent attorney's office.)
The result is that the only people with the skills needed to actually be trusted near actual spacecraft components are all appraching retirement and the kiddies continue to splash around in the "big data" pool, usually burning mountains of computer time to give answers which are not only _wrong_ but 5 minutes analysis of the methods used would show why they're wrong.
Even staff old enough to know better will insert "fudges" into data processing systems instead of investigating why the results aren't as expected, then refining hypothesises to match observations. In one case we had to throw out 25 years of work and start over. When you see this stuff you stop wondering why spacecraft occasionally crash into Mars instead of orbiting it. The miscreants responsible can't be fired as they're tenured staff and they can't be discliplined for the same reason. "Promoting them out of the way" isn't possible either.
Our higher learning institutions are not so much "meritocracy" as "mediocracy", so it's no surprise that pervades downwards.
At some point the best way forward is to resign and become a private tutor.
You get lower stress, the kids who _want_ to learn (from all schools once the word spreads) and the oafs can continue to stumble around their intellectually-vacuous playpen unchallenged.
"I hope Hasaf chooses gaming the system over teaching a less educational class every time."
What happens is that Hasaf takes his/her skills someplace they're appreciated and he gets decent pay.
Meantime the "teachers" who simply show up and rote-teach what they've rote-learned with no actual understanding and who have strong resentment of students who are often smarter than they are, continue to alienate students from wanting to learn.
When kids starts school they love learning. That love is beaten out of them by bad teachers and by peer pressure from kids who've already learned that being bright gets unwelcome attention.
The USA has a very strong anti-intellectualism streak right across the spectrum and school boards are where it's strongest. This is nothing new and it's why the country has been stagnating for decades.
Programming is relatively easy. Programmers have been around for years and it is mostly a clerical job once the requirements are known.
Analysts are the people who tell programmers what to write after laying out the requirements and translating management bullshit into actual technical requirements whilst nailing down the goalposts so you don't get specification drift.
IE: they're there as project managers, translators and to run interference aginst the higher-ups so that the programmers can get on with their fucking jobs. It's a specialist skill and one that most programmers don't suit well.
There are successful analyst-programmers up to N point of complexity, but the falling-down stage is where management decide the analysis part is superfluous and too expensive, so they wipe out that part and try to directly interface with programmers. Ditto programmers who think that they can do what analysts do, without having had analyst and project management training, or any real experience in dealing with C-level duplicity.
A good analyst will make sure a project is on time, on budget and to spec - and ensure the specification actually works.
A better one will anticipate specification drift (within reason) and make sure it can be incorporated into the project without requiring fundamental rewrites or kludges.
An unethical one will not mention that the spec won't work as designed, make sure everything is tightly coded to the letter of the spec and then rake in 4-10 times the original budget on "snagging" fixes, often resulting in the project never working properly and usually having to be dumped entirely for something that already existed and would have done the job. This is the one who will offer the lowest original contract price.
The other analysts will often have been kicked off the tendering process after saying it won't work as specified. Management egos being what they are, they don't want to hear their Rube-Goldberg deisgn is useless.
"My first time driving in Amsterdam I pulled up at a green light because the road was busy on the other side, and the idiot behind me honked at me, went around me, stopped in the intersection, and then the light changed to red and no one from the other road could go straight anymore."
I spent a while living in Amsterdam. I'll guarantee you that driver is either driving on a non-dutch license or if he is, transferred it from a non-dutch, non-western-european one without having to do a local driving test (if he bothered getting a license at all).
"That crawling in turn frustrates drivers who are following and willing to break the limit, leading to a very noticeable increase in the number of aggressive overtakes since the limits have started dropping, as well as driving aggressively close up behind the car/cycle/whatever in front."
I've seen a few car-cams being used as evidence for dangerous driving charges. Putting Mr Angry on Youtube usually has the desired effect anyway as insurance companies are trawling there now, looking for poor driving evidence as ways of reducing their risks.
" traffic would deadlock the moment four cars arrived at a regular four-way intersection"
You do realise that
1: those weird 4-way stops are a uniquely american thing where everyone else would put a roundabout or make one direction have priority.
2: Noone ever arrives at the exact same time. First to arrive is first to move off assuming everyone's going straight on, otherwise straight-on vehicles take priority over turners.
"Besides, all of the robot crashes have been minor fender benders"
Which brings up a very important point.
_EVERY SINGLE INCIDENT_ must be reported if a robot car is involved, even when the robot is not at fault.
What do you think the actual rate of reporting of such dings is, when people are driving? Most people don't even bother to notify their insurance company for a scratch and most countries don't require reporting of non-injury incidents.
If the crash rate of AIs was _really_ twice the rate of humans I'd be extremely surprised. If it was more than _half_ the rate of humans I'd be surprised too.
When one group must report every single incident and another doesn't have to, the status will always look skewed in favour of the non-reporting group. Move it up to the level of crashes which would require onsite police attention or an insurance report and then make the comparison.
Thanks to more rigorous enforcement, what I've seen in several countries is that speeding (outside of freeways) is becoming a thing of the past.
It's creating another problem in that drivers are too busy looking at the speedometer instead of the road, but cruisecontrol and autobraking mitigate that a lot.
The shape doesn't matter, the proportions are the same - and those super-aerodynamic sports cars often have shitty friction figures at lower speeds thanks to the extremely wide tyres they need to keep operating at the higher ones.
Below about 40mph, slab-fronted/sided vehicles have virtually no difference in aerodyamic friction than sleek ones and below 55mph tyre rolling resistance is more important than aero. (narrow high pressure tyres are low resistance and generally better for traction/braking/aquaplane resistance under normal loads but we're conditioned into thinking wider tyres are better after seeing them for decades on high power sports cars)
A single 40 passenger urban transit bus does the same damage to the pavement as several thousand cars at the same speed. Heavy trucks do even more (damage is related to the 5th power of axle loading multiplied by speed)
This means that pushing more people into mass transit can have some unexpected effects on road longevity, but urban planners should have already taken that into account.
"Fast acceleration is not penalized in mileage significantly in a Prius"
It's not penalised significantly in any car. As you mentioned the enemy is braking (engine or friction) and leaving enough space to not have to dab that pedal does wonders to your milage without having to resort to "pulse and glide"
"This is similar to the idea that a driver should not be assuming that the car in front is going to dash through the yellow light, and plan on doing it also rather than realized they might brake to a halt like they are supposed to do."
The only safe way to drive is to assume everyone else is a distracted idiot - which isn't far from the truth. Multi-vehicle crashes are seldom "only one driver at fault" (and a lot of single vehicle crashes aren't single driver faults either).
The road rules are primarily saftety driven and allow humans to make dozens of glaring errors and survive because following distances, etc are all set with the aim of allowing everyone else to see the error and adjust. The fact that there are so many crashes shows just how much humans tend to take everything for granted and the fact that there aren't so many more is because the rules have such wide tolerances to allow other drivers to cope with and avoid even egrarious dangerous behaviour from an individual.
"The driver realizes the light is about to change and accelerates, but not expecting the lead vehicle to stop."
Which is still distracted driving. Consider "The lead vehicle may be stalled" as a starting point and then try arguing with a cop that it's not your fault you rearended a stationary vehicle from a standing start (or your insurance company)
video systems in the car will conclusively show who caused the damage - and I'm pretty sure that any sensible operator will add odour sensors(*) which mean the car will be flagging itself as out of action long before anyone else sits in it.
If you vomit in a taxi around here, you can be assured you'll either be looking at an extra $200-300 charges or being dropped off at the local police station instead of your stated destination, if not both.
"I think I'm going to hurl" is an amazing magic incantation for finding yourself sitting on the side of the road in less than 10 seconds, looking for another ride home. Taxi drivers are amazingly sensitive to the possibility of losing a night's income.
(*) Sensors for these things already exist and are already in use in places like subways.
"if an autonomous vehicle is twice as likely to be rear ended than a non-autonomous vehicle"
When a car is rear ended, unless you're from Mars, the legal position is that it is _ENTIRELY_ the fault of the following driver, no matter what the lead driver did, even if that was a panic stop for no reason on an freeway onramp
(The exact scenario above has happened to me - the "what the F?" reaction slowed me from hitting the brake enough that I did rear end the idiot (unlicensed and uninsured as it turned out) despite being at legal following distance, but it was still my fault for not stopping and my insurance company ate all the costs, including repairs to the car that was illegally on the road (it's a criminal offence here to drive without insurance, let alone without a license and cars are supposed to be impounded on the spot when caught but police didn't bother) and fighting off the driver and passenger's attempts to claim whiplash injuries.
Which means that (just as it is now) the raised insurance charges will go on the meatsack who didn't stop in time, not the computer which did.
Busses and other heavy vehicles are the primary target for increasing automation, not cars, even if Google is developing using cars.
For starters the extra weight of the sensors and heavy-duty processing equipment is negligible on the overall vehicle mass, plus the extra height is useful - and professional drivers are expensive, plus (on busses) vulnerable to assault, etc.
London busses no longer take cash. You prepay your fare before getting onboard. They'll be automatisation targets soon. Mercedes is already pushing its autopiloted trucks for freeway work but they're already quite capable of normal two-lane work and better at manouvering in depots than the average driver thanks to the all-round sensors
"If you have never thought about what you'd do if a couch fell from an overpass, or a tire jumped the median and headed towards you, you are an incompetent driver."
I'll guarantee that if you ask 20 people in a car park those questions 19 will treat you like you just sprouted a second head and the 20th will say "yes, occasionally"
More often, the computer using infrared as well as visible cameras and possibly lidar will have picked up the movement and already be braking long before the metsack even realises what's happening.
Hitting a deer at speed is best avoided. Like cows, a large one can easily penetrate the windscreen and kill the occupants.
Anything faster than 20mph in a residential area can be argued by statistics to be unreasonably high:
At 20mph there's 98% survivability in car vs pedestrian impacts At 30mph that drops to 90% At 35mph it's 50% at 40mph it's under 10% at 45mph it's under 1%
The old "E(k) = 0.5(mass) * (v^2)" equation...
Higher speed roads should be heavily segregated to keep pedestrians and other low speed users off and speeding enforcement in residential areas should be severe. 35mph vs 30mph being a good case in point about why
Roads have a design speed. The relationship between that and the legal speed limit is somewhat elastic.
Slashdot Mirror
It's precisely because of this kind of shenanigan that the EU stepped in to mandate compatibliity requirements - and minimum warranties for consumer equipment.
TTIP is likely to dismantle those consumer protections.
"In 10 years "programming" is going to be like "keyboarding" or "Microsoft Office" on a resume."
In the UK "programming" _IS_ using those programs.
No actual programming skills taught.
Couple that with rote-learned qualifications (especially things like CCNA and MSCE), it means we have to actually test people's knowledge during job interviews, which wastes a lot of time. (it's amazing how many people can't answer the exact same problems used in the exams, when the question is simply worded slightly differently)
More than a few have been highly offended - and mostly because they couldn't answer simple technical questions. One woman applying for a high-level network security position haughtily stated "I just get a man to do that" when faced with technical questions which were explicitly required in the job description and further prodding showed that the impressive list of qualifications she'd supplied was clearly imaginary and got steadily more offended, expecting to get the job simply because she was the only female candidate and "a personal friend" of one of the higher level muckety-mucks. (This was the most blatant one, most people who put fake qualifications on their CVs simply fold their tents and leave when it's clear we're checking their knowledge, but they all keep reapplying for advertised positions with those forgeries still there, clearly unaware that we keep records about the attempts and notify HR to prevent them being considered in other parts of the organisation)
I'm surprised that turtle isn't taught. It's a good way to get started in basic instructions whilst getting instant results.
As for spreadsheets/WP: ew, nasty horrble shite. This is "office/secretarial skills", not programming.
And yet, in the UK, high school computing qualifications ARE all about WP/spreadsheets/MS office
30 years ago one of the computing tasks we were given was a theoretical cpu with 12 instructions (which was 4 less than the original 256-byte CHIPOS system I first learned on a few years previously). We had to step through the assembler _on paper_ to show/understand what it was doing for various exercises and then perform several simple tasks, showing working at all stages.
Highly tedious but instructional.
I work in a space lab where postgrads and RAs are supposedly writing code for flight instrumentation or returned data but it's clear they don't understand computer fundamentals or really understand what they're doing. Pulling those same exercises out and asking them to try it got "What's this shit? I don't understand it, it's too hard".
Is it any wonder that in the last decade we've noticed a substantial deskilling of project staff? (Anyone with decent computer skills is promptly hired away at 3-4 times the salary. The last one ended up working in a patent attorney's office.)
The result is that the only people with the skills needed to actually be trusted near actual spacecraft components are all appraching retirement and the kiddies continue to splash around in the "big data" pool, usually burning mountains of computer time to give answers which are not only _wrong_ but 5 minutes analysis of the methods used would show why they're wrong.
Even staff old enough to know better will insert "fudges" into data processing systems instead of investigating why the results aren't as expected, then refining hypothesises to match observations. In one case we had to throw out 25 years of work and start over. When you see this stuff you stop wondering why spacecraft occasionally crash into Mars instead of orbiting it. The miscreants responsible can't be fired as they're tenured staff and they can't be discliplined for the same reason. "Promoting them out of the way" isn't possible either.
Our higher learning institutions are not so much "meritocracy" as "mediocracy", so it's no surprise that pervades downwards.
At some point the best way forward is to resign and become a private tutor.
You get lower stress, the kids who _want_ to learn (from all schools once the word spreads) and the oafs can continue to stumble around their intellectually-vacuous playpen unchallenged.
"I hope Hasaf chooses gaming the system over teaching a less educational class every time."
What happens is that Hasaf takes his/her skills someplace they're appreciated and he gets decent pay.
Meantime the "teachers" who simply show up and rote-teach what they've rote-learned with no actual understanding and who have strong resentment of students who are often smarter than they are, continue to alienate students from wanting to learn.
When kids starts school they love learning. That love is beaten out of them by bad teachers and by peer pressure from kids who've already learned that being bright gets unwelcome attention.
The USA has a very strong anti-intellectualism streak right across the spectrum and school boards are where it's strongest. This is nothing new and it's why the country has been stagnating for decades.
Programming is relatively easy. Programmers have been around for years and it is mostly a clerical job once the requirements are known.
Analysts are the people who tell programmers what to write after laying out the requirements and translating management bullshit into actual technical requirements whilst nailing down the goalposts so you don't get specification drift.
IE: they're there as project managers, translators and to run interference aginst the higher-ups so that the programmers can get on with their fucking jobs. It's a specialist skill and one that most programmers don't suit well.
There are successful analyst-programmers up to N point of complexity, but the falling-down stage is where management decide the analysis part is superfluous and too expensive, so they wipe out that part and try to directly interface with programmers. Ditto programmers who think that they can do what analysts do, without having had analyst and project management training, or any real experience in dealing with C-level duplicity.
A good analyst will make sure a project is on time, on budget and to spec - and ensure the specification actually works.
A better one will anticipate specification drift (within reason) and make sure it can be incorporated into the project without requiring fundamental rewrites or kludges.
An unethical one will not mention that the spec won't work as designed, make sure everything is tightly coded to the letter of the spec and then rake in 4-10 times the original budget on "snagging" fixes, often resulting in the project never working properly and usually having to be dumped entirely for something that already existed and would have done the job. This is the one who will offer the lowest original contract price.
The other analysts will often have been kicked off the tendering process after saying it won't work as specified. Management egos being what they are, they don't want to hear their Rube-Goldberg deisgn is useless.
They could call them "Top Trumps"
"My first time driving in Amsterdam I pulled up at a green light because the road was busy on the other side, and the idiot behind me honked at me, went around me, stopped in the intersection, and then the light changed to red and no one from the other road could go straight anymore."
I spent a while living in Amsterdam. I'll guarantee you that driver is either driving on a non-dutch license or if he is, transferred it from a non-dutch, non-western-european one without having to do a local driving test (if he bothered getting a license at all).
"That crawling in turn frustrates drivers who are following and willing to break the limit, leading to a very noticeable increase in the number of aggressive overtakes since the limits have started dropping, as well as driving aggressively close up behind the car/cycle/whatever in front."
I've seen a few car-cams being used as evidence for dangerous driving charges. Putting Mr Angry on Youtube usually has the desired effect anyway as insurance companies are trawling there now, looking for poor driving evidence as ways of reducing their risks.
" traffic would deadlock the moment four cars arrived at a regular four-way intersection"
You do realise that
1: those weird 4-way stops are a uniquely american thing where everyone else would put a roundabout or make one direction have priority.
2: Noone ever arrives at the exact same time. First to arrive is first to move off assuming everyone's going straight on, otherwise straight-on vehicles take priority over turners.
The _reported_ rate.
The bigger problem is that human-vs-human trivial incidents normally don't get reports.
"Besides, all of the robot crashes have been minor fender benders"
Which brings up a very important point.
_EVERY SINGLE INCIDENT_ must be reported if a robot car is involved, even when the robot is not at fault.
What do you think the actual rate of reporting of such dings is, when people are driving? Most people don't even bother to notify their insurance company for a scratch and most countries don't require reporting of non-injury incidents.
If the crash rate of AIs was _really_ twice the rate of humans I'd be extremely surprised. If it was more than _half_ the rate of humans I'd be surprised too.
When one group must report every single incident and another doesn't have to, the status will always look skewed in favour of the non-reporting group. Move it up to the level of crashes which would require onsite police attention or an insurance report and then make the comparison.
"but that's what everyone does:"
Thanks to more rigorous enforcement, what I've seen in several countries is that speeding (outside of freeways) is becoming a thing of the past.
It's creating another problem in that drivers are too busy looking at the speedometer instead of the road, but cruisecontrol and autobraking mitigate that a lot.
The shape doesn't matter, the proportions are the same - and those super-aerodynamic sports cars often have shitty friction figures at lower speeds thanks to the extremely wide tyres they need to keep operating at the higher ones.
Below about 40mph, slab-fronted/sided vehicles have virtually no difference in aerodyamic friction than sleek ones and below 55mph tyre rolling resistance is more important than aero. (narrow high pressure tyres are low resistance and generally better for traction/braking/aquaplane resistance under normal loads but we're conditioned into thinking wider tyres are better after seeing them for decades on high power sports cars)
"and reduces wear and tear on the pavement."
A single 40 passenger urban transit bus does the same damage to the pavement as several thousand cars at the same speed.
Heavy trucks do even more (damage is related to the 5th power of axle loading multiplied by speed)
This means that pushing more people into mass transit can have some unexpected effects on road longevity, but urban planners should have already taken that into account.
"Fast acceleration is not penalized in mileage significantly in a Prius"
It's not penalised significantly in any car. As you mentioned the enemy is braking (engine or friction) and leaving enough space to not have to dab that pedal does wonders to your milage without having to resort to "pulse and glide"
"This is similar to the idea that a driver should not be assuming that the car in front is going to dash through the yellow light, and plan on doing it also rather than realized they might brake to a halt like they are supposed to do."
The only safe way to drive is to assume everyone else is a distracted idiot - which isn't far from the truth. Multi-vehicle crashes are seldom "only one driver at fault" (and a lot of single vehicle crashes aren't single driver faults either).
The road rules are primarily saftety driven and allow humans to make dozens of glaring errors and survive because following distances, etc are all set with the aim of allowing everyone else to see the error and adjust. The fact that there are so many crashes shows just how much humans tend to take everything for granted and the fact that there aren't so many more is because the rules have such wide tolerances to allow other drivers to cope with and avoid even egrarious dangerous behaviour from an individual.
"The driver realizes the light is about to change and accelerates, but not expecting the lead vehicle to stop."
Which is still distracted driving. Consider "The lead vehicle may be stalled" as a starting point and then try arguing with a cop that it's not your fault you rearended a stationary vehicle from a standing start (or your insurance company)
video systems in the car will conclusively show who caused the damage - and I'm pretty sure that any sensible operator will add odour sensors(*) which mean the car will be flagging itself as out of action long before anyone else sits in it.
If you vomit in a taxi around here, you can be assured you'll either be looking at an extra $200-300 charges or being dropped off at the local police station instead of your stated destination, if not both.
"I think I'm going to hurl" is an amazing magic incantation for finding yourself sitting on the side of the road in less than 10 seconds, looking for another ride home. Taxi drivers are amazingly sensitive to the possibility of losing a night's income.
(*) Sensors for these things already exist and are already in use in places like subways.
"if an autonomous vehicle is twice as likely to be rear ended than a non-autonomous vehicle"
When a car is rear ended, unless you're from Mars, the legal position is that it is _ENTIRELY_ the fault of the following driver, no matter what the lead driver did, even if that was a panic stop for no reason on an freeway onramp
(The exact scenario above has happened to me - the "what the F?" reaction slowed me from hitting the brake enough that I did rear end the idiot (unlicensed and uninsured as it turned out) despite being at legal following distance, but it was still my fault for not stopping and my insurance company ate all the costs, including repairs to the car that was illegally on the road (it's a criminal offence here to drive without insurance, let alone without a license and cars are supposed to be impounded on the spot when caught but police didn't bother) and fighting off the driver and passenger's attempts to claim whiplash injuries.
Which means that (just as it is now) the raised insurance charges will go on the meatsack who didn't stop in time, not the computer which did.
Busses and other heavy vehicles are the primary target for increasing automation, not cars, even if Google is developing using cars.
For starters the extra weight of the sensors and heavy-duty processing equipment is negligible on the overall vehicle mass, plus the extra height is useful - and professional drivers are expensive, plus (on busses) vulnerable to assault, etc.
London busses no longer take cash. You prepay your fare before getting onboard. They'll be automatisation targets soon. Mercedes is already pushing its autopiloted trucks for freeway work but they're already quite capable of normal two-lane work and better at manouvering in depots than the average driver thanks to the all-round sensors
"So unexpected objects will appear in the road"
They are only unexpected because you weren't paying attention. Automated vehicles are, and looking in all directions (including at the overpasses).
"If you have never thought about what you'd do if a couch fell from an overpass, or a tire jumped the median and headed towards you, you are an incompetent driver."
I'll guarantee that if you ask 20 people in a car park those questions 19 will treat you like you just sprouted a second head and the 20th will say "yes, occasionally"
"Sometimes, going at speed and a deer jumps out "
Sometimes.
More often, the computer using infrared as well as visible cameras and possibly lidar will have picked up the movement and already be braking long before the metsack even realises what's happening.
Hitting a deer at speed is best avoided. Like cows, a large one can easily penetrate the windscreen and kill the occupants.
"The speed limit is set unreasonably low"
Anything faster than 20mph in a residential area can be argued by statistics to be unreasonably high:
At 20mph there's 98% survivability in car vs pedestrian impacts
At 30mph that drops to 90%
At 35mph it's 50%
at 40mph it's under 10%
at 45mph it's under 1%
The old "E(k) = 0.5(mass) * (v^2)" equation...
Higher speed roads should be heavily segregated to keep pedestrians and other low speed users off and speeding enforcement in residential areas should be severe. 35mph vs 30mph being a good case in point about why
Roads have a design speed. The relationship between that and the legal speed limit is somewhat elastic.