genec

ILTB only works from data gathered over time by reputable sources... he then examines the techniques used to gather said data and determines the truth or validity of claims made. (in his lofty opinion...)

Since actual self driving cars are still in the prototype stage, and data is not being readily offered regarding the collision avoidance safety of said vehicles, ILTB can just play curmudgeon and deny any proposed goals presented by "marketeers."

For instance, with regard to AI, he is just liable to respond:
Where is my robot?

And really, the bottom line is that the stuff ain't yet here, marketeers have promised a lot with tech (and even today certain aspects of high-speed internet enabled smart phones don't meet some of the hype) so ILTB is somewhat justified playing the Luddite.

But to look at the question, posed by the OP, in an objective way... "Would a self driving car World make it safe for cyclists?" We can only respond that if such a vehicle obeyed all the laws and rules (unlike a human drivers, who often shortcut said rules/laws), there is a high probability that indeed it would be safer for cyclists.

But at this point, in the infancy of self driving vehicles... nobody really knows how things are going to play out... least of all, all the voices here, speculating on such technology without even consulting any resource about the technology... Heck, we have opinions here based solely on the caveat "But I like to drive...," and "I have heard..."

wphamilton

I think we can justifiably say more than that - AI algorithms have progressed beyond the simple expert system (rules based processing) paradigm that you're implying here.

I'm not speaking as a Machine Learning expert by any means, but I have studied the technology off and on over the years so I wouldn't quite classify myself in that latter group. While it's true that with cars we're still looking at prototypes, but state of the art is already sufficient for this application. That's not saying that it should be easy, because there is a lot of work left for development and testing, and technology will advance, but it's already there.

unterhausen

I just saw something recently where they said that most of the data taken by the autonomous cars on the road actually gets downloaded by hand and then shipped off to data centers in India for supervised learning. It's just too much data to transmit wirelessly or process on-board. Of course, if it was being processed on-board, that wouldn't be supervised learning.

Anyone that has tried to get a machine learning algorithm to learn to recognize things in an image either realizes how fragile the algorithms are or was never interested in noticing that in the first place. It's amazing how often they are right, but then again, they are not right that often. Scale invariance is a big problem, so of course people claim it a lot.

genec

Because of the work yet remaining, there is little data on how well self driving cars actually perform in the real world; just projections. (never mind that we design fighter aircraft based on "projections").

ILTB can still stick to his guns and say... "it ain't there yet." (just as new design fighter aircraft may also need new tuning or pilot training).

While I tend to agree with you, and personally hope self driving cars will be readily available when the time comes for advancing age to dictate that I too should no longer be behind the wheel, ILTB can continue to debate what is and what isn't, and how much is just marketing fancy.

Until there are copious amounts of data showing otherwise... anything at this point is speculation... although admittedly there are some pretty intelligent folks working on getting this right.

BTW I come at this from the hindsight of watching lots of speculators tell the cell phone industry that CDMA would not work, violated the laws of physics, and was flat out doomed. I watched that technology from "the inside" hearing what engineers were telling me it could do, while reading the denials in the marketplace by "experts."
https://forgetmiosotis.blogspot.com/2...f-physics.html

CDMA is the heart of 4G technology, which has placed the internet in our hands though smart phones. (remember dialup wired modems...)

I saw this progress from a couple of refrigerator sized cryptographic vocoders back in my Navy days, to the hand held devices we have today. Technology does indeed "march forward." I saw cell phones progress from crude RF transceivers with a whip antenna to the multi-band multimode, multi-radio devices with no visible antenna that we have today.

Most of us of this current generation have seen computers progress from the '60s era refrigerator size machines with the blinky lights and whirring tape drives to something that we have in our home offices that can down load movies and play games... the latter which were well off on the horizon back in the '60s (pinball anyone?)

I have no doubt that "technology marches forward."

I-Like-To-Bike

Why don't you just post what you think on the subject, instead of paraphrasing my words and/or guessing what I might or might not say?

The bottom line of your recent stilted and awkward posts repeatedly referencing your interpretations of what I might say, is that as much as you don't want to, you agree with my posts on the flimsiness of predictions of substantial reduction in risk to bicyclists or anybody else based on wishful thinking, conjecture and dreamy projections about what might happen with autonomous self driving automobiles.

unterhausen

moderator's note: the subject of the thread is self driving cars and if they would make the roads safer for cyclists. It is not each other. This back and forth is very close to being unacceptable. Please let other people speak for themselves instead of turning their posts into a springboard for strawman arguments. Thanks.

genec

Sure, here is what I think:I also think the technology is advancing faster than you believe.

coominya

They are looking for guinea pigs are they
Like those people who get paid to trial a new drug and end up having seizures, no thanks.

KD5NRH

Lots of situations would be greatly improved with some basic information passed between the vehicles. For example, I've sat in jams on I-30 where it has four through lanes and an exit lane. The actual fender bender was out of view ahead, and people jockeying for whatever lane they thought was going fastest turned the whole thing into a mess. A simple "left two lanes open, right two through lanes and exit lane blocked at this location" would allow both auto-drive systems and driver alert systems for manual drive cars to work toward lining up in the clear lanes, and simply cruise on through rather than playing lane roulette until the blockage finally comes into view, then rushing to cross 2-3 lanes to get to the right ones.
That could be achieved fairly easily with something no more powerful than normal WiFi if the vehicles are also set up to repeat the message to all others in range unless they're outside a certain (maybe speed limit dependent) radius. The actual message need not be particularly human-readable, since the computer will interpret it anyway, and thus the bandwidth of a typical message would be tiny, likely along the lines of a tweet for even a fairly complex situation.

genec

Having cars "connected" has long been a concept, for just the reason you state... that it can make driving smoother. But having a cars moment by moment decision making rely on such a connection is not workable. No central net, and no is there a car signal coming from around the corner in the fog. Nor are people or cyclists required to wear "transponders."

"Mesh networking" should be an adjunct to a fully autonomous vehicle. The car should be able to take car of itself in all conditions presented, and should avail itself to any wider area "news" as needed... This actually is quite similar to using a GPS with traffic notification as a human driver... you plan your route, then while actually driving, you manage the moment to moment situation with traffic, but you "look ahead" to determine if your route is still valid by using "traffic condition info." Google offers such traffic info as does a sub-channel that some GPSs can access (some by subscription, some as a feature).

Here is a link to some thoughts about "connected vehicles..." Or V2V.
v2v | Brad Ideas

coominya

This thread has moved from the sublime to the hilarious. Every advocate seems to have master plan how it will all work out if they just do xyz

genec

Hey, just remember, YOU don't have to use one... and since this is happening in the US and not OZ, it really isn't an issue for you. Besides, they would not work for you... you folks drive on the wrong side of the road.

Rollfast

But they have MUCH hotter cars.


Sheesh. maybe YOU drive on the wrong side?

tyrion

No, we drive on the right side.

Rollfast

If we keep arguing about left and right sides this can only end up in Politics and Religion.

I-Like-To-Bike

Not always so, at least not at Tesla. IT seems to be a bit of a problem there. Or non responsive IT is a good excuse for covering up serious internal problems.
https://www.wired.com/story/cancelin...timely-refund/

tyrion

That's not a technological problem, just good old fashioned bad business.

mr_bill

IT = Information Technology! And lapis philosophorum too!

-mr. bill

I-Like-To-Bike

Did you read the article? One of the myriad excuses used by Tesla was "IT issues", maybe they really cannot track year old accounts, or just didn't think it was an item worth tracking. Perhaps bicycle riders may get the same treatment from their IT folks.

InOmaha

After SkyNet becomes self aware, all the semi-autonomous interconnected software driven vehicles will have an easy time targeting the pedestrians and cyclists.

Before that it might cut down on car crashes. However, the software will need to be very smart or pedestrians and cyclists will be able to easily stop traffic. The result could be increased legislation and design to limit non-car traffic in more areas, thereby further reducing the non rational human factor. You know, things like jaywalking and cyclists running red lights.

I-Like-To-Bike

Or you know, things like walking and cycling anywhere such unprogrammed activities might make it difficult for autonomous cars to navigate without human assistance.

tyrion

I don't care what their excuse is, it's not a technological problem.

InOmaha

If you add human override and assistance, they are no longer autonomous cars, and we're back to where we are now.


I could see problems in most dense cities where cars and people interact quite a bit. As it is now, pedestrians do a risk assessment when dealing with cars. I've been in places like New York where a large enough group of pedestrians can block traffic past the point where they were supposed to stop walking. Because traffic is slow and the risk with a group is low. Someone jaywalking by themselves will look at traffic, gauge the risks invoved and step out and cross at varying speeds depending on how fast traffic is moving. The faster traffic moves, the faster they jaywalk. It's the interaction between two rational/irrational actors evaluating the risk potential that keeps both moving/slowing down/stopping without a negative outcome.


Autonomous cars are not allowed to be irrational. It's hard enough to program the logic involved with operating within the rulebook and no IT programmer is going to program the car to ignore jaywalking pedestrians due to litigation and bad plublicity. So the interaction between car and pedestrian would always be rational car versus rational/irrational human. Once people pick up on the fact that the car will always default to a pedestrian regardless of all other traffic rules, people can act more irrationally and shut down car traffic simply because they need to get somewhere and no longer need to care they could be hit. The car will always stop for them even if the light is green or the car has the right of way on the street.


The bigger the city, and the more human autonomous car interaction, the more likely there will be irrational people stopping cars that have the right of way because they don't care. The result of that is humans will take control over the autonomous cars in non-highway situations to put risk back onto the pedestrians, thereby negating any safety gains pedestrians, cyclists, or drivers my see from the technology AND/OR more roads throughout the city will be designed as highways to limit pedestrians or cyclists so they can't be irrational around autonomous cars. The second method would help ease congestion and likely be supported by a majority of autonomous car owners, and therefore politicians.


My opinion is, some crowded cities like New York might not change at all with the new technologies, while the more spread out cities might actually see even less pedestrian access and more restrictions and limitations on "non-autonomous" vehicles on roads. That way traffic on the side streets in places like LA will flow better.


For the majority of the cities it would be politically easier to limit bicycles to specific streets with bike lanes or push them off on out of the way dedicated bicycle paths so cyclist don't interefere with cars that are programmed to strickly follow the rules. Motorcyles would likely have a similar fate.


The more autonomous transportation becomes, the less reason to give people a individual transportation choices and personal control.

mr_bill

Wow, is there a dystopian science fiction convention in town?

One thing is certain, when an autonomous vehicle kills or seriously injures someone on a bicycle in New York City, NYPD will be there to ticket blitz people on bikes.

-mr. bill

Ninety5rpm

Waymo (Google) is leading the way with a revolutionary approach. Tesla has a very different trial-and-error evolutionary approach. Not sure about the others. But Tesla having issues is inherent to the approach they're taking.