Predictions about autonomous cars

Article says:

Our trust in digital technology companies has become so complete, that when a host of these promised self-driving cars, we believed them. Everyone believed them.

Well, not quite everyone. I didn't. I made a prediction in 2014 that they will become ubiquitous by 2060 and certainly not within a period of 5-20 years as the "experts" were predicting. Why people take note of preposterous predictions from the likes of Elon Musk is beyond me.

More on autonomous cars

I
do not believe that level 5 autonomous cars i.e cars that can drive
anywhere and don't even have a steering wheel, will be possible if they
share the roads with human driven cars. They'll need their own roads,
or at least their own lanes. Even then they won't make an appearance
for many years yet.



Level 4 cars that are restricted to certain
areas and are not suitable for inclement weather conditions will be vastly easier to achieve. But, even here, sharing roads with human
drivers is perhaps an insurmountable challenge.

Autonomous cars and human drivers are incompatible

I've been thinking. It just simply won't be possible to create fully autonomous cars (i.e level 5 cars) that share the same roads as human drivers. Google etc are wasting a lot of their money attempting to do this.



There needs to be a rethink. Autonomous cars need their own lanes. Stop wasting money Google, Musk etc and give it to me instead.



Also see my original blog post on these cars here, and another post here.

Fully autonomous cars

I'm reading a lot in the past couple of days essentially saying that people are being irrational for being afraid to use fully autonomous cars.


I don't agree it's irrational at all. We have to bear in mind that computers can't actually see. It's difficult for them to make sense of their environment; certainly if they merely go by light as we do. Of course, hopefully, they'll use lidar. But there's cheap lidar systems, and expensive one's. You can be sure that they'll use the cheapest ones they can get away with. Their number one priority, after all, is to make a profit.

Pot holes, debris on the road, crisp packets, rabbits running across the road. The number of unknowns the environment can throw at the car is unlimited. How do you program the car to distinguish between an empty carrier bag which it can drive through, and a rock? If they brake too suddenly, cars behind driven by people might well crash into the back.


What happens when ambulances, fire engines etc with the sirens blasting are in the nearby vicinity? What happens with road works? What about the possibility of remote hacking? How are they going to negotiate inclement weather conditions like heavy snow?


And these cars will be complex. Mechanically complex like other cars, but also complex in terms of all the programming. Suppose a few lines of code cause problems? Happens all the time with computer games and they have to issue patches (which might resolve the original problem, but cause a whole new load of problems). But your life is at stake, you cannot afford to have them programmed incorrectly!


It just seems sensible to me to wait until many people have travelled safely in them, and their track safety records exceed that of people driven cars, before getting into one.

Also see my blog post:

Self-Driving Cars

Self-Driving Cars

I tend to pooh pooh predictions regarding new technology and how it will utterly revolutionise our lives. And it seems to me such predictions have a very poor track record. 

 
But I think self-drive cars -- more specifically those that do not allow humans to take over control -- will eventually completely replace human driven cars and utterly revolutionise the world.



Which is not to deny there will be many problems to solve before we get there. For one thing they'll initially encounter a lot of resistance due to the fact that many people -- predominantly men -- just simply love to drive. In addition we presumably will see a mixture of self-drive and human driven cars on the roads at first, and this will create formidable problems. And what of problems such as kids jumping out in front of the cars to see them automatically stop! And I question whether a car which looks like one that Noddy might drive, such as the google self-drive car depicted in the photograph above, will ever become popular.


So ironing out all the problems will take a great deal of time, and we're talking here probably of a few decades. But I'm going to be bold here. I predict that by 2060 no one will drive cars any more on public roads. And practically no one will die any more on the roads.

Update 23/10/14   A very negative article I've just come across:

http://www.slate.com/articles/technology/technology/2014/10/google_self_driving_car_it_may_never_actually_happen.single.html



Article says:

But the biggest issue with the Google car is one that has bedeviled computer researchers for as long as computers have been around: how to endow the machines with the sort of everyday knowledge that humans acquire and use from childhood on.

It's going to be incredibly difficult -- albeit not impossible -- to endow them with that sort of knowledge because they are not conscious and never will be. But presumably people were aware of all these problems. I myself hint at them above.

I certainly don't agree it will never happen. A mixture of normal and self-drive cars will be very problematic. But all cars being self-drive is altogether a different issue. But how do we get to the scenario whereby all cars are self driving without going through the problematic situation of both normal and self-drive cars being on the roads? Perhaps in reality there will be a gradually transition with cars becoming more and more autonomous until the point where they become fully self-driving.

A Further Update 17/01/15

I've just been having an argument with a certain James Burkill on facebook about what computers are capable of. The subject turned to self-driving cars. I showed him this blog entry. He said to me:

Tesla already have self driving cars on the market. And they already drive better than humans. They can see through fog. It can decide on the precise torque required for the forward or rear engine to maximise efficiency. Driving is a perfect application for a machine learning algorithm. There is no shortage of data.


In two years from now all car companies will be offering models with autopilot. By 2020 we will have cars without steering wheels or pedals. By 2022 there will be debates raging over whether we should allow humans to drive due to safety concerns. By 2027 humans will be banned from driving in most western countries.

So his prediction for when no one will drive cars any more on public roads is 2027, compared to my 2060. Quite a bit of difference! I think his prediction is hopelessly optimistic. There again I know virtually nothing about computers. But we shall see.

And yet another Update! 29/08/18

See 2 further blog posts I've made in my other blog on this topic:



https://ianwardells.blogspot.com/2017/08/fully-autonomous-cars.html





https://ianwardells.blogspot.com/2018/08/autonomous-cars-and-human-drivers-are.html