Autonomous Vehicles

[sesevans]

Interesting note below about Google's automated AI tool, an enhanced version of AutoML (machine learning), and how it outperforms human programmers on two industrial strength neural networks. AutoML is a tool developed by the Google Brain team that automates the production of deep learning models. Google making important strides in computer vision with the help of machine learning tools.

The researchers believe that the features learned by their newest application can be reused in many other computer vision applications such as with self driving vehicles and security monitoring.

https://research.googleblog.com/2017/11/automl-for-large-scale-image.html?m=1

 

[sesevans]

...........

I'd definitely buy an EV if it looked like that Honda concept. Looks cooler than any other EV I've seen in recent years.

I appreciate you would like to buy a self driving car like the Honda concept but according to the CEO of Aptiv, the company formed from the split up of Delphi Automotive, and the one with its focus on autonomous vehicle support, the first use cases will be in the commercial market and not the individual market because of the additional costs associated with LIDAR. Aptiv assumes LIDAR use in robo-taxis and clearly does not buy into the concept of no LIDAR AVs. This CEO believes we won't see the individual market for AVs to develop until 2025 when LIDAR costs less than $ 5000.

While I believe this is probably a conservative view, it is still the view of an industry expert.

https://www.bloomberg.com/news/arti...ees-2018-as-dawn-of-robot-taxis-delivery-cars

 

[Nasher]

Thing is how well is LIDAR going to work when everyone around you has it? Surely there will be interference from other LIDAR systems?

 

[sesevans]

menas noothing, every single tesla rolling of the factory and have been rolling of teh factory for several years, drives in shadow mode whether its in auto pilot or not. it is gathering far more data than any other entity. This is extremely important.

you can not map the entire country and expect roads to never change, this is why waymo is cheating and why currently it can not roll it out.

three of your points has nothing to do with waymo and applies to all autonomous cars.

@Glaucus

While we politely disagree on the concept of LIDAR and mapping and rollouts, I still consider Tesla an awesome company in many ways, but one with a very competitive landscape.

By that I mean that the market for "sustainable transport" (Tesla USP) will continue to attract fierce competition from some of the largest tech companies in the world (Google, Apple, etc) which have a superior access to risk capital and talent. These competitors to Tesla will be able to make money from vehicle data and content opportunities threatening Tesla's long term independence. Also Tesla seems completely dependent on Elon Musk, whereas competitors have no single-person risk in the same way.

I recall that Tesla has had some close calls in recent years with access to capital. I have recently been reading that Elon's natural advantage is more with SpaceX than with Tesla and there is some considerable overlap between the two entities: Tesla has already cited help given to them by SpaceX engineers on materials science and engineering. I am sure you would agree that the overlap also includes machine learning and AI that can benefit each.

I wonder if we will see Tesla seeking a closer and formal partnership with SpaceX (which can attract significant capital) and whether this might lead eventually to a merger of the two entitites? Tesla has already merged with its SolarCity unit.

Do you have any comments on this possibility?

 

[sesevans]

Thing is how well is LIDAR going to work when everyone around you has it? Surely there will be interference from other LIDAR systems?

A quick Google search on "LIDAR interference with other vehicles?" yielded this among other searches. You might wish to follow this link and other links on that search query:

https://robotics.stackexchange.com/questions/10770/do-multiple-lidar-systems-in-same-area-interfere

 

[Glaucus]

I wonder if we will see Tesla seeking a closer and formal partnership with SpaceX (which can attract significant capital) and whether this might lead eventually to a merger of the two entitites? Tesla has already merged with its SolarCity unit.

Do you have any comments on this possibility?

from what I remember they haven't had issues with capital for years, but they do funding runs regularly as it costs a lot to build infrastructure.
as for a closer tie to spacex then pretty much no. They share key personnel and they share knowledge and technology(ad i expect when spacex has its internet constellation up, the cars would switch to using that network), but they aren't going to merge or anything like that. For one tesla is publicly traded and spacex is not and will not be for decades.

https://www.bloomberg.com/news/arti...taffers-tesla-and-spacex-share-more-than-musk

 

[sesevans]

from what I remember they haven't had issues with capital for years, but they do funding runs regularly as it costs a lot to build infrastructure.
as for a closer tie to spacex then pretty much no. They share key personnel and they share knowledge and technology(ad i expect when spacex has its internet constellation up, the cars would switch to using that network), but they aren't going to merge or anything like that. For one tesla is publicly traded and spacex is not and will not be for decades.

https://www.bloomberg.com/news/arti...taffers-tesla-and-spacex-share-more-than-musk

Interesting comments. Thanks for sharing the Bloomberg link which is where I assume I first saw the reference to the potential for a more formal business partnership between SpaceX and Tesla, beyond material science and manufacturing assistance which has already occurred and is continuing. And yes the benefit of a proprietary LEO to offer secure data is clear.

I am still of the view that the competition for AVs will become extremely competitive and access to capital and/or having vast capital resources will set apart the winners from the losers. I think it is possible that Tesla will find access to capital tougher in future.

If true, it may be sensible for Tesla to consider a more formal partnership or ownership of SpaceX because I believe the SpaceX competitive moat is better than the Tesla competitive moat. I have read that the global space economy could be worth more than $ 1 trillion, probably much greater than the addressable and highly competitive market for Tesla alone.

As previously mentioned, we already know that self driving vehicles winners will leverage expertise in machine learning and deep learning systems. Space networks will also benefit from this expertise. There seems a lot of overlap between the two applications.

An interesting link today from DigiTimes discusses the likely surge in developing smart cities. The CEO from the Taiwanese company discusses their involvement with sensors and AI and how smart lighting will be an early starting point. As you know, Google has already set up Sidewalk Labs which is exploring the development of a smart city on Toronto's Eastern Waterfront area. Smart self driving cars are discussed too.

http://www.digitimes.com/news/a20171204PD205.html

 

[sesevans]

How powerful and disruptive can the move to autonomous vehicle ride sharing really be?

Short answer: very powerful and very disruptive, according to Trinitonian.com.

Some factors they consider:

1. Oil prices could crash with crude oil demand dropping 30% due to TAAS (transportation as a service.)
2. RethinkX predicts that TaaS “will offer vastly lower-cost transport alternatives — four to 10 times cheaper per mile than buying a new car and two to four times cheaper than operating an existing vehicle in 2021” and that “TaaS will provide 95 percent of the passenger miles traveled within 10 years.”
3. Positive feedback loop for both AV use and cost of insuring AVs.

https://trinitonian.com/the-future-of-autonomous-cars/

 

[Nasher]

They always go on about changes making things cheaper. Since when has that ever been true?

 

[Entai]

Need to look at the long term though.

When desktop computers first came out, they were many thousands of pounds, and everyone said they would be the death of the workforce, and that there would be huge unemployment when computers took over.

A few decades later and computers are everywhere, doing everything, where is massive the unemployment ? Where is everyone having huge amounts of leisure time as the computers are doing everything ? , How cheap are computers now ?

In another several decades, autonomous cars will be everywhere, they will be as cheap, if not cheaper than current mass produced vehicles currently, no one will bat an eyelid, and we will all (those of us still around) still be doing the same old things.

 

[jigger]

Hmm that's fair, but I don't have much faith in a system that just seems to always been on an upwards source of increasing costs for everything.

Cost and need will be the two biggest factors. Personally I don't need self driving sysyem as I'm better driving myself and I wouldn't pay for the extra cost over a non self driving car.

 

[sesevans]

Cost and need will be the two biggest factors. Personally I don't need self driving sysyem as I'm better driving myself and I wouldn't pay for the extra cost over a non self driving car.

Fair enough. It is all about choice.

For me, I prefer to avoid driving when possible as I really dislike traffic queues yet having to focus on the road rather than doing something else that I would prefer. I firmly believe that you can be the best driver in the world but you are squaring up with other human drivers who could well be the cause of an accident affecting you whether you like it or not: drunk driving, road rage, distraction, argument with a passenger, etc etc. If more or all cars were fully AVs one day, it would likely decrease the likelihood of an accident as the AVs would be connected and communicating with each other.

As for cost, if it is proven that AVs are safer than cars driven by humans (in fact I see AVs one day being many more times safer than human drivers), it will have two effects: network effect and insurance reduction. The network effect will spread like social networks----if your friends do it you might be more inclined to join up, and so on. The "network" will actually improve the safety and performance as more miles will be "driven" by AVs and they will learn and improve over time, something that a human is incapable of doing. As to cost, I see ride sharing per mile without a human driver costing much less than privately owned vehicles on a per mile basis. Better safety, more data, fewer accidents and insurance premiums on ride sharing and eventually on privately owned AVs will be significantly less.

Both the network effect and the lower cost should drive more people to use AVs, first ride sharing and eventually owned.

 

[sesevans]

For the Tesla fans I see that Elon Musk spoke about developing their own self driving vehicle chips/processors. Initially they worked with Mobileye and latterly with Nvidia but now wish to develop their own solution entirely.

https://arstechnica.com/cars/2017/1...ips-will-be-the-best-in-the-world/?comments=1

 

[sesevans]

@Glaucus

Beyond the link in Ars that I highlight above, I think you might find it interesting what Elon Musk said this week:

“about half of new cars built ten years from now will be autonomous”

"artificial general intelligence (AGI) will arrive in about seven or eight years"

"But I think there are a lot of ways that AI can be useful short of being god-like".

 

[Participant]

I've ordered the drive pilot package on my E Class and whilst it's not a fully autonomous system like Tesla's, I simply cannot wait to be able to navigate the queues into work without having to brake, accelerate, or steer.

If I could sit and read a book or have a sleep whilst my car took me to places that would be nirvana to me.

The UK economy would also become x% more efficient as more people could work on their way to the office / home, not to mention the reduced commuting time from fewer crashes and idiotic driving decisions.

It's going to be great.

 

[sesevans]

I've ordered the drive pilot package on my E Class and whilst it's not a fully autonomous system like Tesla's, I simply cannot wait to be able to navigate the queues into work without having to brake, accelerate, or steer.

If I could sit and read a book or have a sleep whilst my car took me to places that would be nirvana to me.

The UK economy would also become x% more efficient as more people could work on their way to the office / home, not to mention the reduced commuting time from fewer crashes and idiotic driving decisions.

It's going to be great.

I believe that the first introduction of AVs commercially will be in the form of one or both of the following:

ride-sharing AVs--with capacity for several or more passengers
and
AV trucks for moving goods

An example of the latter is being planned in Racine, Wisconsin because of the construction to begin soon on a major investment by a leading electronics company, Foxconn. Once built it would represent the largest industrial project in State history. Working with the Wisconsin State Dept of Transport, it appears that their plan will include moving people and goods in AVs on carefully monitored roads that will be upgraded to provide for CAV--connected and autonomous vehicles. Communication between vehicles will also be assisted by smart street lights.

In planning a smart city around Racine both in the city and on its road to the airport, this project represents the tip of the iceberg for industrial and commercial applications of AVs. This idea of a smart city can be replicated in many towns and cities around the world.

Today the cost of LIDAR, an essential feature on any self driving vehicle I would want to ride in is still high. This alone would boost the price of the vehicle that any one person might wish to buy, limiting its mass market appeal. I see this price dropping with production at scale but the commercial applications such as ride sharing will be able to absorb this higher cost AV sooner than the individual buyer. After all, the typical individual vehicle is used less than 5% of the time while the ride sharing vehicle can be used 50-60% of the time and pay for itself relatively quickly (the cost of the sensors and LIDAR is offset by the lower cost of no driver to pay for).

https://www.jsonline.com/story/news...-thousands-employees-racine-county/931773001/

 

[jigger]

Honestly the self driving truck is a huge liability. Who gets sued when they cause carnage? Who gets a prison sentence?

 

[Glaucus]

Honestly the self driving truck is a huge liability. Who gets sued when they cause carnage? Who gets a prison sentence?

why would anyone get a prison sentence and it all depends on how each country writes autonomous vehicle laws.
just because there's an accident doesn't mean anyone did anything wrong.
whatever you think trials will happen shortly.

 

[sesevans]

Honestly the self driving truck is a huge liability. Who gets sued when they cause carnage? Who gets a prison sentence?

These are really good questions and their solution is part of the reason why self driving trucks are not universally available today.

The short answer is that these answers are being evolved through discussions between OEMs, insurance companies and regulators.

Keep in mind that your questions are also applicable with human drivers. It is a known fact that human truck drivers can cause "carnage", can be a "huge liability" and are subject to "prison sentences". Remember that there are 1.3 million deaths every year globally on roads today caused by traffic accidents and truck drivers are involved and at fault here too.

AVs have the potential to reduce this death toll substantially. A human truck driver may well be on a deadline to deliver goods and is subject to anxiety, tension, distraction, anger, road rage, drunk driving, etc. An AV robot is not subject to anxiety, tension, distraction, anger, road rage and drunk driving. I recall that when Google began testing AVs in 2009, they said one of the major reasons they were pursuing self driving vehicles was to solve an unmet human need. 1.3 million people die every year, mostly as a result of human error. 1.3 million deaths per year is the same as a Boeing 737 jetliner crashing and killing its occupants every day of the year. If you read these deaths as Boeing 737 jetliners would you believe society and regulators would allow Boeing to build commercial aircraft? Hence there is a real problem on the roads today with human drivers and a solution with the help of technology can be found.

As for AV trucks, note in the link I provided above regarding the Foxconn plant being built in Wisconsin, they are planning AV truck use carefully. Yes they are to be used on public roads but they are planning the roads and street lamps and sensors and LIDAR to be interconnected to achieve safety.

If you have a few minutes, listen to this recent TED talk about How Self Driving Cars work (same concept for trucks) and you may come away more positive about the chances of successful implementation.

https://www.youtube.com/watch?v=Ly92UcnoEMY

Comments?

 

[sesevans]

Apple is beginning to be more open about their plans for AVs. They have, for instance, begun to open source some of their research, which they are usually loathe to do.

https://www.wired.com/story/apple-executive-reveals-more-of-its-self-driving-technology/