[SDP]

This sounds too good to be true, but Comma.ai just announced that they will ship $999 autonomous driving kits by the end of this year.



The kit seems to be iPad-mini sized and the CEO claimed that you will be able to install it yourself.

Here's the video link for more details:

https://www.youtube.com/watch?v=AerjS7PTNYs

Source:
https://techcrunch.com/2016/09/13/co...-of-this-year/

[The Rationalist]

Here is a short video of his self driving car. The video is from Bloomberg business.

https://youtu.be/KTrgRYa2wbI

[AbhiJ]

This has lawsuit written all over it.

I would not trust big OEMS with this technology right now.. leave alone the guy who jailbroke the iphone.

[ecenandu]

Originally Posted by SDP

This sounds too good to be true, but Comma.ai just announced that they will ship $999 autonomous driving kits by the end of this year.

Come on guys, will it be reliable? Who wants to put their life on the line with such systems?

More than the hardware, it is the software that makes the difference. That is why they OEM systems are costly, they have to test and validate the system before making it available for the public.

As a prototype, this product is fine. But to bring such a product in to the market, you need a big team and loads of money.

[SDP]

Originally Posted by AbhiJ

...

I would not trust big OEMS with this technology right now.. leave alone the guy who jailbroke the iphone.

Originally Posted by ecenandu

Come on guys, will it be reliable? Who wants to put their life on the line with such systems?

More than the hardware, it is the software that makes the difference. That is why they OEM systems are costly, they have to test and validate the system before making it available for the public.

As a prototype, this product is fine. But to bring such a product in to the market, you need a big team and loads of money.

Why the skepticism? Did you guys actually see the video/presentation?
The company believes that they are ready to ship the product in 4 months. Why do you think its a prototype?
The CEO actually agrees that their product is not in the league of Google Driverless car or Tesla AutoPilot. Its definitely not targeting the "conventional human driver sitting in the back-seat working on his laptop or chatting on his smart-phone" use-case. More like "You sit in the driver's seat and the system takes care of 90-95% of your driver-job".

Think about it. AI can predict even weather fairly accurately today and that depends on thousands of variables. Would it be really impossible to automate what a human driver does? All the hardware to operate and control the car already exists in the car and with "drive-by-wire" all of those can be controlled by electronic signals. All you need are - eyes & brain equivalents. Comma.ai seems to have taken a different approach towards the 'brain' (the software) part of it. They have tapped into real driving data of n number of actual drivers (data available to them through dash-cam type apps they launched much earlier) and have allowed the brain to "learn" how to drive.

I have my own doubts about this product/technology as expressed in the opening line, but let's just keep the mind open and explore what has been done rather than dismissing it as fraud/prototype. Only a moron would be ready to risk his reputation and career by standing on a stage with a fluorescent green colored box and proclaiming that its ready to be commercially shipped in 3-4 months, when he knows that its nowhere near production ready.

[ecenandu]

Originally Posted by SDP

Why the skepticism? Did you guys actually see the video/presentation?

I have my own doubts about this product/technology as expressed in the opening line, but let's just keep the mind open and explore what has been done rather than dismissing it as fraud/prototype.

Yes, I saw the video.

Fraud/prototype, buddy, prototype isn't bad, in fact it is the first step in the right direction.

I think as a software company, they may have its strengths. But I am skeptical of those DIY kits to be reliable. As a product, this has got little less value because the market for these will be minimal. As a technology, maybe he can sell his product, software, to a car manufacturer and make money.

Watch this video from Nvidia,

https://www.youtube.com/watch?v=HJ58dbd5g8g

[ecenandu]

An article related to the topic.

Uber’s Pittsburgh Project

Second article

The article by Ashlee Vance did not correctly represent Tesla or MobilEye. We think it is extremely unlikely that a single person or even a small company that lacks extensive engineering validation capability will be able to produce an autonomous driving system that can be deployed to production vehicles. It may work as a limited demo on a known stretch of road -- Tesla had such a system two years ago -- but then requires enormous resources to debug over millions of miles of widely differing roads.

This is the true problem of autonomy: getting a machine learning system to be 99% correct is relatively easy, but getting it to be 99.9999% correct, which is where it ultimately needs to be, is vastly more difficult. One can see this with the annual machine vision competitions, where the computer will properly identify something as a dog more than 99% of the time, but might occasionally call it a potted plant. Making such mistakes at 70 mph would be highly problematic.

Source

Lets see, we will find the answer, soon.

I am more than happy to be proved wrong.

Originally Posted by The Rationalist

Here is a short video of his self driving car. The video is from Bloomberg business.

Article from Bloomberg

[AbhiJ]

Originally Posted by SDP

Why the skepticism? Did you guys actually see the video/presentation?

I am a skeptic for the entire self driving car phenomena, not just this guy.
For the Comma AI, the primary problem is that the guy says watch the car like a student driver is driving. To me, thats actually more strenuous than the act of driving.

In real life, with all the billions of variables, I dont trust 100% computers to drive and possibly in the event of an accident scenario, save the life of my family.

I know .. I know.. ABS, traction control already do most of the work, but the decision making power still remains with me. I like it that way. Call me old fashioned, I am ok about it.

I know that computers fly planes and run trains. But both in cases of planes and trains, the idiot cyclist wont be cutting them off anytime soon.

[ecenandu]

https://www.youtube.com/watch?v=iCjNBGDeQJ4

https://www.youtube.com/watch?v=XzKnk2S90tw

[ecenandu]

One year down, this is what I was talking about earlier. He knows his limitations and is playing to his strengths.

https://youtu.be/aqdYbwY9vPU

[ecenandu]

https://www.youtube.com/watch?v=-b041NXGPZ8&t=3670s

Here is an in-depth look at the latest Tesla AutoPilot hardware.

I don't think anyone can beat Tesla when it comes to autopilot, they got the head start and their neural network is pretty good and is getting better every day.

The fascinating thing is their shadow mode, which is always running even if you are not using the autopilot. So, you could be training the neural network every time you go out for a spin in your Tesla. This is a great advantage for Tesla, no other manufacturer has got this much amount of data available to them, no amount of in house testing can replicate the real world data.

The computer vision has become a lot better in the last five years, especially image classification(using convoluted neural networks) and object detection(like YOLO) which will enable us to ditch expensive sensors like LiDAR, which anyway gives us little information.

It is quite interesting to see different approaches used by companies in solving the self-driving problem. I think Comma One and Tesla are using a similar approach to solve the self-driving problem, i.e, using neural networks.

[ecenandu]

An insight into Uber self-driving cars software architecture.

https://www.youtube.com/watch?v=jTio_MPQRYc&t=2365s

[ecenandu]

They have released Comma 3, the first ground-up hardware from the team, the previous generations relied on mobile phone hardware.

I really enjoyed watching the hour-long presentation, the answers are brutally honest and thought-provoking.

Please do watch the whole video if you do get the time.

https://www.youtube.com/watch?v=hbLiehrC2DQ