The first person to hack the iPhone is building a self-driving car • Hacker News Board

Re: The first person to hack the iPhone is building a self-driving car

#401

Post by LoSboccacc » Thu, Dec 17, 2015, 12:52 AM UTC

argonaut wrote: post #303

jpfr wrote: post #37

Prototypical case of the 80/20 rule. He has implemented the happy case. But that system is nothing people realistically would want to drive their cars. What he did is impressive. But the results are not that outlandish for a talented person. 1) Hook up a computer to the CAN-Bus network of the car [1] and attach a bunch of sensor peripherals. 2) Drive around for some time and record everything to disk. 3) Implement so…

Why is deep learning this magic pixie dust you sprinkle on anything and it works? Have the people who are suggesting this actually gotten deep reinforcement learning to work on complex, long-time-frame, real-world continuous control problems before? I would be very surprised if you got deep reinforcement learning to perform well on a self-driving task, even on a highway. If you did, well, your faculty position at Sta…

they're very powerful calssificators and from the outside it seems they can learn to distinguish between arbitrarily complex states.

more states? add neurons! more search space? add layers!

except they have unpredictable resonances, especially multi-layer networks:

http://www.i-programmer.info/news/105-artificial-intelligenc...

they're just starting to understand this, but I believe the myth of the 'do it all dnn' is gonna die. it's time to start thinking about cluster of independent neural network, supervising each an independent aspect of the search space and/or each other.

Re: The first person to hack the iPhone is building a self-driving car

#402

Post by grahamannett » Thu, Dec 17, 2015, 12:58 AM UTC

Daneel_ wrote: post #391

myztic wrote: post #388

(This is completely Off-Topic but it's been bothering me for such a long time now and I never got sufficient answers) Why am I seeing Ubuntu on Screens of developers, experts, et cetera in Cover Stories such as these, most of the time with the 100% plain Ubuntu Desktop with all the craziness that comes with it? It feels like this is the case 90% of the time. Two more (recent) examples I can remember: 1) Fyodor (Guy b…

Most of these people have passed the stage of caring about the desktop itself. I would say they are simply interested in a bash shell that sits on a reliable, well maintained, stable distro with good community knowledge. When you spend all your time at a prompt you don't really care about unity vs cinnamon vs anything else. Debian is stable and has fantastic package management, and Ubuntu builds on that with a vibran…

curious what myztic would recommend as well? ubuntu may not be 'ideal' but it works and installs a lot easier than arguably any other linux distro. i have no problem using debian/arch/etc but if you are just trying to get up and going ubuntu seems to always be the easiest.

Re: The first person to hack the iPhone is building a self-driving car

#403

Post by monksy » Thu, Dec 17, 2015, 1:00 AM UTC

jpfr wrote: post #37

Prototypical case of the 80/20 rule. He has implemented the happy case. But that system is nothing people realistically would want to drive their cars. What he did is impressive. But the results are not that outlandish for a talented person. 1) Hook up a computer to the CAN-Bus network of the car [1] and attach a bunch of sensor peripherals. 2) Drive around for some time and record everything to disk. 3) Implement so…

I would argue that you're half there. You'd have a car that could navigate roads, but ultimately it couldn't get you to where you want to go.

Here has been dedicated on making maps at a quality level that is needed for automous cars. A few issues with the data that is available currently is that the data isn't very detailed, and you're at the mercy of volunteers (tiger(old), openmaps data), or for a company that's main focus isn't maps. (Google)

http://360.here.com/2015/06/02/take-create-autonomous-cars/

Re: The first person to hack the iPhone is building a self-driving car

#404

Post by LoSboccacc » Thu, Dec 17, 2015, 1:08 AM UTC

darkmighty wrote: post #265

1024core wrote: post #205

Like most hard problems, it's easy to pick off the low-hanging fruit and claim that you have solution. Self-driving cars (in some form or the other, under some loose definition of "self" and "driving") have been around since the 20s. But it still remains a vexing problem. It is quite easy to program a car to stay between 2 cars and follow the car in front. It is quite another to have the same car drive on (a) a road…

TBH, since it's a training based system it's "just" a matter of making sure the training set is large enough, including the situations you mentioned (assuming the training method is robust, generalizes well, etc). I would love someone knowledgeable to give an estimate, but I would guess you need at least a handful (10+?1000+?) of examples of each edge case (involving bicycles, pedestrians, weird road designs, street…

what if a driver does a mistake, like running a red, and doesn't get a ticket?

who validates all this data?

attaching a dnn to a driver as a training set is a pipe dream, for now. maybe after we understand how our brain perceives time and build models of future outcomes, we could apply it to build better nn. For now, nn are just best used as classifiers in a controlled environment, not from an environment with unpredictable states.

and especially not in an environment with adversaries http://spectrum.ieee.org/cars-that-think/transportation/self...

Re: The first person to hack the iPhone is building a self-driving car

#405

Post by austenallred » Thu, Dec 17, 2015, 1:18 AM UTC

jpfr wrote: post #37

Prototypical case of the 80/20 rule. He has implemented the happy case. But that system is nothing people realistically would want to drive their cars. What he did is impressive. But the results are not that outlandish for a talented person. 1) Hook up a computer to the CAN-Bus network of the car [1] and attach a bunch of sensor peripherals. 2) Drive around for some time and record everything to disk. 3) Implement so…

> Guy builds a fucking self-driving car. By himself.

> Not that outlandish for a talented person

What planet are you living on? I don't know what you did today, but I played with some jquery animations. This guy drove around in a self driving car that he built himself. It doesn't solve for edge cases? Neither do 90% of CRUD apps. Holy shit.

Give some credit where credit is due. This is not an ordinary or average outcome.

Re: The first person to hack the iPhone is building a self-driving car

#406

Post by grahamannett » Thu, Dec 17, 2015, 1:19 AM UTC

robotresearcher wrote: post #204

username3 wrote: post #56

Earlier quoted context omitted.

Typical twenty year olds don't read all the state-of-the-art papers on a subject before saying they know it all. It sounds more like he's caught up on the latest AI research and fundamentals.

People in their twenties wrote those papers.

The math in any of the paper's he's most likely referring to aren't some theoretical pde maths or abstract algebraic geometry stuff... it's pretty understandable if you can grasp a "graduate level" linear algebra course.

Re: The first person to hack the iPhone is building a self-driving car

#407

Post by mandeepj » Thu, Dec 17, 2015, 1:20 AM UTC

> In the coming weeks, Hotz intends to start driving for Uber so he can rack up a lot of training miles for the car.

Really? I did not expected this from him. Why don't he put his sensors\cameras\kit on few other hundred\thousand cars and pay them some money or get some early adopters.

Re: The first person to hack the iPhone is building a self-driving car

#408

Post by darkmighty » Thu, Dec 17, 2015, 1:29 AM UTC

LoSboccacc wrote: post #404

darkmighty wrote: post #265

Earlier quoted context omitted.

TBH, since it's a training based system it's "just" a matter of making sure the training set is large enough, including the situations you mentioned (assuming the training method is robust, generalizes well, etc). I would love someone knowledgeable to give an estimate, but I would guess you need at least a handful (10+?1000+?) of examples of each edge case (involving bicycles, pedestrians, weird road designs, street…

what if a driver does a mistake, like running a red, and doesn't get a ticket? who validates all this data? attaching a dnn to a driver as a training set is a pipe dream, for now. maybe after we understand how our brain perceives time and build models of future outcomes, we could apply it to build better nn. For now, nn are just best used as classifiers in a controlled environment, not from an environment with unpred…

The vulnerability to sensor error (adversarial or not) is certainly not exclusive to nn based approaches. I commented on the validation problem in the comment above and in another below, and one way to deal with it is simply manual validation (mainly for false positive elimination). Indeed this approach with dnns is already being employed by Mobileye, so I don't think it's a pipe dream.

Sensor failure or well characterized adversarial inputs are actually really easy to deal with -- they are very easy to simulate with a given dataset and self-validate using traditional techniques -- simply make one or more cameras fail (or receive spurious sigs) and verify the output.

It's a good point that probably all autonomous cars will need a contingency plan (probably human intervention and/or blind emergency stops) with non-zero probability -- even if you have a redundant network of cameras around your vehicle a critical number can and will occasionally fail (when you look at the fleet sizes that will be dealt with).

Re: The first person to hack the iPhone is building a self-driving car

#409

Post by unclebucknasty » Thu, Dec 17, 2015, 1:38 AM UTC

tptacek wrote: post #58

bcohn91 wrote: post #46

Earlier quoted context omitted.

Why is the first comment on HN minimizing this truly impressive project? Of course it's not perfect, he's ONE person.

Putting a prototype self-driving car on actual roads without understanding the difficulty of that project seems like a legitimate, substantive criticism, and I don't care how many people are involved in the project.

Now, that's a very valid criticism. I don't care about his personality, but testing the car on live roads is asinine. And, this journalist jumps in and excitedly plays up how he was afraid for his life, etc.

Yeah, instead of going after the sensationalism, how about you discourage him from endangering the lives of others who weren't given the opportunity to make such a stupid decision?

Re: The first person to hack the iPhone is building a self-driving car

#410

Post by nojvek » Thu, Dec 17, 2015, 1:43 AM UTC

I really like his 4 stage definition of self driving car. I don't really care about the fully autonomous like the google car. I've driven the adaptive cruise control VW in europe and that was an amazing experience. The only thing missing was lane control which this guy has done. Personally, where self driving really shines is long trips on the highway. All I really want is smarter cruise control that can stay on one lane and not bump into anything, and ideally send an alarm if it thinks it needs help.