Uber’s First Self-Driving Fleet Arrives in Pittsburgh This Month • Hacker News Board

Re: Uber’s First Self-Driving Fleet Arrives in Pittsburgh This Month

#241

Post by msh » Sat, Aug 20, 2016, 7:03 AM UTC

gambiting wrote: post #116

tcoppi wrote: post #97

Earlier quoted context omitted.

Have you used Google Voice in the past ~year? I have a standard american accent so I might be the best case scenario for it but I haven't seen it miss a word I've said anytime I can remember in a year or two. I mostly use it in the car to reply and send texts, works even with the radio and ventilation on. I've used an Amazon Echo and that is right 99% of the time too. Now, interpreting the words I say is a different…

Well, the thing is, English is not the only language in the world. Google's voice recognition in my native language, Polish, is absolutely abysmal.

Well, my language (Danish) is pretty small and I would say Google voice recognition is notably worse than English. Sometimes it's way of pronouncing names is a bit weird but it is recognized when I said them.

So it might not work for everyone in the world but at some point soon it will properly be there for most national languages.

Re: Uber’s First Self-Driving Fleet Arrives in Pittsburgh This Month

#242

Post by jayjay71 » Sat, Aug 20, 2016, 8:17 AM UTC

argonaut wrote: post #240

jayjay71 wrote: post #235

Earlier quoted context omitted.

I used to work on this - the approach is very different. That's a pretty hollow statement on the internet though, so take it as you will. An interesting aside - it seems like most of Google's team left this year. Chris Urmson left, Anthony Levandowski left, and many of their engineers are gone as well (many went to Otto). It'll be interesting to see what happens now that they've lost all of the original leadership (S…

What is the approach, then?

At a very high level, instead of comparing everything against a map, we try to figure out what it is. For example, the way Google detects traffic lights is using geometry to infer where a traffic light will be. Their maps are so good that they know where the car is, and where the traffic light is, and just use math to figure out where in the image there should be a traffic light. Then they just look at the pixels and figure out which color is lit - red, yellow, green. The way we did it is constantly scan for traffic lights (you can do lots of sanity checks, for instance using GPS are we near an intersection?), which means figuring out whether or not a traffic light is in the image, and then once we detect one figure out what color is the light.

Google's approach is simpler but is highly dependent on the maps. The other approach is easier to scale, but is obviously harder and likely takes more work to get the same results. Something I think a lot of people gloss over is that these maps take up huge amounts of data - not something you can stream over a cell network and probably not even high speed WiFi (maybe you'd download a map for your trip the night before). And also, as I understand it, the maps require regular maintenance.

The real question is how will this scale? And honestly I don't know. My theory is the tech is so hard and the rewards are so great that infrastructure will change to make it easier. By installing smart traffic lights, having cars talk to each other, and adding various road marks specifically for self-driving cars you can drastically reduce the complexity of the problem.

Re: Uber’s First Self-Driving Fleet Arrives in Pittsburgh This Month

#243

Post by user5994461 » Sat, Aug 20, 2016, 3:45 PM UTC

fiveoak wrote: post #210

user5994461 wrote: post #203

Earlier quoted context omitted.

I wouldn't be surprised if driverless cars were way more secure for pedestrians and cyclists. It's really difficult to keep track of pedestrians and cyclists when driving, whereas a driverless car has 360° nightvision.

Eventually, it might be. However just because the car has better sensors doesn't mean it will react better than a human would with less information.

> a human would with less information.

I'm arguing for the case where the human as NO information.

Let's say, there's a bicycle/motorbike somewhat behind your cars.

For a human driver, it can be seen in only one of the 3 rear mirrors at any point in time. Unless it's slightly behind-left or behind-right (e.g. to take over) in which case it's in the dead angle and can't be seen at all.

By turning his head and using the mirrors, a human cannot see most of the 360° around a car. There are obstructed angles, not to mention that the human can only look toward one direction at once.

The self driving cars with 360° vision can see things that a human couldn't. It has a chance to react where the human didn't.

Re: Uber’s First Self-Driving Fleet Arrives in Pittsburgh This Month

#244

Post by puranjay » Sun, Aug 21, 2016, 12:26 PM UTC

kalleboo wrote: post #63

puranjay wrote: post #31

Earlier quoted context omitted.

How much has the recent interest and investment in self-driving cars impacted this estimate? 5 years ago, I would have said that level 4 automation is 20 years away. With all these billions pouring in, I'd cut that projection in less than half. I'm certain that I won't be driving a car ever in 2030

How many billions have been poured into speech recognition? I don't anyone here would trust controlling a car with Siri or Google Voice lest it misinterpret something you said.

You'll be surprised to know how popular speech recognition is among younger people.

Amazon Echo is far from perfect, but it's the perfect example of the rapid improvements in technology + building an interface that "just works"