Wow, this reporter was horrible. He reminds me of those clients who when you show them the early stages of their site after a few weeks of work, point out everything that doesn't work as a huge the-sky-is-falling catastrophic event that is proof of the project's eventual demise. This reporter got to sit in a car that is doing things that a few years ago still seemed like science fiction and yet he had these niggling complaints:
For now, at least, the car only drives routes it's been trained to drive.
Google is pretty good at mapping things.
Since the Google car only just got its learner's permit, it drives accordingly
I think this is meant as a literal statement. The car is able to drive of its own accord, but needs an adult present who can take over if needed. But did you just say that the GOOGLE CAR HAS A LEARNER'S PERMIT?? That's amazing!
Then there was the jerking halt on a side street caused by a car that stopped a little abruptly almost two car lengths ahead.
If you don't think the software is going to err on the side of caution for YEARS after widespread usage, you are mistaken. Eventually we will be so used to trusting these cars that we will probably be napping on the way to work, so who cares that the car hits the brakes a few more times than it should?
And eventually I'm sure these cars will be a model of efficiency, with fast moving currents of cars zipping here and there. One step at a time.
Surprisingly, one thing the car can't do all on its own is use the turn signals.
Hey, how come when I click this button on my site does nothing happen??? THIS PROJECT IS A COMPLETE FAILURE!!!
If Google can get there before a major automaker beats them to it, I'll be really impressed.
The most preposterous statement of all. Please get the reporter on the line and let me place a bet of Google vs all the automakers combined on who will release this technology is first. Assuming that Google hasn't locked up all of the pertinent technologies, this wouldn't be a fair fight.
I'm really disappointed to see this as the top voted reply. Of course the reporter is wrong, of course his article can be ripped apart, of course this is a momentous achievement and the reporter is just too ignorant to know that, and we're all clever enough to see how amazing this is.
The reporter is right, and we're wrong. The question the article is answering is "Are Google's self-driving cars ready?" The article explains in general-interest terms why they're not.
You or I might be tempted to buy that exact car today as a beta, knowing that we'd have to tweak a lot of things, but (a) I don't want to drive in a world where your beta car with your rooted ROMs is driving around (and you definitely don't want to drive on the same streets as my inexperly-rooted ROMS, trust me); (b) the grandparent comment is underestimating the work involved in micro-mapping all of the possible routes, let alone the vast array of possible interactions between other vehicles, humans, animals, new road hazards, weather, etc. all mixed together.
I'm sure it's going to be a few years. Maybe more than a few. But that's okay with me. The stakes are pretty high for two-ton hunks of metal moving at 60 mph. This is something we all definitely want them to get as close to correct as humanly possible before it's available to consumers.
On a slightly meta note, HN comments are weighted by the average karma score of the person leaving the comment, not just the number of up-votes that specific comment received.
So this comment being at the top could easily be down to the poster's karma/average karma.
I'm not convinced that's true, and would be interested to know your evidence. In particular, the commenter in question has no average karma, apparently not having made sufficiently recent previous contributions. Looking at the contributions made, they don't look to be the sort of thing that garners lots and lots of karma, but I only really glanced at them, so that impression could be wrong. But I doubt it.
It's impossible to tell without comment scores being displayed, but my feeling is that any ranking influence from average karma must be small, if any. I'd be interested in any concrete information, even if circumstantial.
If, however, I find that average karma does have an effect, I'm going to stop replying to individuals, stop providing information, stop correcting misapprehensions, and concentrate on only submitting populist items. That's clearly what PG would want, if that's how the ranking systems work.
FTA: For now, at least, the car only drives routes it's been trained to drive. My ride in Washington DC was along a route that Google engineers had driven with the car earlier. Google refused to allow the car to be driven anywhere beyond this well-studied environment, at least not with the media tagging along.
Reporter, did you even ask the Google engineers about that? For a media demo, it is not at all surprising that they would use an already driven route. The reporting here was a bit disappointing.
Reporter, did you even ask the Google engineers about that? For a media demo, it is not at all surprising that they would use an already driven route. The reporting here was a bit disappointing.
Actually, no, this limitation has been well documented (but cleverly brushed aside). The car doesn't see and understand the road ahead, it follows an invisible line on the ground, and can stop if it detects danger. It doesn't read signs.
So yes, demos are impressive, but slightly misleading.
I'm pretty sure there's no "invisible line on the ground". From what I've read, the route simply needs to be mapped at a higher resolution than most roads have been mapped before.
I agree 100%. My commute is over the same route, every day, some 260 days a year. The value of freeing me up to do something else for that hour each day can't be understated.
I live in NYC and the best use-case for this is taxis. Taxi do not take the same route twice. Traffic is a nightmare during different seasonal, temporal, or accident events. So like a Garmin device it should be able to "remap" the trip.
In our lifetimes this will be possible but maybe not in next 1-3 years. (I hope I am wrong)
Google has covered large parts of the (populated) world with Street View. Pictures [1] indicate that the cars were equipped with laser sensors. So Google probably has enough data to have the cars drive new routes.
But thats not the point. Most of peoples daily driving routine is the same routes over and over, commuting to and from work.
Going even further: I'd imagine that you could train the car on a specific route by driving it manually. Such information could be crowdsourced, much as Google already does with its traffic data on Maps.
I believed that as well, but now I'm not so sure. The density of data gathered by the Velodyne on the self driving cars is far higher than the SICK lasers in that picture. Also, the scene may have changed in the interim.
Yes, but we're still talking cars that measure metres by metres, and streets with lots of extra room. As you mention, the scene may have changed - so trying to preplan the whole route to the inch will only hurt you. What is needed is basic data above GPS, and I think the Street View lasers can be that.
I'm sorry I was unclear - I meant that the scene could have changed since the Street View team was last through.
The Google Maps data is surprisingly bad - we saw several presentations about the Car, and they highlighted how awful it really is.
I don't think that they're preplanning. I think that they're analyzing the scene for differences to simplify and speed calculation. Having good a priori data is really powerful.
You might be surprised. It's a bit more than having to just program in new GPS coordinates. Imagine the car drives the route once collecting tons of data with the laser range finder and other sensors on the car. Now they let the car sit for a few days while they process all the data. Finally the car can drive the route. There's still some pretty hard challenges to solve in making those offline algorithms realtime. Maybe some of it is just waiting a couple years for faster processors, but I'm guessing they might have some pretty intense engineering challenges ahead.
I don't think he meant doing it was an easy task, but rather that a large majority of the time when you take your car you go on a route you've already taken in the past, and if the car can self train its routes when in manual mode then you only need to drive it once / the first few times to go to work/groceries/school/...
So many of the comments here are pointing out the flaws in the reporter's account and, quite rightly, scoffing at their ignorance.
But there's a lesson to learn: he's an ordinary person, and this is what he thinks about the demo!
So don't just dismiss it as completely missing the point. Learn and understand that this is how the normals think. This is an important lesson for we who write code.
That is such a software development moment :) Probably too easy to implement, that's why they forgot it.
But the fact that the car often needs the human to intervene is a bit more interesting. Why would a human better know if there is enough space for two cars to pass each other in a narrow street?
If I was google I'd be erring on the side of extreme caution. The first public accident that happens with one of these cars is going to get plastered over every media outlet in the world.
That they had the car in "super safe driving" mode for a media event seems entirely reasonable, even if the car probably could've navigated the narrow street with 95% confidence. The obvious path of attack is targeting the safety of an automated driver. Lose voter confidence and the whole thing comes crumbling down to be delayed another decade of legislation.
Turn signals are a small subset of the "signal knowledge and intent" lights/outputs that a self-driving car might want to use to communicate with the humans and, some day, other self-driving cars around. The team might be working in the direction of a more general implementation of such things rather than hooking up a temporary system for just two sets of dumb on-or-off lights.
Probably depends on the accuracy of positioning and real time 3d modeling of the environment. I wouldn't be surprised if it wasn't down to the centimeter level yet.
The system likely has a bubble of safety around it to allow for sudden changes close to the car. This makes the "car" far larger than it would actually be.
The lack of turn signals makes me highly suspicious. There is more to this omission than meets the eye, I am quite sure of it. As you say, adding turn signals in should be straightforward. The computer presumably has its turns planned far enough in advance. So why doesn't it?
I am actually even less inclined to trust these so-called "self"-driving cars now. Not like I even trust the human-driven variety all that much...
They've made a self-driving car. The merely annoying parts are going to be the fun bits! Everything else is going to be difficult, if not bordering-on-impossible. I find it very difficult to believe that there is any technical barrier to computer-operated indicators.
Wouldn't it be sufficient to wire in a computer controlled switch in parallel to the mechanical switch that the human uses to control the turn signals?
Or just add a new set of turn signals independent of the built-in ones.
Or it could just be that the turn signals aren't wired in to the computer system. On my vehicle, the wiring for the turn signals is pretty basic, there is no computer involved.
FTA: "But the biggest step will be to create a car that will let me just sit in the back seat with no-one at all in the driver's seat. . . . If Google can get there before a major automaker beats them to it, I'll be really impressed."
A major automaker has as much chance of beating Google as a 350lb man has of beating Usain Bolt at the 100m dash. The automakers aren't really known for doing cutting-edge software, you know.
That part made me grin. Most non-technical people I've talked to have expressed surprise that Google is the company furthest along with this technology. They see the problem as one of vehicular design rather than large-scale data analysis. Automakers are pretty good at the former, but nobody has proved themselves to be better at the latter than Google.
No major automaker will produce self-driving cars because self-driving cars greatly reduce the need for private car ownership. A two-car family might be able to manage with a single car if that car can drive itself to pick up individuals. Groups could pool their money to buy a single car the same way that fractional ownership of jets happens today but on a mass scale. The number of cars parked unused could go down substantially. Self-driving cars are a huge benefit for society and cities especially but a disaster for car companies.
They may be a disaster for the car companies, but they'd be a huge success for the only car company to have one, if only one car company had one. Thus is how competition can drive adopting new technologies even when they make their industry less profitable.
Today's driving scenarios will be served by a small pocket of tomorrow's automated power, granted.
If we have self driving cars, fast switching and routing, we may find ourselves in possession of a packet switched human transportation network. I think we'd need a lot of cars in that world. They might become commoditized like the PC was, but still a profitable business.
Absolutely. Google must realize that its supremacy in search/ads will eventually come to an end, so they're making a really smart, long-term investment here.
Self-driving cars would be awesomely useful, and Google has exactly the right resources (top AI engineers) to make this happen.
One day, we'll all likely have cars with AI powered by Google.
Or you could view it that they are trying to extract an extra hour or two of eyeball time per person per day. Since you no longer have to drive the car, you could be using your android tablet on the commute to work....
That's a good point. It's a way to make more money from the users they already have. It's certainly in line with their current ad-based business model, but man, talk about thinking outside the box...
I guess if they get Google AI in your car, it will probably come with lots of other Google products, too.
Google AI: Elsurudo, we are about to go past Joe's coffee and they are offering a free cookie with any large drink. Traffic is good today so we have time if you would like me to stop there?
Taking click through adverting to the next level, delivery of the actual customer.
> The data from each situation would be ingested and analyzed so the car could learn what to do in the future. Those lessons could, hopefully, be applied to a broad range of driving conundrums.
This is huge. Maybe a handful of cars are learning there way around slowly, but if they have legitimate automatic learning systems, imagine two or three thousand cars all learning together.
This is a major piece in the overall puzzle that the article doesn't really explore. Google's greatest strategic advantage is networked systems - each additional driverless car is a mobile, sensor-packed learning system for all other driverless cars. I strongly doubt that the plan is for each car to 'learn' how to handle each new driving situation independently - as soon as one car learns a behavior, they all learn that behavior.
That level of networked learning system is probably not fully mature, but it's certainly not far from reality.
This won't be the last critical review that Google gets. History has shown that people are incredibly quick to dismiss and attack things they don't understand.
A good (recent!) tech example would be the iPhone: back in 2008; it's kind of funny now, but my first impression of it was "huh, where's the stylus?" and I wasn't alone. The amount of "No stylus, so it's hard to use" FUD being spread around was incredible.
Now my two year-old son tries to swipe on my computer dumbscreen to go to the next photo. The "People of the land.... morons" opinion curve is probably going to follow the same trajectory; they will dismiss and criticize until they can't remember how crappy driving a car really was.
Can the lack of automatic turn signal be due to some legal implications? Like for example, the car falling in the same position as other car with assisted driving technologies.
Or perhaps it's just a quick way to change the car's route?
For now, at least, the car only drives routes it's been trained to drive.
Google is pretty good at mapping things.
Since the Google car only just got its learner's permit, it drives accordingly
I think this is meant as a literal statement. The car is able to drive of its own accord, but needs an adult present who can take over if needed. But did you just say that the GOOGLE CAR HAS A LEARNER'S PERMIT?? That's amazing!
Then there was the jerking halt on a side street caused by a car that stopped a little abruptly almost two car lengths ahead.
If you don't think the software is going to err on the side of caution for YEARS after widespread usage, you are mistaken. Eventually we will be so used to trusting these cars that we will probably be napping on the way to work, so who cares that the car hits the brakes a few more times than it should?
And eventually I'm sure these cars will be a model of efficiency, with fast moving currents of cars zipping here and there. One step at a time.
Surprisingly, one thing the car can't do all on its own is use the turn signals.
Hey, how come when I click this button on my site does nothing happen??? THIS PROJECT IS A COMPLETE FAILURE!!!
If Google can get there before a major automaker beats them to it, I'll be really impressed.
The most preposterous statement of all. Please get the reporter on the line and let me place a bet of Google vs all the automakers combined on who will release this technology is first. Assuming that Google hasn't locked up all of the pertinent technologies, this wouldn't be a fair fight.