Hacker News Reader

Comment by ChuckMcM | Hacker News Reader

Remember this is a company with billions of dollars in the bank, and collecting street imagery is a very parallelizable task.

Lets do some basic math. Lets say you pay a driver $10/hr to drive 8 hrs a day at an average speed of 35mph in the city. That is about 250 miles covered per driver, per day and $80. So you do that for 10 days and that is 2,500 miles for $800.

Now you build 10,000 cars with street view like capabilities, call it $40,000 per car, perhaps you get a car maker to help you perhaps not. that is 400M$ for cars, and 10,000 drivers is $8M for 10 days and 250 million miles. You haven't even spent close to a billion dollars and you've got street view like imagery for every major city on the planet in 10 days. Run this program for a year and you cover every road twice with your 10,000 drivers.

Its certainly doable by Apple if they choose to go that route.

View on HN

The sensor pile on top of google cars costs somewhere near $100 000, the car bolted under it is the cheap part.

Also, driving is the cheap part - structuring all the gathered data and images is a nontrivial task, you simply can't do it in a few months (as apple would need) if you don't have the know-how and software tools for this.

It would be kinda neat to have Apple and Google street view mobiles going around and pseudo-competing with each other.

Apple has the money to do it no question. However it takes more than just money. On the data collection side there's designing, building and integrating the sensor system on the car. That has to be tested. Then it needs to be replicated many times. The data collected needs to be stored somewhere. Either locally on the car or in the cloud. Either way that has to be figured out. You need to hire drivers, protect yourself from liability etc. While all of the these pieces are doable they're non-trivial. Especially when you consider the number of people involved. And that's just collecting the data.

Now Apple has to process the data and make that available to the maps app. And there needs to be software written for the maps mobile apps. Perhaps desktop too. All that needs to be tested. This "presentation" side of the problem involves a bunch of people too.

So how long do you think it takes a company the size of Apple to accomplish all that? I'd set a lower bound of one year. So for 12 months Google maps still has street view over Apple and will probably improve it (especially if they know Apple is creating a competing product).

While money isn't an issue the amount of time, energy and other resources required to make a product as refined as Google street view is enormous. Can Apple do it? It probably comes down to leadership. Steve probably could have done it. I wonder if the current leadership can.

>you've got street view like imagery for every major city on the planet in 10 days.

It took Google years to get Street View right, with massive amount of engineering talent and financial capital committed along the way.

12 months would be extremely challenging even for Apple. As for 10 days... well let's just say I'd be a bit unhappy, if ChuckMcM was my program manager :p

You overlook a couple of things, first is that Streetview has already been 'done' so there are a number of things which are more straight forward (do you recall how quickly Microsoft duplicated it relative to Google's effort?) and the second is that yes doing it in a year would be extreme.

That said, if you had been unemployed for over a year and I said, I've got a really simple job for you that will last at least a year, you drive this car up and down every street in the town at the legal limit. I'll pay you $400/week with benefits. Probably pretty appealing to you.

> Lets do some basic math. Lets say you pay a driver $10/hr to drive 8 hrs a day at an average speed of 35mph in the city. That is about 250 miles covered per driver, per day and $80. So you do that for 10 days and that is 2,500 miles for $800.

Don't forget fuel charges! ~$40 per day (at 3 USD per gallon for a car that gives 20 miles per gallon).

I think it'd be far quicker & cheaper to license street view data from Google, even if they start collecting their own in the background.

Google have spent many years perfecting the collection & analysis of the data, Maps needs fixing yesterday.

It will take considerable coordination to direct 10.000 drivers across the country or globe. Not sure if you can pull this off in a year, maybe it takes a year just to get a 1000 drivers on the road?

Your "basic math" fails on many levels.

1) "an average speed of 35mph in the city" is not going to happen. In Manhattan that's 15 MPH. In midtown Manhattan it's 1.7 MPH. So 10 MPH on average would be more realistic.

2) You forgot gas

3) You forgot tickets

4) You forgot car insurance

5) You forgot car maintenance

6) You forgot car accidents and the cost of replacement

7) You forgot tolls

8) You forgot lawsuit costs

9) You forgot parking costs (and time to drive from and to parking)

10) You forgot the costs of the hiring process (try hiring 10 drivers in 10 days)

11) You forgot the cost of routing of all these cars (you have to organize 10,000 cars somehow)

12) Your guess of $40,000/car for design, approval, production of a specialized car is way off.

13) You forgot server costs (stitching the images)

14) You forgot storage costs (storing, transferring) millions of photos.

15) You forgot IT and management costs to support all this shit. $8M will buy you a few executives in charge of this project.

16) Your estimate of $10/hr for the driver salary is way off (assuming it's just a driver, and not a trained technician). Average salary of a cab drive in NYC is $25/hr. Chicago - $17/hr. Los Angeles - $17.5/hr. That's not counting the benefits - health insurance, social security, etc.

Heh, part of the goal of 'simplfying' is so that you can reason about things. Most people don't really conceptualize what it means to have 100 billion dollars of 'cash on hand'.

We can make the numbers more accurate. Hell if Tim Cook hired me and gave me the budget for it we could knock this out in a couple of years. I'm sure there are at least half a dozen folks out there who could. The simple point remains though which is that you can solve this problem with money, and you can make the solution time arbitrarily short by scaling the money.

Apple could solve this problem with money by licensing Google's existing service. Actually building a competing service can't be solved with money alone. Just think of the number of people involved.

FWIW a street view 'type' car (differential GPS, 360 view camera, drive train encoding, wide band celluar/wifi scanner, and 100 TB of on board storage is significantly less than $40,000 these days. It is easier if you're just pulling ground data (no need to optically stabilize the cameras).

Yes, you listed (some of) the components. Building a working scalable system is something else though.

Oooh, wanna test me? Went through a pretty deep dive on the streetview stuff when I was working where you are but I'd be willing to bet you I could build you a 3rd gen car that would out perform the current ones for a lot less than $40K. I'd get dibs on Rod though, that man is hecka talented with a CNC mill.

Sure, Google, couldn't figure out how to make it cheaper than $100K per vehicle, but ChuckMcM on the internets sure can do it for "a lot less than $40K".

Reply on news.ycombinator.com