Hacker News Reader

Comment by falcolas | Hacker News Reader

Dear Car Companies:

If you would really like to advance the state of the art for self driving technology, I invite you come up to the mid-west, especially in the winter months, and make your toys work under those conditions.

Working up here will push your technology quite far, with snow, slush, worn away lane markers, cold temperatures, wind, potholes, washboard roads, ice, clogged sensors, blocked GPS, and any number of the above challenges thrown together in random and unpredictable patterns.

If you can get to level 4 or 5 automation in those conditions, making them work in the eternally sunny portions of CA will be a snap.

Sincerely, a Montanan (who is annoyed they had to deal with snow, ice, blocked sensors, and cold temperatures just this morning)

View on HN

Dear falcolas,

If you really need to understand why car companies will not take your advice, you should look at their sales numbers. Car companies don't sell many cars in Montana. They sell a lot of cars in urban concrete jungles and suburbs like those that exist in California and basically every major city. Solve one major city (especially one with hills like San Francisco), and you've likely solved most major cities, worldwide.

California has 14.5 million cars registered and 2 million new cars are sold in the state each year. There aren't even 14.5 million people in Montana. It's an incredibly easy market to "disrupt" as it were, and they have people in California that are known early adopters to lean on for buying, testing and validating their product. The people of countryside and similar locales are recalcitrant and will see products like these as extraneous, "taking their jobs" and "killing their freedom of driving."

So once they've tackled the low hanging fruit market of California, maybe they might consider expanding into the harsher wilds. But, similar to public transit, the infrastructure worth investing in is infrastructure most people will use, and unfortunately that is not a self-driving Ford F-150.

Signed, Common Business Sense.

Car companies sell plenty of cars in Chicago, New York, Toronto, Boston, Montreal, Minneapolis, Philadelphia, Detroit, Denver, Baltimore, and Pittsburgh.

Exactly! And Montana is a safe place to develop the tech in a lower density/risk environment.

It is not necessary for them to operate at full Montana conditions to have a successful initial business. The nice thing about machine learning is that it gets better and better as more data is collected.

So the rational path here is to launch in moderate temperature areas with beta software and eventually refine the tech to operate in adverse condition areas.

Pushing or expecting perfection from the beginning is how thousands of products have died before launching.

The best way to do product development is getting it out of the door as early as possible (in a reasonably safe state) with a competent subset of early adopter users and if possible in an idealized environment/use case. From there you expand and heavily capitalize marketing.

This is what Facebook did when they launched in Harvard and restricted it to only Ivy-league schools. Before slowly expanding to other universities and then the public at large. They called it the bowling pin strategy.

"If your project doesn't experiment first your entire project will be an experiment" - Joe Armstrong (creator of Erlang)

Most major cities in the world? All those major cities that don't get snow and icy roads? Major cities like Chicago, new York, Moscow? Any big city in northern Europe? What about Canada?

I think the snow question is a good one.

18-30 million people in Moscow... That's a big market!

Average salary in Moscow is $14000/year.

Never said they had to own the smart driving car. Lots of people riding public transit out there.

Facebook succeeded for years by starting with Ivy league schools before expanding to other US universities, then all western mainstream users, then expanded internationally.

Expecting them to get it right everywhere at launch time is a bad idea.

US cities are easy. Try Naples, Italy.

They can try Naples after they make a billion dollars in US cities. That's always how this stuff works. I've learned that the hard way living in Canada kilometres from the US but having totally different timelines for access to technology. Sometimes it's out of company ignorance but sometimes it makes sense from both a business and marketing perspective.

Do you really think the companies working on self-driving cars are unaware that there are portions of the world where inclement weather and poor road conditions exist? People keep bringing this up in every self-driving car thread as though it's some great insight.

yes, car companies know about snow. and i'm sure they'll solve it when they're ready to. Problems aren't typically tackled by solving every aspect at once, and the fact that nobody is demonstrating anything outside of perfect conditions in southern california should tell you a lot about the state of the technology.

AFAIK, there's a new huge testing facility for self-driving cars outside of Detroit for this exact reason. Will update with a link if I find it.

Edit: http://ns.umich.edu/new/multimedia/videos/23020-u-m-opens-mc...

And more recent investment: https://www.forbes.com/sites/dougnewcomb/2016/07/27/michigan...

You catch low hanging fruit first before even you think about palm trees. In other words, take care of most common use cases first before you run out of breadth chasing the extreme and difficult scenarios.

Yes! Come to Canada and try out your tech. California is so soft.

God it is so true.

There's such an amazing cultural comparison to be made between I-70 over Vail pass (in Colorado) and I-80 over Donner pass in California.

Vail pass rarely closes and is open for vehicle traffic in blizzard-like conditions - but you're on your own. You'd better be well prepared and be the competent driver of a well-maintained, capable 4WD vehicle.

And so, in a blizzard, Vail pass will be nearly empty and the occasional driver will be well-equipped for the trip.

In California, they enforce chain control on I-80 for all vehicles. This has the (amazing, stupefying) perverse outcome of allowing everyone to drive over Donner pass in a blizzard. Further, they are emboldened by the chains to drive quite poorly in conditions they have no idea of how to drive in.

The result is, during a full-blown blizzard, I was driving next to some moron in a RWD camaro trying to bob and weave through slower moving traffic. Dense traffic. Dense traffic largely made of other rear wheel drive cars operated by drivers with no snow driving experience.

California.

The reason they have to enforce chain control over Donner pass is because most Californians live in warmer climates and as such do not have snow tires. Very different to an average Colorado resident who swaps to snow tires during winter. What's the other option ? Just leave it open fully knowing that Bay Area drivers will be driving RWD, summer tires in deep snow ?

This comment gets made in every single self driving car thread. These companies are full of the smartest engineers in the world. They know conditions are different outside of California. But that doesn't mean they shouldn't be focused on California right now.

Did you see the Ohio transportation bill that just passed?

Kasich proudly mentions he signed it to create barriers to driving automation.

There are parts of the Midwest that are more inhospitable to driving automation than just the weather.

Reply on news.ycombinator.com