Hacker News Reader

I have a feeling truly driverless operation will be slower than we think.

I seriously believe it'll be here in no more than 3 years, maybe 5 if there are regulatory hurdles. There's just too much value to capture for everyone to be half-assing it.

EDIT: Citations

"Google Self-Driving Car Project Monthly Report | December 2015" https://static.googleusercontent.com/media/www.google.com/en...

"We’re currently averaging 10,000-15,000 autonomous miles per week on public streets"

There are ~100K Model S vehicles. Somewhere around VIN 55K is when autopilot hardware was introduced (http://www.teslamotorsclub.com/showthread.php/44787-Autopilo...). Let's conservatively say that there are 40K Model S vehicles out there with self-driving hardware that are refining their data, and consequently, algorithms Tesla is updating. 40K vehicles * 12K miles/year/vehicle (average) = 40,000,000 miles/month (!!!) in "autonomous experience" Tesla is gaining. (Fun fact: The average human drives ~500K miles in their lifetime)

"According to Fred Lambert from Electrek, Musk referred to the first round of Model S owners with Autopilot as "expert trainers" that will collect and deliver tons of important data for Tesla's intelligence network. Musk said the system should improve each week as the driving algorithm constantly updates, adding Model S drivers could add around 1 million miles of new data every day."

http://www.techinsider.io/elon-musk-explains-tesla-autopilot...

That's a lot of progress being made each month.

Never underestimate the volume of regulatory hurdles. I'd bet on decades of needless delay. Half of it caused by the first time people hear on the news about a self-driving car in an accident and ignore all the statistics showing they're actually far safer.

I think regulatory hurdles will be introduced by politicians not due to safety concerns, but based on jobs. The conversation will switch from "immigrants are stealing our jobs!" to "robots are stealing our jobs!"

Two can play at the propaganda game. "Human drivers cause X% more deaths on the road; can your family risk having humans at the wheel?"

Technology companies have been remarkably bad at playing the propaganda game. We, as an industry, have done such a bad job of security that trusting the computer to drive you is going to be a powerful problem to overcome. "You've replaced your credit card because of a computer security failure, now think about that when getting into your computer controlled car. Can your family risk having a hacker at the wheel?"

Except for people who have no choice (the elderly, whom have no one else to transport them and can no longer drive themselves, or people who have gotten DUIs ironically enough).

Well, we do elderly transport in our neck of the woods (not sure about the urban folks). People currently sometimes have no options, but those groups probably won't overcome the regulatory wall that will show up. Plus, the truck driver is one of the most common jobs in almost every state, and it is a job that cannot be outsourced. Its going to be a rather large hit to the economy. I would imagine which side a politician would come down on.

I will think it might become an issue when I see a Google Car in North Dakota in the winter. Until then, I'm not sure the technology for a national roll-out is really there.

Self-driving trucks are also a thing

I know that. That's why I mentioned it.

One probalem is that driverless cars don't have to be better than humans. They have to be better than human drivers with computer assist. Or at least cheap enough to jsutify the extra carnage.

Truck driver is most common job in a goodly number of US states, so its going to be disruptive of a job that is basically impossible to ship overseas.

We'll have to see who can bribe, whoops I mean donate, the most money. Self driving cars is being backed by Google, Apple, Tesla, GM, Uber, Lyft...that's a lot of weight to be thrown around.

I actually think most of the push will come from transportation/logistics companies, where labor expenses tend to be quite high. Imagine if trucking companies could replace their drivers with self-driving trucks, for example. They would massively increase their profitability.

Never underestimate the regulatory influence of insurance companies. They are the ones who will call the shots when it comes to driverless car adoption, and I suspect they'll be all for it. Every year, we lose about 30,000 people on the roadways in the US alone. There are massive incentives in almost all corners of society to eliminate human drivers as rapidly as possible.

I'm what you'd call an "enthusiast" driver, but I'm not going to stand in the way of autonomous cars even though it will mean the end of an activity I enjoy. The upside is just too great to reject for the sake of what amounts to a hobby. We're about to find out if Ferry Porsche was right when he said, "The last car on Earth will be a sports car."

Insurance companies make money based on the float of premiums they take in vs what they pay out in claims. If the risks reduce substantially, so will the float.

Insurance companies thrive on quantifiably, risky propositions of which driving is a perfect example.

I will bet you $1,000 that there will be ZERO driver-less cabs in Chicago by New Years Day 2021.

Your statement that there's "too much value to capture" is absolutely correct. The value that is being captured is human labor.

There are entrenched political interests that will slow down adoption of this that will make climate-change deniers look like kindergarten bullies.

I agree that the adoption rate will be slower than the 3-5 years rate stated by the commenter above.

I do think, though, that of the major cities, many people would agree that Chicago is among the most resistant to changes in labor due to customer preferences, due to its unique political system.

That said, though, imagine how absurd it would have been 6 years ago to imagine the market share that the Chicago yellow cab companies are losing, to newcomers that didn't even exist yet. I don't think many people, especially cab companies and Chicago legislators, could have predicted it.

The swell of opposition in legislators against cheaper technological answers to current labor solutions is heroic- the question that remains is, 'How long before consumer demand for the radically cheaper and/or more efficient option makes the opposition moot?'.

The answer usually seems to be: 'Sooner than everyone thinks.'

The first satellite was launched in 1957. The first man landed on the moon only 12 years later in 1969. But that didn't meant that a trip to Mars was in the cards even decades later.

Uber was a matter of applying well understood techniques and waiting for a critical mass of GPS enabled mobile devices with data support. Self driving cars still need to overcome fundamental challenges. It could happen in five years, or 50 years might pass without breaking through the final hurdles. Remember, the last 10% is 90% of the work.

I set a reminder for Jan 1 2021 to see if your prediction is correct. Now i guess we just wait.

Check out an app called : iCalledIt. You can call out your predictions and set reminders :)

I somehow doubt it will be working by 2021...

Are you going to 'call it'?

Elon Musk just did:

"In ~2 years, summon should work anywhere connected by land & not blocked by borders, eg you're in LA and the car is in NY"

https://twitter.com/elonmusk/status/686279251293777920

The people who own the medallions and consequently are politically entrenched are not the people who will be put out of jobs.

Putting large numbers of people out of decent jobs is not something politically savvy people do in cities.

I disagree that that driverless cars will get tied up in regulation. You know who is excited about driverless cars? AARP, the most powerful lobbying organization in America, whose members are the most consistent voters in local elections and who are afraid of being isolated when they lose their ability to drive.

http://www.aarp.org/home-family/personal-technology/info-201...

Regulations are largely local matters (at least on this matter,) and AARP is a federal powerhouse, not a local one.

Parent's comment still stands.

National organizations have influence in local policy when it affects issues they care about. Normally, at least for very contentious issues, this doesn't happen (e.g. abortion, global warming, gay marriage) because those matters aren't handled locally.

An issue like this would absolutely be a priority for AARP and its members. Those members, as correctly noted, are likely voters.

Source: I work with local governments across the country.

It makes sense that Florida is one of the states aggressively allowing the testing of self-driven vehicles.

You're already wrong though.

How do you define "cab"? How do you define "driverless"? Do you mean from arbitrary point A to arbitrary point B, or can there be only pre-defined connections between hubs? Must they travel above ground on ordinary streets, or specialized streets? How many people can they hold? And how do you define "driver"? Under certain interpretations of your bet, the MTA system already satisfies those requirements.

There's not a single legal Chicago taxi-cab that is driver-less.

Don't be fatuous.

Yes, trains and cars count as different things because cars can go from arbitrary point A to arbitrary point B on normal roads.

What is your point?

My point is that it's important to be precise about the terms of a bet like that.

Because the word "cab" is too ambiguous for you. Sigh.

Adoption could be slower than we think, but I don't think regulatory issues will be that big, the whole logistics industry will push it and many more. I also think that possible profits are the stronger incentive than preserving human labor

It is going to take longer than expected to have a perfect driverless car but when it is here the takeover is going to be extraordinarily quick.

I believe the regulatory hurdles will actually go the other way -- suddenly it is going to be very hard to be a human driver on a public road. It is the regulations which will force human drivers from the road, not self-driving technology. Someone can provide a better analogy than I can, but it would be the equivalent of trying to hand deliver a packet rather than send it across the internet.

>I seriously believe it'll be here in no more than 3 years, maybe 5 if there are regulatory hurdles. There's just too much value to capture for everyone to be half-assing it.

We wont have full blown self-driving cars amounting to any more than 10% of the traffic even in 2025.

Maybe!

Per capita vehicle miles have been dropping since 2004:

Article: http://www.ssti.us/2014/02/vmt-drops-ninth-year-dots-taking-...

Graph: http://www.ssti.us/wp/wp-content/uploads/2014/02/2014-VMT-ch...

> Unlike other past dips in driving, this recent downward shift has had no clear, lasting connection to economic trends or gas prices. Evidence suggests that the decline is likely due to changing demographics, saturated highways, and a rising preference for compact, mixed-use neighborhoods, which reduce the need for driving. Some key factors that pushed VMT upward for decades – including a growing workforce and rising automobile ownership – have also slowed considerably. SSTI released a report last September outlining the many contributing factors, with references to supporting literature.

You mention 10%. 10% of total miles driven in 2025 is going to be drastically more than 10% of total vehicle miles driven today (Baby boomers aging out, who drive much less in retirement).

I agree. Anyone who thinks it is coming soon is deluding themselves.

The technology maybe reasonably close but the policies and logistics aren't. The responsibility for car crashes and insurance isn't defined and needs alignment with government policy around the world. And there are still concerns about the behaviour of the self driving cars i.e. it drives to the rules which no one does and how does it handle random scenarios e.g. mattress flies off truck in front.

It's taken years for car avoidance and emergency braking technologies to come to market and even those aren't widely available still.

And it isn't even close. Just the other day Google released a bit of PR that indicates their current cars will pull over and stop if it starts to rain. That basically means their cars aren't usable in all but the driest locales. And handling the rain is exponentially harder than handling good weather driving.

They completely punted on snow as well IIRC.

So all of the truck drivers will move to the rainy coasts and Canada/Alaska, and all of the driving-labour-sensitive industries will move to the sunnier parts of the continent? :P

I still predict that the first major adoption of driverless tech will come for long-haul trucking, not the every-day commute.

Think remote flat desert highways. Trucking companies have strong economic incentives to make it happen, having a person in the driver's seat that has to take frequent legally mandated breaks is slower and more expensive than a machine.

Driverless cars are Google hoopla. I think it's more likely that cars slowly come equipped with increased automation, like adaptive cruise control and automatic braking. As they become cheaper, regulators will mandate some of these features. These are cheap fixes that make driving safer. Drivers and insurers will both like them. But complete computer control of the whole car is a much bigger hurdle and I don't see why we will simply skip to that point rather than implementing cheaper, effective automation first.

I've only seen that autopilot work on a straight, gravel driveway. No way it works on my driveway: A steep incline followed by 90º turn.

Where is my fancy self driving car supposed to park? Cause until that's a solved problem there will always be a place for Lyft.

Yours? Or the one you use with a ridesharing service?

If its yours, your own garage. If you're in an urban area, I'd assume ridesharing services would use lights out parking garages when they're overcapacity.

A car sits idle 95% of the time. That's why GM invested in Lyft. They see a future where there's a need for drastically fewer vehicles to service mobility needs. Yes, yes, rush hour handwaving. You use economic incentives to solve that.

Self-driving cars don't need to park at a destination. They just find the next rider. Privately owned self-driving cars can drive themselves to a carpark that's arbitrarily inconvenient to get to (which will greatly increase congestion, so let's hope that self-driving cars turn out to mainly be used as taxis).

General Motors just invested $500 million in Lyft, and will partner to provide vehicles today and autonomous vehicle technology as it develops in the future.

GM has already committed to autonomous cars in the past, but it got more real when they threw half a billions dollars in the ring and partnered with Lyft.

Driver-less cars will take off at sunny locations like California, Nevada, etc.

But it will take longer for other weather conditions you will find in winter at other locations - a combination of snow/rain and evening sun rays are a showstopper for current LIDAR and camera based systems.

> Let's assume self-driving cars take off in <5 years. That means that this isn't "winner take all". It means its a commodity that's going to be driven down to razor thin margins. In that case, neither Lyft nor Uber have much of a business, and VCs are better off burning their money in a barrel.

Possibly. But Uber raised 1B at the same time. So that suggests that the VC world still believes in the Uber model (or that they think Uber comes out as one of the winners in self driving cars). The fact that Lyft failed to get any traditional investors suggests that those investors believe there is a significant difference.

Lyft and uber are competing in 2 completely different spaces. Uber is most similar to a company like the nasdaq and lyft is pretty similar to a cable company that existed in a liquid market.

Uber is an information arbitrageur. They make markets and have high velocity. By this I mean they are well capitalized and all of their money is spent on people and developing software. Google collects and organizes the worlds information, uber it's movements.

Lyft, like fed x, runs ligistics and information analytics but largely inside of its own narrow domain here being last mile transportation. Lyft focuses on comfort and style and is modeled on something like virgin airlines. It will continue to make money (i have reversed my thinking here) but not a lot as the industry becomes commoditized.

Uber will pivot into something like a CDN in a few years doing end to end delivery or will become software vendors of autopilot technology.

Obviously, I dont know what each company will do, but they are super different

This is wrong. The analogies you make are tenuous. They're doing business in the exact same space. One is much bigger than the other.

Their differences are due to their relative positioning in the market, i.e. due to the opportunities they have as a consequence of their relative scale.

>>Possibly. But Uber raised 1B at the same time. So that suggests that the VC world still believes in the Uber model

No, it simply suggests that VCs behave as a herd, and one well-known VC firm investing in a company suddenly makes the others want to invest in it too due to FOMO (fear of missing out).

That answers why a second VC might join a round. It doesn't explain how rounds work in general though. I'm not convinced we can hand-wave over the first VC joining. They still have criteria they're using, and they seem willing to take on a lot of long-shots. The fact that none of them picked up Lyft's latest round still says something.