Tesla’s Summon-Your-Car Feature Spurs U.S. Safety Inquiry | Hacker News Reader

Tesla’s Summon-Your-Car Feature Spurs U.S. Safety Inquiry

bloomberg.com

45 pointskgwgk35 comments

I worry that Tesla's self-driving ambitions put their entire mission (sustainable transit) and products at risk. Being conservative just isn't in their DNA—and you NEED that DNA to work on autonomy.

Advanced Summon is a good example of this problem. You might say it needs to be deployed to get the data needed to refine it. But they could have spent more time silently collecting data, comparing what the computer would do with what real drivers do. And they could have at least staged the rollout to happen much more slowly.

I just don't think they should have released it this widely in the current state. It's disingenuous to label it as beta and ask users to uncover the kinks with real risk to life and property.

"But they could have spent more time silently collecting data, comparing what the computer would do with what real drivers do."

From their Autonomy day presentation, I thought this is how most of their releases go though? They run it in shadow mode, then launch when the accuracy gets high enough?

That said, I do agree with you on the sentiment regarding Tesla's self-driving ambitions potentially putting their sustainable transit mission at risk.

You're absolutely correct. I'm just saying they could have spent even more time doing it. I wonder what the gating success metrics were for Tesla for this feature.

Electric cars may be less polluting, but that doesn't make them "sustainable transit". It's a band-aid on the many problems caused by car-centric development, which does nothing to solve the fundamental issues:

https://www.donkey.bike/wp-content/uploads/2018/02/autonomou...

If the same car can do 20 trips on its own that graphic looks completely different.

Because now 20 cars will hold less than 20 people on the average. 20 cars on the road with many of them autonomous may only have 15 or maybe only 10 people.

It’s a good point, but counting the taxi driver as increasing efficiency is disingenuous.

Taxis are an invalid alternative.

The truly "Green" alternatives are electric scooters, bikes, busses, and subways.

Electric scooters seem to be the winner in point-to-point movement within a city, although we're going to need to figure out proper safety regulations moving forward (scooter speed limits, road vs sidewalk, etc. etc.) Although still immature, the explosion of scooter-based inner city travel gives me a lot of optimism to that methodology.

20 people on 20 scooters will be more space efficient, cheaper, and environmentally friendly. They move slower than cars, but seem ideal for urban point-to-point travel under 2 miles. Apparently, Segway had the right idea (but the tech wasn't mature enough. Modern Li-Ion batteries + Scooter rentals + Scooter Ride Sharing apps didn't exist in the 00s, but today make it a reasonable technology to use).

-------------

Given a good enough bus network, scooters should work for Point-to-bus-to-point travel, or Point-to-subway-to-point travel, since they're small enough to carry in a backpack.

I think this is just wildly out of touch. Yes, a family of 4 with two small children, getting around on buses and scooters carried in their backpacks.

Children in my area are largely carried around in school busses. I couldn't drive in High School and relied upon walking to the bus stop.

There are unorganized groups ("groups of friends") who travel together from bus-stop back into their neighborhoods. This is just walking. And I don't think I grew up in a very urban area.

---------

Scooters are helpful when walking further distances for personal city travel. But... when it comes to school, busses are extremely efficient at the job. Probably the best at the job: a large group of people are getting to the same location at roughly the same time.

The feature is not ready. The only reason Tesla pushed it out this early is that they want to recognize some of the Autopilot deferred revenue in Q3 to mask a YoY decline in revenue driven by lower ASPs, which damages their growth story.

Damaged growth story? Tesla delivered more cars this quarter than ever before: https://www.cnbc.com/2019/10/02/tesla-tsla-3q-2019-productio...

Lower ASP is to be expected as people are buying many more (and increasing) Model 3 than Model S/X.

The headline of your link literally says "falling short of forecasts". That supports the "damaged growth story" assertion, doesn't it?

No I don't think so. Deliveries are still going up and up! Just not quite as high as the TOP end of the Wall st. forecast.

Selling more cars is great, but YoY reduction in revenue will not justify the valuation of TESLA stock.

Depends how you define "reduction in revenue." YoY growth in units but reduction in per-unit profit is often a very good sign. See Apple when they introduced the iPod mini. Average sale price went down but number of units continued to increase.

Total revenue.

Apple did not make less total revenue when they introduced a lower-priced iPod. The claim above was that Tesla would have had a reduction in total revenue without being able to book pre-sales for this half-baked feature.

From the 10-Q Q2 Quarterly Report;

Deferred revenue related to the access to our Supercharger network, internet connectivity, Autopilot and FSD features and over-the-air software updates on automotive sales with and without resale value guarantee amounted to $1.19 billion and $882.8 million as of June 30, 2019 and December 31, 2018, respectively.

Deferred revenue is equivalent to the total transaction price allocated to the performance obligations that are unsatisfied, or partially unsatisfied, as of the balance sheet date. Revenue recognized from the deferred revenue balance as of December 31, 2018 was $113.5 million for the six months ended June 30, 2019. From the deferred revenue balance as of January 1, 2018, revenue recognized during the six months ended June 30, 2018 was $44.5 million.

Of the total deferred revenue on automotive sales with and without resale value guarantees, we expect to recognize $567.0 million of revenue in the next 12 months.

Even at low speed, Tesla seems to still have trouble avoiding stationary obstacles.

All of the videos I have seen of the summon feature have been pretty bad. A lot of time the vehicle can't seem to figure out what to do. I imagine that will get better with time, but if I had a Tesla I wouldn't be using it now.

I'm surprised they released the feature as-is.

If you use it with a little common sense, it works great. I just got the software update today and tested it out in a parking lot that had minimal foot traffic, worked perfectly.

I've used it three times so far. In all cases, it was an area with no traffic and I was very close.

One time I stopped the car out of an abundance of caution.

It's really a party trick more than a complete feature.

There is a selection bias towards the videos posted and shared the most. That being, the videos showing things going wrong are spread more.

I've used mine a few times and had no issues. One of the times, a cop stopped and watched as my car made its way towards me. He then waved and moved on.

Many others have had no issues with it as well. it is "beta", and yeah, you need to have common sense when using it of course.

The headlines going around when it first came out of "Tesla smart summon gets into accident" was so misleading. It was another driver not looking when backing out and backed into the Tesla. That happens thousands of times daily in parking lots all over. Parking lots are a terrible, unsafe place to begin with.

I am pretty happy with mine. It isn't perfect, but it obviously isn't the nightmare that a lot of people with no actual experience with it are claiming.

Tesla will have tons of data to improve it in the short term.

Does the summon-your-car feature use the same technology that full self-driving robotaxi mode will use when it's released in under a year?

Not sure why anyone is surprised anymore by Tesla releasing unfinished features and then polishing. That's literally what they do.

We're the 'training data'

https://news.ycombinator.com/item?id=21141149

https://news.ycombinator.com/item?id=21105388

I could be wrong about this but it seems the way autonomous driving works is that it needs a ton of real-world failures to learn from. There seems to a mostly-lay-person that there is no way around getting it out there and having people test it out in very odd parking lots with lots of babies toddling around it.

This is the process of evolution, can it be artificially simulated in a way that makes this unnecessary? Can we evolve without people dying or having sex?

Do it in a controlled environment or simulation, there's no reason or excuse to risk real world babies.

But then how do you train it for real world situations where there's real world babies running around?

https://outline.com/V89rFR

It's sad. They will force regulators to act and it will affect legitimate companies working on Self-Driving. It seems most of other companies working on self-driving understand there is a lot of responsibility in pushing half-baked product on the street.

This is why we cannot have nice things.

Do you remember the time when Uber run over a pedestrian?

It’s pretty reasonable to regulate 2 metric tons steel bodies moving among fragile creatures.

Beta tests where bugs mean dead people aren’t cool.

I would definitely not put Uber in the list of responsible ones. As far as I know Waymo and Cruise for example are taking security extremely seriously.

Reply on news.ycombinator.com