There's two tiers-- Enhanced Autopilot (EAP), and FSD. To own FSD, you need to have already purchased EAP. However, you do not need to purchase FSD to have EAP.
Enhanced Autopilot is (in simple terms) lane keeping + adaptive cruise control. It's designed only to work on highways, and does not need the HW3 board nor will owners that only purchased this get the new hardware.
FSD owners will get the HW3 board so that Tesla can fulfill the promise of providing FSD that the owner originally purchased with their car (or after-the-fact). FSD will (in theory) handle self-driving in most cases, such as regular street driving with stop lights, stop signs, roundabouts, etc.
Why should we believe Tesla that HW3 is actually capable of FSD this time?
https://www.techrepublic.com/article/tesla-to-release-fully-...
If they don't have FSD by mid-2019, that's two years off that target. Deadlines slip, companies overpromise, this is a hard problem, all that jazz, but by early 2020, people who leased Teslas that they were told would have FSD are going to be deciding what their next move is. If part of the calculation is "FSD still ain't here yet, but Tesla is promising a free upgrade to HW4 which they say will really lay the foundation for FSD this time Elon pinky-swears," I would submit that it's gonna matter.
When you select full self-driving capability in the Tesla configurator, it states explicitly and prominently that this isn't enabled (or ready!) yet, and is subject to legal approval.
Anyone that buys a Tesla, thinks it is fully autonomous, and dies as a result is failing natural selection. Third parties (pedestrians and other drivers) on the other hand are victims, but have there been any examples of this with Tesla cars? My not-full-self-driving AP2 Model X can detect pedestrians, and will also emergency stop if it detects things. This is aside from the fact I, as the driver, should be emergency stopping if I see a forthcoming collision with a pedestrian!
Think of Autopilot like a horse. You still need a driver, but at least you've got another set of eyes on the road. Sometimes it will notice things you don't, sometimes it will get spooked by things it shouldn't do, but two sets of eyes/two brains are better than one.
The things that could happen are that either Tesla ships full self-driving (FSD) and better delivers when it does (i.e. it has to work reliably).
Or it puts off the customers again, promising a new hardware 4, 5, 6, ... capable of FSD - in which case people better read the instructions (and pay attention at all time lest they end up in a divider or under a trailer).
You'd have to elaborate as of why this would have been the main reason (vs other possible reasons...).
I swapped in October to a 3PE, now I'm seeing that another upgrade is incoming, and that THIS one is required for FSD.
I've spent plenty of money with this company, but handing over $5,000 more dollars for a feature that is unproven is not something I feel like I should need to do; especially when I'm clearly advertised to at the time that I can buy the feature later for $6,000.
If fully autonomous driving becomes available, and it costs more than what was advertised, they are going to have a big problem.
Can you go into depth about why this comparison is inappropriate, why it unlikely to be solvable in this way?
(I can certainly understand “not now”, but your claim seems rather stronger than that).
We’ve made extraordinary progress, but, ultimately, we’re still training systems in extremely simplistic ways to solve (conceptually) simplistic tasks. Localizing objects (into a set of categories no less) is only a fraction of what your brain does when looking around. Moreover, we still struggle with this. One of the Tesla accidents was due to a segmentation error (two simimilary colored and textured objects were merged).
You'd need a storage rack in your frunk to record any amount of raw sensor data. That's before you consider their production AP hardware doesn't have cycles to spare to do any of that recording.
Source?
> You'd need a storage rack in your frunk to record any amount of raw sensor data. That's before you consider their production AP hardware doesn't have cycles to spare to do any of that recording.
I have no idea on what you base your claims, but without more information it is hard to figure out what you have in mind...
What curious people have found is that the car just sends a basic disengage report with maybe 10s of h264 video or spaced far apart raw images from the cameras.
You don't (always) need full resolution with 30 fps raw frames for training your model though. If you are looking for missing exit, stopping at a traffic light, etc. you need front facing camera (so 3) and a few raw frames (and the sensor is 12 bits not 16). One frame is 1.8MB raw, but if you train on YUV images, this would be even smaller. With a 200MB buffer in memory, you can keep 100 frames at any time (and you have 8GB local storage for storing while waiting to upload)
You said "any meaningful amount of data", I guess this is subjective!
https://electrek.co/2017/06/14/tesla-autopilot-data-floodgat...
... until it mis-view an exit lane and attempts to crash/kill you, that is (see events of 2018).
That is kind of deal breaker for me, when talking about 1+ ton metal box moving me around I will take "slow to market but not blind" instead.
I'm still waiting for Tesla to accept reality and put lidar (make the option more expensive if you have to and you probably do, whatever)
The way Elon Musk tells it, they put it into production once it was safer per mile than human drivers. Which seems like a legitimate point.
You use something when it's better than what you would have to use in the alternative, not when it's 100% perfect with no possibility of ever making a mistake under any circumstances. Which is probably not even actually possible to do.
The problem is if human drivers kill more than 30,000 people in a year, it's not news, because it's been that way forever. But one autonomous car kills one person and it's the top story.
I'd caution you against believing anything that he says. It's easy to extract from the data an unfair comparison that shows the group with Tesla AP being safer, but comparing it against cars in totally different driving context or against the safety of motorcycles is just completely deceptive. The stats indicate that Tesla w/AP enabled on freeways is less safe than Tesla w/o AP enabled on freeways.
That's all marketing fluff.
The only U.S. insurance company offering such a discount is some random startup insurance company available to only a small percent of the population. The company is about 2 years old, requires significant tracking of your driving behavior, and it's entirely unclear why they are offering the discount.
The only UK insurance company offering such a discount said “it was too early to say whether the use of the autopilot system produced a safety record that justified lower premiums. It said it was charging less to encourage use of the system and aid research.” It made this decision working closely together with Tesla, it wasn't based on any actuarial data.
Also, from that article, it looks like the NHTSA found that crashes were 40% lower in cars with autopilot. [2] I don't know enough about insurance and how it works to know if that counts at all, but it shows that it's not all marketing fluff.
[1] https://www.businessinsider.com/tesla-liberty-mutual-create-...
[2] https://www.businessinsider.com/tesla-autopilot-cuts-crash-r...
"Liberty Mutual Insurance offers discounts for various electronic safety features; such as Adaptive Cruise Control, Adaptive Headlights, Collision Preparation Systems (including automatic braking), Blind Spot Warning, Lane Departure Warning and Rear-View Cameras"
I'd be interested in seeing crash data for Teslas compared with other $50-$130k cars less than 5 years old.
I don't know it well enough to know if there were other factors at play, but it wasn't just that Teslas in general were less likely to crash, only the ones with AP hardware installed.
But Tesla knew it would certainly be perceived as such.
How? I'm not trying to instigate something here, but a quick search of this isn't showing anything except a bunch of groaning about how unfair it is that Tesla gets all kinds of tax breaks for its cars because it's the only one selling electric at any kind of scale.
https://www.businessinsider.com/tesla-liberty-mutual-create-...
"The firm's InsureMyTesla program is an example of Tesla working directly with an insurance partner to mitigate risks and keep rates low."
I thought subsidize meant that they would be basically paying the difference.
If there's no money changing hands in this partnership, then either the insurance companies are really dumb and are just throwing money away, or there's some truth to the stats that tesla keeps bringing up (40% fewer crashes in cars with autopilot) (or there's something else i'm missing!)
Insurance companies may do this because it is a strong signal the customer is gullible to questionable upsells. Maybe they've found that statistically they can sell them more things like overpriced umbrella insurance through direct mail marketing pack ins with their car insurance bill.
(edit: opps, I misread AP as FSD)
It's also only used/active when it is reasonably able to work. Therefore the miles used are inherently better/safer as it is (much less whiteout / torrential AP driving).
This is a luxury human drivers do not have, short of "do not drive those miles at all".
I have always hated this metric. It's disingenuous, and selective.
https://www.businessinsider.com/tesla-autopilot-cuts-crash-r...
They may not be taking the hardest miles, but the only way to improve the overall average is to be doing better than the status quo on the miles they are taking.
The thing is that I can take precautions to increase my safety margin - don't drive impaired, drive in good weather conditions, get lots of sleep, don't be aggressive, don't follow closely, etc. etc. etc.
I don't have a Tesla, but I imagine there's nothing I can do to voluntarily make it safer. It functions normally until it doesn't.
In that sense these Level 2 systems are a lot like q-tips. The box says you shouldn't stick them in your ear, but everybody knows that's what you buy them for. If everybody were to follow the directions, the consumer desirability of the product would plummet.
Even though you're not supposed to take your hands off the wheel, level 2 systems are often referred to as "hands off". And during a 60 Minutes interview Elon Musk is seen taking his hands off the wheel. Unofficially the point of Level 2 automation is that you can take your hands off the wheel. Officially, your hands must remain on the wheel. The unofficial usecase is what actually gets consumers enthusiastic, but the official usecase is what companies like Tesla cover their ass with.
They're having their cake and eating it too.
It's common knowledge that most people think they're better than average drivers, so we tend to compensate and believe we're no better than the statistical average driver. But yes indeed, most of us in fact are better than average drivers whenever we're doing our best to drive carefully.
They were comparing crash and fatality statistics against things like motorcycles.
So no "over a billion" isn't plenty.
Are they safer? Possibly. Statistics show they are safer? That's bollocks. There's not enough data to say yet.
There are lots of things wrong with auto-pilot, but it is far from "garden-variety".
I would advise you to sample one before drawing conclusions.
Tesla’s do great in traffic.
is it flawless, no, but it does work in conditions I didn't find fun driving in. An example was two lane highway, at night, in the rain, with enough jokers with badly aimed lights or not dimming their lights. It did very well and let me feel confident looking away from the road if the oncoming car's lights were too bright.
I did not buy the FSD suite, I don't have a desire for that level of autonomy and I am not confident anyone will pull it off soon. I mean I have seen examples where they map it all but that isn't the same.
Here is a video from Cadillac showing your statement is false: https://www.youtube.com/watch?v=_rxW68ADldI&t=0m39s
Tesla will speed up or slow down and change lanes automatically. This is not the same. Also, Supercruise is only enabled in very limited circumstances. You can use autopilot anywhere even if it isn't officially recommended.
You are still required to engage the turn signal; Tesla will speed up / slow down to execute the merge, eventually, and will steer the merge for you but does not make the decision to do so autonomously (though it'll recommend a new lane to you with a blue lane line on your dash).