Talk:Tesla Autopilot

Function specifications
I found the table row labeled "Freeway Interchanges" to be less than informative. First, "Freeway Interchanges" is not a function. The table doesn't say what the car is supposed to do at a freeway interchange, nor does the article have a section on the topic. The second point is that there is no distinction between the unspecified feature set between HW1 cars and HW2 cars. As I understand it, HW1 cars are supposed to take freeway exits under navigation when the car is in the land adjacent to the exit. This feature does not currently exist. HW2 cars supposedly one day will exit one freeway and merge onto another.

Also the use of the phrase "hands free" with HW1 Autopilot is misleading since the feature requires hands on the wheel. I propose removing "Hands-Free On-Ramp to Off-Ramp" from the table completely, as it is both misleading and duplicative of other features listed in the table.

Also, I think it would be helpful to distinguish between features that exist or are planned. For example, Autopilot was announced in October of 2014, and there is still no freeway interchange feature that actually exists.

I'll add lane departure warning to the table. Dr. Conspiracy (talk) 13:15, 22 March 2017 (UTC)


 * Google "freeway interchange tesla autopilot" and you will see many sources say it is a feature in HW1 and HW2. The hands-free is also feature since it since it works without 100% contact with your hands, but hand-free is for a limited time. I added the word "limited" to the table. --Frmorrison (talk) 15:00, 22 March 2017 (UTC)

Beta status during Joshua Brown incident
The text says that Autopilot was in Beta status during the incident, but that's misleading. Autosteer was in Beta but Tesla wasn't claiming that TACC or AEB were in Beta. Nobody is claiming that the accident had anything to do with the car's inability to steer properly. NHTSA found that TACC was a feature that allowed the car to follow other vehicles and that neither TACC nor AEB were designed for cross traffic detection, nor did Tesla claim that they were. NHTSA also investigated whether drivers understood how the system was supposed to be used and found that overwhelmingly drivers knew how it was designed to be used.

Claiming that it wasn't Tesla's fault because it was in Beta is beyond misleading, and is downright irrelevant. The NHTSA report should be cited, as well as how NHTSA found that it was working as designed and worked well even when compared to the best such systems on competing vehicles. Hagrinas (talk) 20:20, 24 October 2017 (UTC)

Use of primary reference is poor
https://www.tesla.com/autopilot changes too frequently and old references do not back the text, nor is the page dated. Wikipedia policy is to prefer secondary sources rather than primary sources for references. WP:PRIMARY This is a reminder that the policy makes sense. Daniel.Cardenas (talk) 05:54, 15 May 2019 (UTC)

False claims
I'm amazed how many false claims are in this article stemming from quoting Tesla's overly wishful thinking. Tesla's claims should be prefixed with "According to Tesla ...", rather than stating them as fact. Example malarkey about hardware 3:
 * April 2019 Full Self Driving computer (FSD)

That what was said about hardware 2. Should we put FSD tag on hardware two tag also? If we change everything where it says FSD to PSD, partially self driving, it is a hundred times more times likely to be true. Daniel.Cardenas (talk) 17:43, 21 December 2019 (UTC)
 * I've gone through the article and sanitized and no longer feel it is full of false claims. Daniel.Cardenas (talk) 15:21, 29 December 2019 (UTC)

Feature table
https://en.wikipedia.org/wiki/Tesla_Autopilot#Feature_Table The third column, 2016 EAP & FSD column, is wrong towards the bottom. Issue stems from being different answers depending upon if it is EAP or FSD. Options to fix: Thoughts? I'm leaning towards option one at this point. Thanks,  Daniel.Cardenas (talk) 18:36, 6 January 2020 (UTC)
 * 1) Ignore and over next few months people will have upgraded to FSD HW3 and just have one column for FSD.
 * 2) Split columns out for 2016 EAP vs FSD.
 * 3) Just put two answers in for last 3 rows, like is done in last row.

HW3 and 2019
https://en.wikipedia.org/wiki/Tesla_Autopilot#HW3 I think it is important to note that last year Tesla said AP would be feature complete. A goal that has been missed. Important to track AP successes and failures. I reverted a change to make the date 2020. What do you think? Thanks! Daniel.Cardenas (talk) 14:41, 10 January 2020 (UTC)
 * Section expanded and added a couple of references to make it more clear.  Thanks, Daniel.Cardenas (talk) 14:50, 10 January 2020 (UTC)

It appears this article makes confusion between partial automation and self-driving car
According to the NTSB, "If you own a car with partial automation, you do not own a self-driving car. So don’t pretend that you do"

https://www.theguardian.com/technology/2020/feb/25/tesla-driver-autopilot-crash

Overly promotional
I think this article could use some TLC, as it currently reads like an advertisement written by Tesla. I've highlighted a few areas of particular concern:

Full self-driving
Calling the HW3 a "full self-driving computer", and calling existing features "partial" self driving abilities, is misleading - currently Tesla has a level 2 "hands off"/partially automated driver assist system. I understand that Tesla themselves sell this as "Full Self Driving Capability", but it's important to be clear about the true current state of the technology. See: https://en.wikipedia.org/wiki/Self-driving_car#Definitions

Safety
Many sources mention issues that have been raised surrounding safety of the Autopilot system, including issues with Tesla's self-reported safety claims, NTSB investigations, inability to see stopped vehicles, and claims that the "Autopilot" name itself is deceptive and dangerous.

HW1, HW2, HW3
The History, Driving features, and Technical Specifications sections all repeat a lot of the same information, and devote undue weight to the differences between the hardware versions.

Fluff/promotional material
There is a lot of apparent fluff and promotional material throughout the article, such as the "Future development" and anecdotes in the Public debate section, and a separate Awards section for one award.

I'd love to hear any thoughts or feedback, and look forward to improving this article together. Stonkaments (talk) 19:09, 29 April 2020 (UTC)

Analyst
Sonkaments reverted edit about Morgan Stanley indicating that only Tesla is an automaker making money with there autonomous software? Is that in dispute? Is there another automaker making money? Then promotes another analyst that says Tesla is last. Very bias in my opinion. Daniel.Cardenas (talk) 16:01, 30 June 2020 (UTC)
 * Autopilot is explicitly *not* autonomous software. It is a level 2 ("hands off") advanced driver-assistance system. Of course there are numerous other automakers selling cars with advanced driver-assistance systems:
 * https://en.wikipedia.org/wiki/Advanced_driver-assistance_systems#Implementations
 * https://www.cars.com/articles/which-cars-have-self-driving-features-for-2020-418934/
 * https://www.autoblog.com/2019/11/03/repairing-advanced-safety-systems-create-conflict/
 * As far as true level 4 autonomous vehicles, to my knowledge the only project that has starting serving paid customers is Waymo - https://www.theverge.com/2019/8/26/20833215/waymo-self-driving-car-taxi-passenger-feedback-review Stonkaments (talk) 19:43, 30 June 2020 (UTC)
 * I fail to see how your comments relate to the bias editing. Daniel.Cardenas (talk) 20:29, 30 June 2020 (UTC)
 * Can you be more specific on the bias you see in the article? I was focused on the specific claim in your edits, and it seems clear to me from the sources that:
 * 1) Tesla Autopilot is an advanced driver-assistance system, not autonomous driving software
 * 2) there is a competitor (Waymo) that is actively monetizing their autonomous cars.
 * Thus the claim that Tesla is the only company “fully monetizing its autonomous driving assets at scale" seems wrong on both fronts, and doesn't seem fit for inclusion.
 * The Nagigant study you're referring to seems to be the most respected third-party analysis of the autonomous driving industry. There is a well-documented pattern of optimism and hype in the industry, so it's important to rely on neutral and objective sources. Sell-side analysts are frequently biased towards an overly optimistic view of the companies they cover.Stonkaments (talk) 21:25, 30 June 2020 (UTC)
 * Sorry, I should have been more clear. Instead of making money, I should have said making a profit.  Waymo is losing close to a billion dollars per quarter, if my memory of readings is correct.  In terms of bias editing you are only putting negative info in the lead.  Without your edit there is balanced negative and positive.  The negative is the earlier sentence.  How do you define "autonomous driving software".  It meets the definition of level 2 autonomous system.  I don't think anyone should consider losing a billion dollars per quarter as “fully monetizing its autonomous driving assets at scale" or making money.   Quite the opposite is true, Waymo is losing a staggering amount of money.  Navigant study respected?  By whom? Daniel.Cardenas (talk) 21:56, 30 June 2020 (UTC)
 * WP:NPOV calls for "representing fairly, proportionately, and, as far as possible, without editorial bias, all the significant views that have been published by reliable sources on a topic." Navigant's studies of the autonomous driving industry have been widely cited throughout the years,  and the fact that they place Tesla in last place is clearly a significant view, and worthy of inclusion.
 * As for the analyst's claim, as I noted above, I don't believe a sell-side analyst's opinion should be considered reliable or significant, due to their tendency to write overly optimistic notes on the companies they cover. Plus, as I also noted earlier, many other auto manufactures are also making a profit on their level 2 advanced driver-assistance systems, so the analyst's claim is false and misleading, and does not belong in the article. Stonkaments (talk) 22:33, 30 June 2020 (UTC)
 * WP:NPOV calls for "representing fairly, proportionately, and, as far as possible, without editorial bias, all the significant views that have been published by reliable sources on a topic." Navigant's studies of the autonomous driving industry have been widely cited throughout the years,  and the fact that they place Tesla in last place is clearly a significant view, and worthy of inclusion.
 * As for the analyst's claim, as I noted above, I don't believe a sell-side analyst's opinion should be considered reliable or significant, due to their tendency to write overly optimistic notes on the companies they cover. Plus, as I also noted earlier, many other auto manufactures are also making a profit on their level 2 advanced driver-assistance systems, so the analyst's claim is false and misleading, and does not belong in the article. Stonkaments (talk) 22:33, 30 June 2020 (UTC)

Inaccurate description of HW3
The article refers to HW3 as a “Tesla designed SOC” which is of course not the case.

The HW3 was designed and produced by Samsung, though arguably Tesla had some input in the design.

It’s an SOC that uses standard ARM72 cores and standard ARM Mali GPUs.

The article gives the impression that Tesla designed the entire chip themselves, whereas in reality they picked some of Samsung/ARM’s “off the shelf” parts and combined them into an SOC.

The best comparison would be when say.... Nokia, is making a new cellphone and buys an SOC from Samsung. They come up with some specifications (price, battery life, speed) and together with Samsung pick the parts that meet the pricepoint and the particular performance needed.

There is a rather huge difference between DESIGNING an SOC (A multi billion dollar venture that takes 3-4 years) and ordering a customized version of an existing SOC, which is the case here and which the article should reflect.

A good comparison would be a fast food meal. There is a difference between peeling all the potatoes, cutting them into fries, making hamburger patties and baking buns. And going to McDonald’s and picking items off their menu, which is what Tesla did.

192.38.141.210 (talk) 13:26, 16 July 2020 (UTC)
 * That's a good point - do you by chance have a link to a reliable source describing the design process for HW3? Stonkaments (talk) 21:01, 16 July 2020 (UTC)

That is an inaccurate description of HW3. Every boutique chipmaker combines off the shelf parts along with custom design. The bulk of the NN processing is not done by the Arm chip, which Samsung licensed from Arm. Daniel.Cardenas (talk) 01:43, 27 September 2020 (UTC)

Removed some Beta tags
Some of the features like summons and autosteer are not listed on Tesla's web page as beta: https://www.tesla.com/support/autopilot, but I now know that in the car these features are listed as beta. If I would have known that prior to my edit, I would NOT have removed the beta markers. Daniel.Cardenas (talk) 01:45, 27 September 2020 (UTC)

Categorizing the Safety Concerns section into HW1 and HW2?
What do think of categorizing the Safety Concerns section into HW1 and HW2? And in the future there will be a hardware 3 subsection. Daniel.Cardenas (talk) 18:52, 23 December 2020 (UTC)

FSD focus?
FSD will grow in interest. If more people are coming here to learn about FSD, should article primarily focus on that? Should article be renamed FSD? Autopilot can be about HW1+. FSD should be about HW3 or latest hardware. Article split might make sense in the future. Daniel.Cardenas (talk) 17:44, 11 March 2021 (UTC)


 * We can have a separate article Tesla Full Self Driving, but this article should not be renamed in my opinion. These are different things, and Tesla Autopilot is very notable by itself. I think we should also make clear the distinction between basic Autopilot and FSD. --Ita140188 (talk) 03:04, 13 March 2021 (UTC)
 * I was also thinking about creating a new article. Should be a good idea. Mrconter1 (talk) 07:07, 28 September 2021 (UTC)

Can we remove the "cleanup" tag from the History section now?
Can we remove the cleanup tag from the history section now? I have done some work on that section. If the tag needs to remain, what other work needs to be done? ReferenceMan (talk) 15:57, 26 May 2021 (UTC)
 * Right now it is basically a collection of one-liners organized in semi-chronological order, so no I would say it isn't cleaned up. It needs to read like how a section of an encyclopedic entry normally reads, that is with well-constructed and full paragraphs that convey progressions like in Incidents section.QRep2020 (talk) 23:18, 26 May 2021 (UTC)
 * I have done a number of improvements, and it is now much less a collection of one-liners, although not perfect. Any further suggestions you have would be welcome.ReferenceMan (talk) 14:47, 28 May 2021 (UTC)
 * I should have some time tomorrow to give it a close proofread. It does look better though. QRep2020 (talk) 23:04, 29 May 2021 (UTC)
 * There are multiple citations to not only the Tesla website but also to Teslarati and the Tesla Motor Club message board - independent reliable third-party sources are needed. There is also a citation from the Federal Aviation Administration in the middle of the third paragraph for some reason. QRep2020 (talk) 03:57, 31 May 2021 (UTC)
 * ✅ I believe I have resolved those issues. Please review.  What other changes are necessary to remove the "cleanup" tag? ReferenceMan (talk) 15:57, 31 May 2021 (UTC)
 * What other changes are necessary to remove the "cleanup" tag? ReferenceMan (talk) 01:58, 5 June 2021 (UTC)

Outstanding crashes
There is a reference in the text to outstanding crashes where there might or might not be a relation to the Tesla autopilot. I feel like its important to even have reference to ongoing cases, but the pure reference to an external article is not covering it sufficiently. This is also true as the list might change over time and also in terms of a near term content creation and provision ability for Wikipedia. --Alexander.stohr (talk) 08:48, 9 June 2021 (UTC)
 * Sure, feel free to start that subsection in the new section I have created from the former Crashes subsection. I am guessing the Woodlands incident would go in this Outstanding Crashes subsection then, yes?

Merge Tesla Dojo?
Should Tesla Dojo be a standalone article, or would it be better to merge here (or maybe History of Tesla, Inc.)? Your input would be appreciated; the discussion can be found at Talk:Tesla Dojo Stonkaments (talk) 11:52, 20 August 2021 (UTC)
 * Second. QRep2020 (talk) 15:51, 20 August 2021 (UTC)

The Woodlands, Texas, USA (April 17, 2021)
The section is erroneously popular because initial report said no one was in the driver's seat. Yes it was an error. We don't need to put mistaken "autopilot use", when autopilot was not in use in article about autopilot. Can put info in Tesla Model S notable crashes. Daniel.Cardenas (talk) 17:49, 9 December 2021 (UTC)
 * The latest update from the NTSB investigation, per the https://www.cnbc.com/2021/10/21/ntsb-driver-seat-occupied-in-fatal-tesla-crash-in-spring-texas.html article, stated that a full investigation was still underway, one "includ[es] Tesla’s advanced driver assistance system, the postcrash fire, occupant egress, and results of the driver’s toxicological tests." Therefore, the reasonable hypothesis, as claimed by the sources regarding Autopilot being involved in this incident, has not been discredited. QRep2020 (talk) 01:58, 10 December 2021 (UTC)
 * Ridiculous to put crashes in this article as long as use of autopilot has not been discredited. Should only put articles in autopilot article that are about autopilot.  Daniel.Cardenas (talk) 16:53, 10 December 2021 (UTC)
 * In this case, it has been discredited, so the section about the Woodlands, TX, crash should be removed in its entirety, because it was theorized that Autopilot was used, but that theory was later proven wrong. Unknown0124 (talk) 19:49, 20 May 2022 (UTC)
 * And yet Consumer Reports (known Tesla haters), believing the initial narrative that Autopilot was active during the crash, go out and egg on more people to do more of those stunts and post videos of them on places like TikTok. Unknown0124 (talk) 19:46, 20 May 2022 (UTC)
 * I've removed the section. As concluded by the NTSB investigation, Autopilot was not involved in this accident. Thus there's no reason to keep it in the article.
 * If someone wants to move the information somewhere else, the latest version of the article with the section still present is here. --Veikk0.ma 00:30, 16 February 2023 (UTC)
 * I see you reverted my deletion of this massive section of non-Autopilot related information. Unless there's some kind of a real argument for keeping it, I'll probably just do it again. "The section explains the situation just fine" is completely irrelevant when the situation has nothing to do with Autopilot in the first place. This article is about Autopilot, it is not a noticeboard of car accidents involving Teslas.
 * If this car accident really needs to be mentioned, we should dedicate maybe 1-3 sentences to it at most, just to make it clear that it wasn't Autopilot-related. We certainly don't need 6 paragraphs, 10,000 characters and an animated GIF to illustrate an event that has nothing to do with the subject matter of the article. --Veikk0.ma 19:40, 16 February 2023 (UTC)
 * Disagree, it is fine as it stands. I presented an argument in the recent revision purpose: "There are likely users who do not know that the NTSB made this conclusion after all of the coverage, and where else on here would they learn otherwise?" Additionally, the accident got an inordinate amount of attention by the media - the subsection reflects that. QRep2020 (talk) 22:33, 16 February 2023 (UTC)

Toyota
I don't see how an item of Toyota following Tesla is relevant as a rebuttal to expert analysis of the technical merits, it's a synthetic combination. --StellarNerd (talk) 20:14, 20 June 2022 (UTC)


 * The fact about Toyota following Tesla's vision approach (as argued in the source material) is relevant to the article but was presented in the wrong way, agreed. QRep2020 (talk) 20:54, 20 June 2022 (UTC)

"Third" versus "second" motorcycle crash
The language regarding the three motorcycle crashes is a bit complicated - both of the sources claim that there were three cases involving a motorcycle and a Tesla that was using Autopilot. Yes, the CNN article also includes something about Riverside police saying that the Tesla in that case did not hit the motorcycle while the rider was on it, but it does so without addressing the tension that I think ReferenceMan and Ptrnext are concerned with. Does the sentence as it stands constitute Original Research then? Or maybe there needs to be other sources included? Or can the point be rephrased somehow to dissolve the tension entirely? QRep2020 (talk) 23:48, 25 December 2022 (UTC)


 * I rephrased it as "multiple" crashes without mentioning a specific number. I think that's a reasonable compromise that does not contradict the references while accurately discounting the Riverside crash. Cheers, Mliu92 (talk) 17:43, 18 January 2023 (UTC)

Flawed Animation?
The text says "The Tesla was proceeding west on Artesia against the red light when it struck the Civic, which was turning left from Vermont onto Artesia."

but the accompanying animation shows the Civic emerging from the straight lane, rather than a left-turn lane. 98.51.94.150 (talk) 09:13, 22 January 2023 (UTC)


 * Speaking of the animations, who made them exactly? Can we get some more, pretty please? QRep2020 (talk) 11:26, 14 April 2023 (UTC)

Technical and information over-load
The start of the article is over-long, overly-technical and serms to read like a poor public relations hand out. Given that the safety of hands-free/self-driving car remains a major issue of concern, should not the 'Safety, statistics and concerns' section be placed nearer the top of the article - and given greater importance? — Preceding unsigned comment added by 95.151.2.44 (talk) 20:34, 9 May 2023 (UTC)

Elon Musk says
Given the continued, demonstrably false claims by Elon Musk that fully autonomous vehicles are coming soon, stretching back to 2013, I do not believe we should place any further credibility in Musk's predictions and references that primarily repeat his statements should be discounted. In addition, news organizations have been asked to view Musk's words skeptically. I have taken the time to document these prior claims and determined whether or not they were met, based on the simple criterion "Did a Tesla vehicle meet SAE Level 5 autonomy by the date claimed?", as noted below.

Mliu92 (talk) 14:44, 9 August 2023 (UTC)
 * Beautiful. QRep2020 (talk) 01:43, 10 August 2023 (UTC)
 * Hey Mliu92, I would love to have this chart see the light of day. If we hesitate to incorporate it here, would you mind if I brought it over to Criticism of Tesla, Inc.? QRep2020 (talk) 05:51, 1 October 2023 (UTC)
 * @QRep2020 Sure, either way, no objections to incorporating the chart. Cheers, Mliu92 (talk) 06:05, 1 October 2023 (UTC)

This could actually be a good article
It summarises a lot of information at a good level of detail. Unfortunately, the article as a whole reads rather hit-pieceish.

I would encourage readers interested in the technology, especially safety conscious potential buyers, to seek information directly from actual users and experience the system themselves if possible. This article does, sadly, not provide a balanced description (it seems to be a recurrent issue in the automobile industry, I recall similar criticism when seat belts were made mandatory, or when ABS and airbags were introduced, for instance). 85.160.32.101 (talk) 10:07, 23 September 2023 (UTC)

FSD Capability
Tesla changed the name of FSD to FSD Capability. Should I rename a few of the occurrences to that? - Daniel.Cardenas (talk) 16:25, 14 June 2024 (UTC)


 * Maybe in a few places with the full name spelled out, but 'FSD' works fine and is still used consistently by RS. QRep2020 (talk) 19:33, 14 June 2024 (UTC)

Split
People complained about the size of the article, so I took to divvying it up. That said, I encourage users like Mliu92 to help summarize the many, many Autopilot accidents and fatalities at the Fatal and nonfatal crashes subsection. QRep2020 (talk) 13:37, 15 June 2024 (UTC)