Autonowashing

Source

On the 16^th Tom Krisher reported that US Opens Formal Probe Into Tesla Autopilot System:

The U.S. government has opened a formal investigation into Tesla’s Autopilot partially automated driving system after a series of collisions with parked emergency vehicles.

The investigation covers 765,000 vehicles, almost everything that Tesla has sold in the U.S. since the start of the 2014 model year. Of the crashes identified by the National Highway Traffic Safety Administration as part of the investigation, 17 people were injured and one was killed.

NHTSA says it has identified 11 crashes since 2018 in which Teslas on Autopilot or Traffic Aware Cruise Control have hit vehicles at scenes where first responders have used flashing lights, flares, an illuminated arrow board or cones warning of hazards.
...
The agency has sent investigative teams to 31 crashes involving partially automated driver assist systems since June of 2016. Such systems can keep a vehicle centered in its lane and a safe distance from vehicles in front of it. Of those crashes, 25 involved Tesla Autopilot in which 10 deaths were reported, according to data released by the agency.

On the 19^th Katyanna Quach reported that Senators urge US trade watchdog to look into whether Tesla may just be over-egging its Autopilot, FSD pudding:

Sens. Edward Markey (D-MA) and Richard Blumenthal (D-CT) put out a public letter [PDF] addressed to FTC boss Lina Khan on Wednesday. In it, the lawmakers claimed "Tesla’s marketing has repeatedly overstated the capabilities of its vehicles, and these statements increasingly pose a threat to motorists and other users of the road."

These are ridiculously late. Back in April, after reading Mack Hogan's Tesla's "Full Self Driving" Beta Is Just Laughably Bad and Potentially Dangerous, I wrote Elon Musk: Threat or Menace?:

I'm a pedestrian, cyclist and driver in an area infested with Teslas owned, but potentially not actually being driven, by fanatical early adopters and members of the cult of Musk. I'm personally at risk from these people believing that what they paid good money for was "Full Self Driving". When SpaceX tests Starship at their Boca Chica site they take precautions, including road closures, to ensure innocent bystanders aren't at risk from the rain of debris when things go wrong. Tesla, not so much.

I'm returning to this topic because an excellent video and two new papers have shown that I greatly underestimated the depths of irresponsibility involved in Tesla's marketing.

Let me be clear. Tesla's transformation of electric cars from glorified golf carts to vehicles with better performance, features and economy than their conventional competitors is both an extraordinary engineering achievement and unambiguously good for the planet.

Family members drive a Model 3 and are very happy with it. This post is only about the systems that Tesla tells regulators are a Level 2 Automated Driver Assist System (ADAS) but that Tesla markets to the public as "Autopilot" and "Full Self-Driving".

Four years ago John Markoff wrote about Waymo's second thoughts about self-driving cars in Robot Cars Can’t Count on Us in an Emergency:

Three years ago, Google’s self-driving car project abruptly shifted from designing a vehicle that would drive autonomously most of the time while occasionally requiring human oversight, to a slow-speed robot without a brake pedal, accelerator or steering wheel. In other words, human driving was no longer permitted.

The company made the decision after giving self-driving cars to Google employees for their work commutes and recording what the passengers did while the autonomous system did the driving. In-car cameras recorded employees climbing into the back seat, climbing out of an open car window, and even smooching while the car was in motion, according to two former Google engineers.

As someone who was sharing the road with them, I can testify that seven years ago Waymo's cars were very good at self-driving, probably at least as good as Tesla's are now. But Waymo had run into two fundamental problems:

• Over-trust or complacency. Markoff wrote:
  

  Over-trust was what Google observed when it saw its engineers not paying attention during commutes with prototype self-driving cars. Driver inattention was implied in a recent National Highway Traffic Safety Administration investigation that absolved the Tesla from blame in a 2016 Florida accident in which a Model S sedan drove under a tractor-trailer rig, killing the driver.

  The better the system works most of the time, the less likely the driver is to be paying attention when it stops working.
• The hand-off problem. Markoff wrote:
  

  Last month, a group of scientists at Stanford University presented research showing that most drivers required more than five seconds to regain control of a car when — while playing a game on a smartphone — they were abruptly required to return their attention to driving.
  
  Another group of Stanford researchers published research in the journal Science Robotics in December that highlighted a more subtle problem. Taking back control of a car is a very different experience at a high speed than at a low one, and adapting to the feel of the steering took a significant amount of time even when the test subjects were prepared for the handoff.

  But as I wrote at the time:
  

  But the problem is actually much worse than either Google or Urmson say. Suppose, for the sake of argument, that self-driving cars three times as good as Waymo's are in wide use by normal people. A normal person would encounter a hand-off once in 15,000 miles of driving, or less than once a year. Driving would be something they'd be asked to do maybe 50 times in their life.
  
  Even if, when the hand-off happened, the human was not "climbing into the back seat, climbing out of an open car window, and even smooching" and had full "situational awareness", they would be faced with a situation too complex for the car's software. How likely is it that they would have the skills needed to cope, when the last time they did any driving was over a year ago, and on average they've only driven 25 times in their life?

Waymo, the company that knows the problem best, understood seven years ago that "Full Self-Driving" was decades away. But Elon Musk started selling customers "Full Self-Driving" five years ago, two years after Waymo had decided it was too hard to be feasible in the medium term. Neither Waymo nor Tesla have an excuse for missing the problems of complacency or hand-off. They were well-understood in aviation and covered in 1997's Humans and Automation: Use, Misuse, Disuse, Abuse by Raja Parasuraman and Victor Riley:

Most automated systems are reliable and usually work as advertised. Unfortunately, some may fail or behave unpredictably. Because such occurrences are infrequent, however, people will come to trust the automation. However, can there be too much trust? Just as mistrust can lead to disuse of alerting systems, excessive trust can lead operators to rely uncritically on automation without recognizing its limitations or fail to monitor the automation's behavior. Inadequate monitoring of automated systems has been implicated in several aviation incidents — for instance, the crash of Eastern Flight 401 in the Florida Everglades. The crew failed to notice the disengagement of the autopilot and did not monitor their altitude while they were busy diagnosing a possible problem with the landing gear

That was published seventeen years before Waymo figured it out.

Musk was not just selling a product he couldn't deliver, he was selling an investment that couldn't deliver — the idea that a "Full Self-Driving" Tesla would make its owner a profit by acting as an autonomous taxi. Tesla's marketeers faced a choice that should not have been hard, but obviously was. They could either tell Musk to back off his hype (and get fired), or go along. Going along required two new marketing techniques, "Autonowashing" the product and "Econowashing" the investment.

Autonowashing

Mahmood Hikmet's must-watch YouTube video Is Elon Musk Killing People? is an excellent introduction to the thesis of Liza Dixon's Autonowashing: The Greenwashing of Vehicle Automation:

According to a recent study, “automated driving hype is dangerously confusing customers”, and, further, “some carmakers are designing and marketing vehicles in such a way that drivers believe they can relinquish control” (Thatcham Research, 2018). Confusion created by OEMs via their marketing can be dangerous, “if the human believes that the automation has more capability than it actually has.” (Carsten and Martens, 2018). The motivations for this are clear: “Carmakers want to gain competitive edge by referring to ‘self-driving’ or ‘semi-autonomous’ capability in their marketing...” (Thatcham Research, 2018). As a result, a recent survey found that 71% of 1,567 car owners across seven different countries believed it was possible to purchase a “self-driving car” today (Thatcham Research, 2018).

Dixon uses three case studies to illustrate autonowashing. First, media headlines:

Over the past decade, terms such as “autonomous”, “driverless”, and “self-driving” have made increasing appearances in media headlines. These buzzwords are often used by media outlets and OEMs to describe all levels of vehicle automation, baiting interest, sales and “driving traffic” to their respective sites. It is not uncommon to come across an article discussing Level 2 automation as “autonomous” or a testing vehicle as “driverless”, even though there is a human safety driver monitoring the vehicle and the environment

Second, the Mercedes E class sedan:

In 2016, Mercedes-Benz launched a new advertising campaign called “The Future” in order to promote the new automated features launching in its E-Class sedan. The campaign stated:

“Is the world truly ready for a vehicle that can drive itself? An autonomous-thinking automobile that protects those inside and outside. Ready or not, the future is here. The all new E-Class: self-braking, self-correcting, self-parking. A Mercedes-Benz concept that's already a reality.”

The headline of one of the ads read, “Introducing a self-driving car from a very self-driven company.”

Mercedes pulled the campaign, in part because it appeared just after a fatal Autopilot crash and in part because consumer groups were pressuring the FTC.

But primarily Dixon focuses on Tesla's marketing:

It is explicitly stated on the Tesla website and in the vehicle owner's manual in multiple instances that the driver must keep their hands on the wheel and their attention on the road ahead (Tesla, 2019b, 2019a). Despite these statements, Tesla is the only OEM currently marketing Level 2, ADAS equipped vehicles as “self-driving” (The Center for Auto Safety and Consumer Watchdog, 2018).

In October 2016, Tesla announced that “all Tesla vehicles produced in our factory...will have the hardware needed for full self-driving capability at a safety level substantially greater than that of a human driver” (Tesla Inc., 2016a) (see Fig. 2). This announcement also came with the sale of a new Autopilot option called “Full Self-Driving Capability” (FSD). Tesla stated that customers who purchased the FSD upgrade would not experience any new features initially but that in the future, this upgrade would enable the vehicle to be “fully self-driving” (Lee, 2019). This option was later removed, but then subsequently reintroduced for sale in February of 2019.

Dixon Fig. 3

The message that "Full Self-Driving" was actually available, or at least just around the corner, was reinforced in many ways, including promotional videos but primarily by Elon Musk personally hyping the capability:

Tesla's CEO Elon Musk has promoted “Full Self-Driving Capability” on his personal Twitter account, in one case stating “Tesla drives itself (no human input at all) thru urban streets to highway to streets, then finds a parking spot” without clarifying that this feature is not yet enabled (@ elonmusk, 2016). Further, Musk has been seen in multiple TV interviews (Bloomberg, 2014; CBS News, 2018) removing his hands from the wheel with Autopilot active. In one of these examples, he did so and stated, “See? It's on full Autopilot right now. No hands, no feet, nothing,” as he demonstrates the system to the interviewer, who is sitting in the passenger seat (Fig. 3) (Bloomberg, 2014). This behavior is at odds with appropriate use, and is explicitly warned against in the Tesla Owner's Manual (Tesla, 2019a).

Screen grab

Tesla's 2016 "Paint It Black" advert was even more explicit.

Lets get real. It is half a decade later and Gabrielle Coppola and Mark Bergen have just published Waymo Is 99% of the Way to Self-Driving Cars. The Last 1% Is the Hardest:

In 2017, the year Waymo launched self-driving rides with a backup human driver in Phoenix, one person hired at the company was told its robot fleets would expand to nine cities within 18 months. Staff often discussed having solved “99% of the problem” of driverless cars. “We all assumed it was ready,” says another ex-Waymonaut. “We’d just flip a switch and turn it on.”

But it turns out that last 1% has been a killer. Small disturbances like construction crews, bicyclists, left turns, and pedestrians remain headaches for computer drivers. Each city poses new, unique challenges, and right now, no driverless car from any company can gracefully handle rain, sleet, or snow. Until these last few details are worked out, widespread commercialization of fully autonomous vehicles is all but impossible.

Musk wasn't alone in having excessively optimistic timelines, but he was alone in selling vehicles to consumers based on lying about their capabilities. This is bad enough, but the story Hikmet tells is worse. You need to watch his video for the details, but here is the outline (square brakets are timestamps for the video):

Tesla's description of "Autopilot" and "Full Self-Driving" reads:

Autopilot and Full Self-Driving Capability are intended for use with a fully attentive driver, who has their hands on the wheel and is prepared to take over at any moment.

In other words, when these automated systems are in use the driver must monitor their behavior and be ready to respond to any anomalies. Dixon writes:

There is a long-standing consensus in human-automation interaction literature which states that humans are generally poor monitors of automation (Bainbridge, 1983; Sheridan, 2002; Strand et al., 2014). Partial vehicle automation requires a shift in the role of the user, from manual control to a supervisory role. Thus, the demand on the user for monitoring increases

Humans aren't good at the monitoring task; because the system works well most of the time they become complacent. In Tesla's Q1 2018 earnings call Musk explained the problem [29:30]:

when there is a serious accident on autopilot people for some reason think that the driver thought the car was fully autonomous and we somehow misled them into thinking it was fully autonomous it is the opposite case when there is a serious accident maybe always the case that it is an experienced user ... the issue is more one of complacency like we just get too used to it

Thus it is necessary to equip vehicles with Driver Monitoring Systems (DMS), which ensure that the driver is actually paying attention to their assigned task. This has long been a standard in railroad practice. Hikmet's story is essentially about Tesla's DMS, and the conflict it posed between the need to ensure that customers were "fully attentive" at all times, and Elon Musk's irresponsible hype.

Other car companies' DMS are effective. They combine:

• Capacitative sensors ensuring that the driver's hands are on the wheel.
• A camera system looking at the driver, with image processing software ensuring that the driver is looking at the road.
• Infra-red illumination ensuring that the camera system continues to operate at night.

The problem for Tesla was that these systems are not just relatively expensive but also highly intrusive. They would, for example, have been triggered during Elon Musk's televised demonstrations, and completely destroyed the "Look Ma, No Hands" message that Tesla had "Full Self-Driving". They would have constantly reminded customers that the high-tech features they had paid for did not actually exist.

Tesla's solution to this dilemma was to implement a DMS using a torque sensor to determine whether the driver's hands were on the wheel. This suffered from two problems, it did not determine whether the driver was looking at the road, or even in the driver's seat, and the torque needed to activate the sensor was easy to provide with, as Consumer Reports did, a roll of tape [18:13]. Hikmet reports that specifically designed "Steering Wheel Boosters" are easily purchased on-line.

Musk's explanation for why Tesla hadn't adopted the industry standard DMS technology emerged in an April 2019 interview with Lex Freedman [31:19]:

Freedman: Do you see Tesla's Full Self-Driving for a time to come requiring supervision of the human being?
Musk: I think it will require detecting hands on wheel for at least 6 months or something like that. ... The system's improving so much, so fast, that this is going to be a moot point very soon. If something's many times safer than a person, then adding a person the effect on safety is limited. And in fact it could be negative. ... I think it will become very quickly, maybe towards the end of this year, I'll be shocked if its not next year at the latest, having a human intervene will decrease safety.

Steve Jobs notoriously possessed a "reality distortion field", but it pales compared to Musk's. Not the "end of this year", not "next year at the latest", but two years after this interview NTHSA is investigating why Teslas crash into emergency vehicles and Mack Hogan, writing for the authoritative Road and Track, started an article:

if you thought "Full Self Driving" was even close to a reality, this video of the system in action will certainly relieve you of that notion. It is perhaps the best comprehensive video at illustrating just how morally dubious, technologically limited, and potentially dangerous Autopilot's "Full Self Driving" beta program is.

Why can Musk make the ludicrous claim that Autopilot is safer than a human driver? Hikmet explains that it is because Tesla manipulates safety data to autonowash their technology [20:00]. The screengrab shows Tesla's claim that Autopilot is ten times safer than a human. Hikmet makes three points:

1. Tesla compares a new, expensive car with the average car, which in the US is eleven years old. One would expect the newer car to be safer.
2. Tesla compares Autopilot, which works only on the safest parts of the highway network, with the average car on all parts of the network.
3. Tesla doesn't disclose whether its data includes Teslas used in other countries, almost all of which have much lower accident rates than the US.

NTHSA's 2017 report on an early Autosteer fatality depended upon data Tesla generated. Elon Musk tweeted [24:57]:

Report highlight: "The data show that the Tesla vehicles crash rate dropped by almost 40% after Autosteer installation"

Two years later, after they FOIA'ed the data, Quality Control Systems Corporation published a report entitled NHTSA's Implausible Safety Claim for Tesla's Autosteer Driver Assistance System:

we discovered that the actual mileage at the time the Autosteer software was installed appears to have been reported for fewer than half the vehicles NHTSA studied. For those vehicles that do have apparently exact measurements of exposure mileage both before and after the software's installation, the change in crash rates associated with Autosteer is the opposite of that claimed by NHTSA - if these data are to be believed.

For the remainder of the dataset, NHTSA ignored exposure mileage that could not be classified as either before or after the installation of Autosteer. We show that this uncounted exposure is overwhelmingly concentrated among vehicles with the least "before Autosteer" exposure. As a consequence, the overall 40 percent reduction in the crash rates reported by NHTSA following the installation of Autosteer is an artifact of the Agency's treatment of mileage information that is actually missing in the underlying dataset.

Source

Musk is high on his own supply. The interview continues [32:57]:

Freedman: Many in the industry believe you have to have camera-based driver monitoring. Do you think there could be benefit gained from driver monitoring?
Musk: If you have a system that's at or below human-level reliability, then driver monitoring makes sense, but if your system is dramatically better, more reliable than a human then driver monitoring does not help much.

Musk's reality distortion field is saying Autopilot is "dramatically better ... than a human". Who are you going to believe — Musk or the guys in the 11 emergency vehicles hit by Teslas on Autopilot?

As Timothy B. Lee reported nearly a year ago in Feds scold Tesla for slow response on driver monitoring:

The National Transportation Safety Board, a federal agency tasked with investigating transportation crashes, published a published a preliminary report Tuesday about a January 2018 crash in Culver City, California. For the most part, the report confirmed what we already knew about the incident: a Tesla Model S with Autopilot engaged crashed into a fire truck at 31 miles per hour. Thankfully, no one was seriously injured.

But near the end of its report, NTSB called Tesla out for failing to respond to a 2017 recommendation to improve its driver monitoring system.

What the 2017 report said [42:04] was:

monitoring steering wheel torque provides a poor surrogate means of determining the automated vehicle driver's degree of engagement with the driving task

They recommended manufacturers "develop applications to more effectively the driver's level of engagement". Five of the manufacturers responded; Tesla didn't.

Source

Musk is famously contemptuous of regulation, but this May Tesla finally made a slight concession to reality [37:51]. Four years after the NTSB told them to do better and two years after Musk claimed it "wouldn't help much", they implemented a camera-based driver monitoring system. Unfortunately, it uses the cabin camera, which wasn't designed for the job. Not merely does it lack infra-red capability, so it doesn't work at night, but Hikmet also shows it being fooled by a picture [38:25].

This Potemkin system likely won't be enough for China. Simon Sharwood reports that, under new rules announced this month:

Behind the wheel, drivers must be informed about the vehicle's capabilities and the responsibilities that rest on their human shoulders. All autonomous vehicles will be required to detect when a driver's hands leave the wheel, and to detect when it's best to cede control to a human.

And, as a further illustration of how little importance Tesla attaches to the necessary belt and braces approach to vehicle safety, in May Telsa announced that their Model 3 and Model Y cars will no longer have radar. They will rely entirely on image processing from cameras. Removing a source of navigation data seems likely to impair the already inadequate performance of Autopilot and Full Self-Driving in marginal conditions. The kind of conditions that someone who takes Musk at his word would be likely to be using the systems.

As Hikmet says, people have died and will continue to die because Elon Musk regards driver monitoring as an admission of failure.

Econowashing

The stock market currently values Ford at around $50B, General Motors at around $70B, Toyota at around $237B and Volkswagen around at $165B. It currently values Tesla at about 1/3 more than these four giants of its industry combined. That is after its P/E dropped from a peak of almost 1,400 to its current 355, which is still incredible compared to established high-tech growth companies such as Nvidia (P/E 70). The market clearly can't be valuing Tesla on the basis that it makes and sells cars. That's a low-margin manufacturing business. Tesla needs a story about the glorious future enormous high-margin high-tech business that will shower it with dollars. And that is where econowashing comes in.

Musk is repeatedly on record as arguing that the relatively high price of his cars is justified because, since they are will soon be might eventually be completely autonomous they can will might generate a return on the investment by acting as autonomous taxis when they aren't needed by the owner. For example, here is Musk more than two years ago:

I feel very confident in predicting autonomous robotaxis for Tesla next year. Not in all jurisdictions because we won't have regulatory approval everywhere, but I am confident we will have regulatory approval somewhere, literally next year

And here is Musk this year:

In January, after Tesla stock shot up nearly 700 percent over the course of a year, Elon Musk explained how shared autonomous vehicles, or SAVs, can help justify the company's valuation.

Speaking hypothetically on a fourth-quarter earnings call in January, Musk laid out a scenario in which Tesla reached $50 billion or $60 billion in annual sales of fully self-driving cars that could then be used as robotaxis.

“Used as robotaxis, the utility increases from an average of 12 hours a week to potentially an average of 60 hours a week,” he told investors on the call, according to a Motley Fool transcript. “So that’s, like, roughly a five times increase in utility.”

There are already companies that claim to be high-tech in the taxi business, Uber and Lyft. Both initially thought that robotaxis were the key to future profits, but both eventually gave up on the idea. Both have consistently failed to make a profit in the taxi business.

Source

Transport economist Hubert Horan started his epic series Can Uber Ever Deliver nearly five years ago with this table. It shows his prediction that Uber's economics simply didn't work in the taxi business. It is still looking good.

So Uber needed to econowash itself. Horan's second part Understanding Uber’s Uncompetitive Costs explains how they did it:

Uber dealt with this Catch-22 with a combination of willful deception and blatant dishonesty, exploiting the natural information asymmetries between individual drivers and a large, unregulated company. Drivers for traditional operators had never needed to understand the true vehicle maintenance and depreciation costs and financial risks they needed to deduct from gross revenue in order to calculate their actual take home pay.

Ongoing claims about higher driver pay that Uber used to attract drivers deliberately misrepresented gross receipts as net take-home pay, and failed to disclose the substantial financial risk its drivers faced given Uber’s freedom to cut their pay or terminate them at will. Uber claimed “[our} driver partners are small business entrepreneurs demonstrating across the country that being a driver is sustainable and profitable…the median income on UberX is more than $90,000/year/driver in New York and more than $74,000/year/driver in San Francisco”[4] even though it had no drivers with earnings anything close to these levels.[5]
...
An external study of actual driver revenue and vehicle expenses in Denver, Houston and Detroit in late 2015, estimated actual net earnings of $10-13/hour, at or below the earnings from the studies of traditional drivers in Seattle, Chicago, Boston and New York and found that Uber was still recruiting drivers with earnings claims that reflected gross revenue, and did not mention expenses or capital risk.

Horan has just published the 27^th part entitled Despite Staggering Losses, the Uber Propaganda Train Keeps Rolling explaining the current state of the process:

In order to prevent investors and the business press from understanding these results, Uber improperly combined the results of its ongoing, continuing operations with claims about valuation changes in securities of companies operating in markets they had abandoned. To further confuse matters, Uber and Lyft both emphasized a bogus, easily manipulated metric called “Adjusted EBITDA profitability” which does nor measure either profitability or EBITDA.

Part Twenty-Seven returns to an important question this series has discussed on multiple occasions—how can a company that has produced eleven years of horrendous financial results and failed to present any semi-coherent argument as to how it could ever achieve sustainable profitability, still be widely seen as a successful and innovative company? One aspect of that was discussed in Part Twenty-Six: the mainstream business press reports of Uber’s financial results are written by people who have difficulty reading financial statements and do not understand concepts such as “profitability.”

The primary driver of the huge gap between Uber’s positive image and its underlying economic reality was its carefully crafted and extremely effective propaganda-based PR program. This series has documented the origins and results of this program in great detail over the years. [3] In the years before objective data about Uber’s terrible economics became widely available, these accounts were designed to lead customers and local officials into believing that Uber was a well-run and innovative company producing enormous benefits that justified its refusal to obey existing laws and regulations and its pursuit of monopoly power.

Uber propaganda is still being produced since the company needs to give potential investors and the general public some reason to believe that a company with its problematic history and awful financials still has a promising future.

Horan goes on to discuss two recent examples of Uber propaganda, Maureen Dowd's fawning profile of CEO Dara Khosrowshahi, and a Wall Street Journal editorial entitled How Uber and Lyft Can Save Lives based on more of the bogus "academic" research Uber has a track record of producing.

Musk's continual hyping of the prospect of robotaxis flies in the face of the history of Uber and its competitors in the non-automated taxi business. Even if robotaxis worked, they'd be a lot more expensive to buy than conventional taxis. They'd eliminate paying the driver, but the Uber driver is lucky to make minimum wage. And they'd incur other costs such as monitoring to rescue the cars from their need to hand-off to a non-existent driver (as has been amply demonstrated by Waymo's Phoenix trial). If Uber can't make a profit and can't figure out how to make a profit even if cars drive themselves, skepticism of Musk's robotaxi hype was clearly justified.

Now, Estimating the energy impact of electric, autonomous taxis: Evidence from a select market by Ashley Nunes et al puts some real detail behind the skepticism. They compare Autonomous Taxis (ATs) with Conventional Taxis (CTs) and Personal Vehicles (PVs):

The findings of our paper are fourfold. First, we illustrate that an AT’s financial proposition, while being more favorable than CTs, remains — contrary to existing discourse — less favorable than PVs. ATs impose a cost of between $1.42 and $2.24 per mile compared to $3.55 and $0.95 per mile incurred when using CTs and PVs respectively. Second, we identify previously overlooked parameters, the most notable being capacity utilization and profit incentive, as significant impediments to achieving cost parity between ATs and PVs. Omission of these parameters lowers AT rider costs to as low as $0.47 per mile. Third, we document that rebound effects do not require cost parity between ATs and PVs. We show that AT introduction produces a net increase in energy consumption and emissions, despite ATs being more expensive than PVs. Fourth we identify and quantify the technological, behavioral and logistical pathways — namely, conformance to AT-specific energy profile, ride-pooling and ‘smart deployment’ — required to achieve net reduction in energy consumption and emissions owing to AT deployment.

For the purpose of critquing Tesla's econowashing, it is only necessary to consider Nunes et al's financial analysis. Their model includes the full range of cost factors:

Expenditures considered when estimating consumer cost include vehicle financing, licensing, insurance, maintenance, cleaning, fuel and, for ATs specifically, safety oversight (16,22). Requisite safety oversight is assumed to decrease as AT technology advances. We also take account of operator-envisioned profit expectations and fluctuations in capacity utilization rates that reflect demand heterogeneity.

AT cost estimates also consider heterogeneity in vehicle operational lifespan and annual mileage. As the pro-rating of fixed costs over time impacts the financial proposition of ATs, both factors warrant attention. Mileage heterogeneity considers vehicle recharging requirements that may limit vehicle productivity and subsequently, profitability (23,24). Productivity may be further impeded when vehicle electrification is paired with vehicle automatization owing to increased vehicular weight, sensor load and aerodynamic drag, all of which limit vehicle range (25).

We also consider consumer travel time in terms of hourly wages and thus transform differences in travel time to money units (19,26). Literature suggests that productivity benefits are realized through the re-allocation of time to paid or leisure activities that replace the demands of driving on attention. Envisioned benefits include would-be drivers performing other valued activities (19).

They explain the difference between their analysis and earlier efforts thus:

Our financial results admittedly differ from past studies demonstrating cost competitiveness of ATs with PVs (12-14,19). The primary reason for this is that our model accounts for capacity utilization considerations and operator-envisioned profit expectations. Although the inclusion of these factors ‘worsens’ an AT’s financial proposition, their consideration is timely and consistent with commercial fleet operator business practices (20,21).

Conclusion

Why does Elon Musk keep lying about the capabilities, timescales and economics of his self-driving technology? After all this time it isn't plausible that someone as smart as Musk doesn't know that "Full Self-Driving" isn't, that it won't be in 6-24 months, and that even if it worked flawlessly the robotaxi idea won't make customers a profit. In fact, we know he does know it. In Tesla's most recent earnings call Musk said (my emphasis):

“We need to make Full Self-Driving work in order for it to be a compelling value proposition,” Musk said, adding that otherwise the consumer is “betting on the future.”

And last night he tweeted:

FSD Beta 9.2 is actually not great imo

Why Elon Musk Isn’t Superman by Tim O'Reilly suggests why Musk needs people to be “betting on the future.”:

Elon Musk’s wealth doesn’t come from him hoarding Tesla’s extractive profits, like a robber baron of old. For most of its existence, Tesla had no profits at all. It became profitable only last year. But even in 2020, Tesla’s profits of $721 million on $31.5 billion in revenue were small—only slightly more than 2% of sales, a bit less than those of the average grocery chain, the least profitable major industry segment in America.

O'Reilly should have noted where 56% of those profits came from:

Tesla’s revenue and bottom line were helped by the sale of $401 million in emissions credits in the fourth quarter to other automakers who need them to meet regulatory standards.

He continues:

Why is Musk so rich? The answer tells us something profound about our economy: he is wealthy because people are betting on him.
...
despite their enormous profits and huge cash hoards, Apple, Google, and Facebook have [P/E] ratios much lower than you might expect: about 30 for Apple, 34 for Google, and 28 for Facebook. Tesla at the moment of Elon Musk’s peak wealth? 1,396.

The insane 1,396 P/E, and the only slightly less insane current 355 P/E depend upon investors believing a story. So far this year Musk has lost 22% of his peak paper wealth. If Tesla had dropped to Google's P/E Musk would have lost 93% of his peak paper wealth in 7 months. He would be only 7% as rich as he once thought he was. Preventing that happening by telling pausible stories of future technologies is important to Musk.

Update 24^th December 2021: Brian McFadden has figured it out!:

Referring to this New York Times story