This is a story about science, education, YouTube and (most importantly) money. It’s a story about what happens when the desire to educate comes into conflict with the desire to get paid. Finally, it’s a story about how some of the most valuable companies in the world are using educational YouTube as a vector for misinformation. And, to unpack all of this, we’re going to talk about one of the biggest science channels on the platform: Veritasium. In July 2021, the science YouTube channel Veritasium released a video titled Why You Should Want Driverless Cars On Roads Now.
It featured the channel’s host and founder Derek Muller making a characteristically compelling argument for the embrace of autonomous vehicles. Attempting to debunk the anxieties some people have about letting driverless cars loose on public roads, Muller asserts that self-driving vehicles are actually far safer than their human-driven alternatives and have the potential to both save lives and greatly improve contemporary urban life. A large part of the video is delivered from the backseat of one of a fleet of driverless taxis currently operated in Arizona by Waymo, the successor to Google’s self-driving car project. All in all, it amounts to a glowing review of currently-existing driverless car technology in general and Waymo’s implementation of that technology in particular.
Muller so thoroughly punctures any concerns that one might have about autonomous vehicles that it’s hard not to come away from the video as an enthusiastic evangelist for self-driving cars. For a YouTube channel which derives its name from the Latin word for “truth” to be able to make a video which is so uncompromisingly positive—so lacking in caveats or caution—must be reflective of a highly refined technology, right? Well, no. It’s actually reflective of little more than the size of Waymo’s PR budget. For, as Muller admits 30 seconds in, this is a sponsored video—paid for by Waymo and very clearly designed to contribute to their present PR strategy. To describe the video as an advert is perhaps to slightly miss the point. The primary goal here isn’t to promote Waymo’s taxi service to potential customers.
That’s only actually available to unvetted members of the public in a handful of Phoenix suburbs so most people can’t use it anyway. Instead, the video is a more general attempt to reassure viewers about a technology that has been attracting increasingly negative headlines. In December 2020, Uber lost confidence in its long-standing dream of automating its ride-hailing service and sold its self-driving car division to a competitor. A few months later, the poster child for techno-optimism, Elon Musk, similarly revealed that he ‘didn’t expect it to be so hard’ to implement full self-driving capabilities in Tesla vehicles.
And, in May of this year, embarrassment came knocking at Waymo’s own door when a video circulated online of one of its self-driving taxis getting confused by some traffic cones, blocking two lanes of traffic and then trying to escape the support team that came to the aid of the trapped passenger. Such stories have combined with a steady drip of crashes and near-crashes involving cars with self-driving capabilities to make some people feel that driverless vehicles might be, at best, an over-ambitious pipedream and, at worst, highly dangerous. In September 2021, Morning Consult revealed that ‘the share of US adults who believe autonomous vehicles are safer than regular cars’ had fallen by 5 points to just 22%. It’s no surprise that Waymo, which is currently looking to expand its experimental taxi service to new cities, would want to push back on these perceptions; should opposition to self-driving cars gain too much traction, securing permits to test these vehicles on public roads is likely to become much more difficult. That they might turn to a YouTube channel as part of their pushback campaign might, until just a few years ago, have seemed less likely. Nevertheless, those of you who have watched some of my previous videos or who follow me on Twitter will know that I’ve been keeping an eye on this growing trend of companies, billionaires and elite institutions using sponsorship deals with purportedly educational channels to promote their own highly-selective and self-interested vision of the world for some time.
Earlier this year, we took a look at a video by ex-Vox journalist Johnny Harris to foreground this emergent practice in which a creator goes beyond integrating a brief advert for a product or service into their video and turns off their critical faculties entirely, working with a company or institution to craft a video which, from start to finish, is little more than corporate propaganda. Now, as a rule, I don’t think we owe it to creators who make these wholly-sponsored fluff pieces to engage with the minutiae of their arguments. Whatever the merits of Veritasium’s other work—and there are many—the choices surrounding what information to include in this particular video are not being made with reference to what will leave viewers with the most-rounded understanding of autonomous vehicles but, instead, with reference to Waymo’s current PR goals. This is, in short, a completely disingenuous and unserious contribution to the debate around driverless cars. That being said, I often struggle to convince others of this.
Often as a result of the good will that creators have built up over the years, people might accept that a sponsorship tips the scales a tiny bit but will be hesitant to believe that an entire video could be completely worthless. In fact, Muller himself sums-up many people’s attitude towards these kinds of corporate partnerships when, following his revelation that the Veritasium video is sponsored by Waymo, he gives a little shrug as if to say, “no big deal”. Today, then, we are going to engage with the arguments put forward in Why You Should Want Driverless Cars On Roads Now.
We’re going to look at Muller’s invocation of dubious comparisons with other technologies, his misrepresentation of statistics and the highly, highly selective approach he takes to discussing the capabilities of Waymo’s vehicles. What we’ll find is that, while the video doesn’t contain any outright falsehoods, its central argument relies on Muller being extremely careful about what he does and doesn’t say. By the end of the video, hopefully you’ll have a sense of how damaging these pieces can be to our collective understanding of the world, why we should be sceptical of anyone who describes themselves as an educator involving themselves in the creation of such content and why we should resist the further normalisation of such practices. To reiterate, while we’re going to be focussing primarily on Veritasium’s partnership with Waymo today, Muller and his team are far from the only science, tech and/or educational YouTubers to periodically give corporate sponsors the opportunity to influence the editorial content of their videos. In fact, the video we’re discussing today was itself just one part of a larger PR campaign in which Waymo recruited five YouTube channels and tasked them with creating videos which would push back against declining trust in self-driving cars. Some of these channels helpfully included some screen recordings of what one of the creators described as a ‘roundtable’ and what others might try to characterise as a “press event” but, comparing extracts from the event to the resulting videos, seems to have involved little more than Waymo staff briefing the sponsored creators on what to include in their videos.
The uploads which resulted from this campaign varied depending on the size, resources and styles of the channels making them. Shannon Morse’s video simply involves her talking to camera with the occasional bit of stock footage overlaid to make things visually interesting. The video released by Snazzy Labs favours showing clips from the briefing event itself, allowing Waymo specialists to explain how the company’s vehicles operate in their own words. The most entertaining video is by Jason Silva, largely because of how sensationalist and superlative-laden the whole thing is.
Silva doesn’t seem to entirely understand the technology that he’s been told to talk about and so is reduced to drawing upon Arthur C. Clarke to describe self-driving cars as ‘indistinguishable from magic’. The Veritasium video is clearly that which has had the most time and resource pumped into it. This is reflective of both the channel being much more established than the others and, presumably, the much larger payment that would therefore have come attached to the sponsorship deal. Interestingly, where other channels have apparently been asked to include the hashtag ‘#WaymoAmbassador’ either in the description to their videos or in the video itself, the team at Veritasium seems to have negotiated that term away.
This is interesting in that it hints at a desire to not want to be seen to be too close to the company the video will discuss even if the content of the video itself remains dubious. What really sets the Veritasium video apart is that it is the only sponsored video in this campaign in which someone actually rides in a Waymo car. Interestingly, this does not work in the video’s favour. In an attempt to showcase the car’s ability to deal with tricky situations, Muller instructs it to drive through a car park. In the event, there aren’t actually any other moving vehicles around.
Nevertheless, there is a person pushing a trolley along the pavement. Despite this person being a decent distance away and being well within the bounds of the pavement, the car comes to a screeching halt—the wheels actually squeal it’s so abrupt. Muller tries to turn this into a positive, using it as an opportunity to talk about how the Waymo software calculates various different possibilities of what might happen in any given scenario. Ultimately, though, it’s not convincing. One is left to wonder whether the software’s reasoning is so uncertain that it stops every time it sees a person on the pavement just in case they leap out into the road? Or, even if we accept the possibility that the man pushing the trolley might have suddenly jumped out in front of the car, why was it going at such a speed as to require it to stop so dangerously? While the Veritasium video may be a more accomplished production than those released by the other creators involved in this campaign, however, the key talking points remain broadly the same. Throughout each of the videos, the same statistics—many of which we’ll critique later—crop up time and again.
We repeatedly hear about how 94% of serious road accidents in the US are caused by human drivers (more about that figure later) and how the Waymo software has “driven” a cumulative 20 million miles (more about that figure later too). Several of the same cutesy anecdotes are repeated by different creators. In fact, if one watches through all five videos, one finds clips of Waymo employees at the briefing event priming the creators with talking points which later crop up in videos by the other channels framed as opinions that the hosts have come to of their own accord. It would be entirely possible to scrutinise Veritasium’s contribution to this PR campaign by playing through the video in question and ridiculing it for its general cheesiness. There’s one moment, for example, where, having supposedly come to the end of his planned ride in the Waymo vehicle, Muller is apparently having so much fun that he just can’t bear to get out. He therefore calls the support team and arranges to extend his ride.
He tells the operator that ‘I just completed my ride but, like, I don’t wanna get out of the car. I just wanna keep driving’. Putting aside the weirdness of the fact that extending his ride involves him having to get out of the car, wait for it to drive around the block and then come back and pick him up again, are we supposed to believe that the initial plan for the day’s filming was for him to just be left on some random residential street in Phoenix? Nevertheless, while the silliness of this presentational veneer is worth noting, what I’m really interested in today are the more sincere arguments that the video makes for the swift and widespread adoption of driverless cars and the statistics which supposedly substantiate those arguments.
For, some people might posit that sponsored videos like this can be a win-win for viewers and the sponsoring companies. Viewers are able to gain an insight into emerging technologies, the argument might go, whilst companies are able to increase brand recognition. Digging into the claims that Muller makes in this video, however, destroys this premise entirely, with any claims to education being swiftly overshadowed by mountains of corporate fluff and misinformation. Having pieced together some of the financial backdrop to the creation of this video, we can finally get on to examining some of the claims that are made about Waymo’s self-driving cars within it. As mentioned previously, the video kicks off with the channel’s host, Derek Muller, acknowledging the sponsorship deal with Waymo before climbing into the back of one of the company’s vehicles and setting off on his ride around the Phoenix suburbs.
Finally, at just over a minute in, he begins to build the foundations of his argument. The preconception at the heart of this video is that self-driving cars are underdeveloped, can occasionally be dangerous and that we should therefore be cautious in how and when we allow them to be used. Muller evidences the prevalence of this view through showing us the results of a YouTube community poll which revealed that 40% of Veritasium’s audience believed that self-driving car technology wouldn’t be “ready” for at least another 10 years. Highlighting the lack of driver in the moving car that he’s sat in, Muller makes it clear that his goal in this video will be to prove those doubters wrong. Following a brief potted history of Waymo—and, interestingly, long before his (frankly dubious) explanation of how the company’s vehicles work—Muller’s first approach to doing so is to foreground two other ways in which automation is used in our lives which may have once been seen as alarming or dangerous but which most people now routinely accept as safe.
He begins with lifts, suggesting that concerns surrounding self-driving cars can be likened to the public response to the introduction of automated elevators. He reports that ‘when people started putting in driverless elevators, well the public was very concerned and they didn’t want to ride in those elevators’. Now, this obviously isn’t a very good comparison. As Muller quickly admits, elevators only have to travel in one dimension and a lift shaft is a highly controlled environment relative to the multitude uncertainties of public roads.
Interestingly, this section is also entirely plagiarised. A quick Google search reveals it to have been lifted beat-for-beat from a 2015 Planet Money interview with architectural historian Lee Gray. We won’t have time to go into Veritasium’s somewhat disingenuous approach to referencing and citing evidence for this video today, however it just seems unnecessarily rude not to have credited the source here. Muller’s second choice of comparison is more successful, though not for the reason he thinks it is. He continues by discussing Autoland, the automated guidance system that is used to land aeroplanes in conditions of low visibility.
Muller recalls that I saw this particular landing where a plane is coming in into Vienna and it’s just so foggy that the pilots can see almost nothing. I mean, this is the view from the cockpit. And yet they make a textbook landing.
Right on target. So, how do they do it? The answer is, the pilots didn’t do it. It was a Cat III Autoland procedure: the plane just came in and landed itself essentially. The automation of plane landings is obviously far more complex and therefore impressive than the operation of an elevator. The problem with using it to make the case for the roll-out of driverless cars is that Autoland is far more limited and requires the involvement of many more human beings than Muller cares to mention. For one, the suggestion that the plane “landed itself” ignores the fact that a good deal of the technology which makes these landings possible is on the ground.
Those of you that have flown to or from large airports may have noticed an array of antennae at the end of the runway. Collectively, these are called the localiser and they essentially fire out a beam which, alongside other systems, guides the plane to the centre of the runway. The result is that Autoland procedures can only be accomplished at airports with this kit; it’s far from a case of the plane simply responding to the environment around it. Additionally, Autoland involves a lot of human input.
There are several videos on YouTube of pilots flying the exact same model of plane featured in the clip Muller shows us and, watching them, one finds it to be an incredibly involved process. They’re not even just monitoring the landing, either, pilots are actively involved in adjusting the speed of the aeroplane, the flaps, deploying the landing gear and much more. This is before we consider the huge number of people involved in monitoring and making safe the airspace and runway. As we’ll see later, these geographical limitations and the involvement of significant numbers of human beings actually makes Autoland pretty comparable to Waymo’s vehicles. And, admitting these caveats doesn’t make the technology any less impressive. Unfortunately, at least in part due to the video being sponsored by Wamyo, Muller is hellbent on both pretending that automation technology is more capable than it actually is and on pretending that human involvement in its operation, rather than being essential, has the potential to be fatal.
In a pretty distasteful turn, the video draws our attention to the case of Asiana Airlines Flight 214. As Muller describes it Asiana Airlines Flight 214 was on final approach to San Francisco. Attempting to manually land the plane, the pilot accidentally left the throttle at zero and, by the time they realised and tried to abort the landing, it was too late. The plane crashed into the runway seawall and split in two.
The implication of Muller’s version of events is that the pilots of Flight 214 were too egotistical to make use of automated systems in landing the plane and that, if they had done so, the crash could have been avoided. The problem is that the pilots were making use of automated systems in the form of what’s called an Autothrottle. More than this, the official investigation placed the onus for the crash on that very technology.
On publishing the final report, the acting chairman of the investigating body told the BBC that: In their efforts to compensate for the unreliability of human performance, the designers of automated control systems have unwittingly created opportunities for new error types that can be even more serious than those they were seeking to avoid. To use the crash of Flight 214 as proof that autonomy is good and humans bad, then, is frankly ludicrous. This tragic episode is an argument for more caution, not less. There is likely a brilliant discussion to be had about what other implementations of automated and autonomous technology can teach us about how, why, when and where we as a society might want to deploy driverless cars. Unfortunately, again at least partly due to the sponsoring relationship with Waymo, the Veritasium video chooses to do away with such nuance.
By misrepresenting automated systems as more capable than they actually are, the video implies hesitancy over driverless cars to be entirely irrational. And, in the next part of the video, Muller doubles down on this argument. In the interests of not making a long video even longer, we’re going to largely skip over a couple of sections of the Veritasium video. In the first, Muller explains the five levels of autonomous driving and, in a barely-concealed dig at Tesla’s autopilot function, suggests that the supposed safety benefits of driverless cars only kick in once you reach full autonomy. He tells us that ‘all these drivers were trusting the technology too much.
Which makes almost fully-autonomous vehicles potentially more dangerous than regular cars. This is why Waymo decided that the only safe way to proceed is with a car that has at least Level 4 autonomy’. Footage from the Snazzy Labs video (which also makes the same point in the exact same way) suggests that this potshot at Tesla is drawn entirely from the briefing with Waymo staff. The following section sees Muller take a tour around Waymo’s depot and offices. This is fairly unremarkable, although it is the one part of the video where we get a hint at the considerable number of human beings required to ensure that Waymo’s taxi service is safe. Alongside storing the cars, Muller reveals that the depot is ‘also where people monitor all the rides in progress’.
There is then a bit more vlog-style footage as well as the beginnings of an interview with the woman in charge of Waymo’s ride-hailing operations who, unsurprisingly, thinks Waymo is great. With all this over and done with, Muller returns to the central argument of the video: that self-driving cars are safer than their human-driven alternatives. This time, however, rather than trying to convince us that autonomy is safe, he settles for arguing that human drivers are less capable than we might like to think.
This section is a bit “death by statistics”. We learn that ‘74% of people believe they are above average drivers’, that ‘in the 20th century, 60 million people were killed on the road, that’s basically an extra World War’s worth of deaths’ and, finally, that ‘the National Transportation and Safety Board has identified human error as the cause of 94% of accidents’. Unlike in most of their other videos, Veritasium doesn’t provide any sources for these figures, and a brief use of a search engine gives some indication as to why this might be. The figure suggesting that 74% of drivers believe themselves to be “above average” appears reputable, having come from a short 2003 article in the Journal of Safety Research.
After this, however, things become more questionable. The idea that 60 million people were killed on the road in the 20th century (and the comparison to World War 2) is lifted straight from Wikipedia. Wikipedia in turn cites an infographic produced by Information is Beautiful in 2012 to accompany an exhibition at the Wellcome Collection in London. The infographic’s creators acknowledged at the time that there may be ‘some inevitable double-counting, broad estimations and ball-park figures’ and, consulting the spreadsheet which details the origins of each statistic, one finds ‘road traffic accidents’ to be one of those causes of death for which no source is cited. This therefore appears to be a cautionary tale about relying too heavily on Wikipedia.
Whilst it might reveal a slight lack of research rigour, however, whether the 60 million deaths figure is verifiable or not doesn’t really affect the meat of the video’s argument—people dying is tragic no matter the quantity. The more important figure in that respect is the claim that 94% of driving accidents are the result of human error. This statistic was included in Waymo’s briefing to the five creators it recruited as part of this PR campaign but has also previously been used by Muller to make the case for self-driving cars in a video released in 2017 sponsored by BMW. Unfortunately, when we actually read the report which the figure is taken from, we find Muller and Waymo’s weaponization of it to support the roll-out of driverless cars to once again be a gross misrepresentation. The statistic is drawn from a two page document released by the National Highway Traffic Safety Administration in February 2015. Drawing on a body of data surrounding car crashes gathered eight years previously, the datasheet does indeed state that ‘the critical reason [for a crash] was assigned to drivers in an estimated 2,046,000 crashes that comprise 94 percent of [...] crashes at the national
level’. The following sentence, however, continues that ‘in none of these cases was the assignment intended to blame the driver for causing the crash’. This may seem confusing or contradictory at first, but Phil Koopman, an Associate Professor at Carnegie Mellon University who specialises in self-driving car safety, explains it well in a 2018 blog post.
See, labelling the driver as the “critical reason” for a car crash does not necessarily mean that the crash took place due to negligence or malice. Instead, the “critical reason” for an accident is akin to the “final point of failure” leading up to the crash. Say, for example, someone is driving perfectly diligently along a rural road when they unknowingly pass a stop sign which has been concealed by an overgrown bush. When they reach the following crossroads, the road markings have also deteriorated to the point where the driver understandably believes it to be their right of way. They therefore continue straight ahead without stopping at which point they collide with another vehicle. Now, if we wanted to assign “blame” for this crash, we’d probably say that it was the fault of whichever authority is in charge of trimming the hedge and maintaining the road markings.
Nevertheless, the conditions for assigning the “critical reason” in the NHTSA study were such that, as long as there was some possibility that the driver could have seen the other vehicle approaching their side and reacted to avoid the crash, the driver was still marked down as the “critical reason” for the incident. In short, while the 94% figure does include some people who were drunk driving or asleep at the wheel, it also includes people whose sole failing was not quite being able to react sufficiently after suddenly finding themselves in a difficult and dangerous scenario. The overtone of this section of the Veritasium video is clear: that greater adoption of driverless cars has the potential to reduce the number of road accidents in the United States by up to 94%.
Muller tells us that ‘most of these errors are impossible for a machine to make’ and, soon after, that ‘if you have a vehicle that has, you know, LiDAR and Radar and 29 cameras, you’re just not going to hit them’. It’s also heavy on emotional manipulation. Muller stresses that ‘every year, when people are backing out of driveways or parking spaces, in the US, up to 200 people are killed; and it’s frequently older people or children, the children of the drivers’. The problem is that the adoption of driverless cars wouldn’t be a solution in anywhere close to all of these cases.
In our example of the neglected crossroads, an overgrown hedge and lack of road markings would be just as much of a problem for an autonomous vehicle as it would be for a human being. From the heights of Muller’s misleading 94%, Koopman suggests that autonomous vehicles might be able to ‘[cut] fatalities in half if we can achieve parity with an average well behaved human driver’. As we’ll see in the following section, that’s still a big if, with driverless cars far less capable than Waymo and Muller imply.
Furthermore, none of this takes into account the new risks and dangers that would undoubtedly accompany greater use of autonomous vehicles. In this section, then, we again find a refusal to approach the topic at hand with the slightest degree of nuance. A non-sponsored video about driverless cars could have featured a really interesting conversation about what problems autonomous vehicles will solve and those they won’t.
Instead, Veritasium chooses sensationalism and misrepresentation. Of course, there’s a chance that the misrepresented statistics in this section are largely the result of confirmation bias—of stumbling across a figure that supports the argument that Waymo has asked Veritasium to make and simply not questioning it—rather than an active decision to leave out information that complicates their position. When it comes to the video’s highly, highly selective explanation of how the Waymo vehicles work, however, such excuses begin to seem less conceivable. For a video which is attempting to convince us that driverless cars are a safe and reliable technology, Veritasium’s video is pretty light on technical explanation. In fact, it’s not until 9 minutes and 16 seconds in that we’re provided with any kind of overview of how Waymo’s vehicles actually work.
At this point, Muller gives us a brief tour of some of the systems through which the cars monitor their surroundings. We learn about the LiDAR system: ‘up here in the very prominent top, there is a 360 LiDAR so it can see all around the car’. We learn about its use of cameras: ‘there are 29 cameras around this vehicle which gives you full 360 vision’. We also learn how it deals with emergency vehicles: ‘there’s also a microphone up on top to listen to what’s happening in the environment and, if there are sirens, then the car will pull over to the side of the road’. And then… well, that’s it. As Waymo and Veritasium see it, that is all the information that one needs to understand how these cars work.
As you can perhaps tell from my tone, that’s not true. See, one goal of this video is to convince us that, if it wasn’t for technophobic politicians, negativistic NIMBYs and red tape loving bureaucrats, you would be able to hail a Waymo taxi in major cities across the US. And, for that argument to be truly convincing, we have to believe that, if those cultural and political hurdles could be overcome, you could drop a Waymo car in New York or Tampa or Detroit or Denver and it would just work.
As such, in his technical rundown, Muller suggests that the vehicles operate by simply monitoring the environment around them and making decisions based on that. He tells us that: the way the LiDAR works is it shoots out invisible laser beams, scanning around millions of times a second. And then it detects the reflection and how long it takes to come back allows you to determine how far it is to that object. So what it’s doing is, like, painting a 3D picture of the world. This is certainly one, very important, aspect of how Waymo’s vehicles work; the LiDAR and other sensors allow the cars to identify other vehicles, pedestrians, the colour of traffic lights and any other “live” aspects of a driving situation.
Yet, what Muller completely fails to mention is that the cars also require highly detailed maps to be prepared for any area they want to operate in. As Waymo themselves describe it in their Safety Report: Before our cars drive in any location, our team builds our own detailed three-dimensional maps that highlight information such as road profiles, curbs and sidewalks, lane markers, crosswalks, traffic lights, stop signs, and other road features. Rather than rely on GPS, the Waymo Driver cross-references our pre-built maps with real-time sensor data to precisely determine their location on the road. In a similar mode to the earlier misrepresentation of Autoland systems, then, what we find here is an inflation of the capabilities of Waymo’s vehicles whilst also refusing to mention significant human involvement in building, processing and verifying intricate and diligently-labelled scans of every street the cars will drive down.This is before we even begin to take into account the team which monitors any Waymo journeys which are currently taking place or the support teams which roam the city prepared to spring into action when things go awry Without wanting to pre-empt my conclusion too much, we have here a perfect example of why we should be sceptical of supposedly educational creators handing over their platforms to companies like Waymo: where a non-sponsored video might have given us a complete and balanced insight into how Waymo’s vehicles work, Veritasium’s deal with the company sees them leaving out vital information in order to make their sponsor look good. Geographical limitations (and a failure to mention them) again become relevant in the penultimate section of the video.
Prior to this, there is another vlog section—the hilarious car park incident mentioned earlier—and a discussion of the ethical questions surrounding whose life a driverless car should choose to save should a vehicle find itself forced to make a choice between hitting two people. Muller’s argument here is incredibly weak, consisting of him just trying to reassure us that such events are rare and that we therefore shouldn’t lose too much sleep over it. He tells us that ‘the reality is that 99% of accidents aren’t like that’. I hope it’s sufficient to say in response that rare things still happen. In 2015, The Economist reported that the odds of an aeroplane crashing were one in 5.4 million,
but, as a society, we still dedicate significant resources to try to stop that from happening. Where geography once again becomes an issue is when Muller returns to trying to convince us that Waymo’s software is—to borrow the phrase the company loves so much they’ve trademarked—’the world’s most experienced driver’. He does do partly through continuing his cringeworthily soft-ball interview with the head of Waymo’s ride-hailing operations. Perhaps the lowest moment in this interview sees Muller make the following suggestion: ‘if that’s true, it means, like, every vehicle that’s not on the road is, kind of, a worse situation. Do you know what I mean?’. But he also throws a couple more statistics at us.
Muller reports that ‘these vehicles have way more experience than any human driver, because they’ve now accumulated data over 20 million miles of driving on public roads. If you were an average driver, you’d have to drive for 1000 years to accumulate that sort of experience’. Waymo loves to quote this statistic in their promotional literature. What they seem less keen to reveal is exactly where these 20 million miles of driving took place. For, throughout the Veritasium video, Muller simply refuses to mention that the reason that he filmed this video in the suburbs of Phoenix, Arizona is because that’s the only place in the world where Waymo’s taxi service operates.
While, in August 2021, the company began testing in San Francisco, a human safety driver is behind the wheel in every vehicle and prospective passengers are forced to sign a nondisclosure agreement. That’s not to say that the vehicles have only driven in these two cities; the Waymo Safety Report claims that the cars have driven in more than 25 cities across 10 US states. Nevertheless, the proportion of those 10 million miles that have been driven in different places is highly significant. See, as Mahmood Hikmet highlights in one of his videos about Waymo, the primary reason that the company is seeking to introduce its vehicles to new cities at the moment is because new environments bring new and unforeseen challenges. As Hikmet puts it, These vehicles are under a test and they’re going to fail, they’re going to run into edge cases. Some of them might seem really, really obvious.
But, it’s really hard to account for all of those edge cases during initial testing. So, what Waymo are hoping to achieve here in this public rollout is, they’re hoping to run into more of these edge cases. In short, letting the car drive up and down the same roads again and again eventually has diminishing returns—it might be amazing at driving on those particular roads, but still terrible elsewhere. We see this with Waymo’s vehicles in relation to weather. Phoenix has pretty favourable conditions for driving: it is generally clear and bright.
The Safety Report does stress that the vehicles have been tested in other conditions ‘from sunny Phoenix, Arizona to rainy Kirkland, Washington, across the snowy Upper Peninsula of Michigan, through Death Valley heat, and in foggy San Francisco’. Nevertheless, it would seem the company is still not entirely confident in the vehicles’ ability to operate in all weathers. A 2019 blog post by the company’s Chief Safety Officer revealed that Waymo’s taxis are ‘designed to come to a safe stop’ under certain bad weather conditions, and a Morning Brew article from July 2021 highlighted that certain conditions require the reintroduction of safety drivers behind the wheel. As with the statistic suggesting that 94% of accidents are the result of “driver error”, then, while the 20 million mile figure might appear to be a slam dunk in favour of allowing Waymo carte blanche to launch a taxi service in every city in the world, the truth is far more complicated. And, as we’ve come to expect from Veritasium in this video, none of those crucial questions are asked. Instead, as in the case of Waymo vehicles requiring highly-detailed, human verified maps of any road they drive on and that of the taxis being geofenced into a small area of Phoenix, the sponsorship deal with Waymo ensures that the video declines the opportunity to properly educate us about this technology in favour of leaving out any information which shows the company’s vehicles to be anything less than perfect.
Perhaps the most frustrating thing about the multitude of omissions and misrepresentations contained within this video is that admitting that the technology still has limits wouldn’t make it much less impressive. While Muller hams up his enjoyment of being in the Waymo vehicle a bit, I can’t imagine it would be anything but really cool to ride in one of these cars. They are an impressive feat of engineering. That being said, Muller rounds his video up by predicting some of the ways in which driverless cars might change the world. He tells us that transportation will get cheaper. Think of all the wasted value in the cars that spend over 95% of their time parked.
We can regain a bunch of time and feel happier because commuting and being stuck in traffic sucks. We can reduce traffic because vehicles will have better awareness of each other. As I bring this video to a close, as well as wrapping-up my points about the devastating effect corporate sponsorships have the potential to have on educational YouTube, I want to ask a few questions about whether driverless cars would really have the impact that Muller claims or whether there might be better alternatives to these problems already in existence. Throughout the videos sponsored by Waymo as part of their PR campaign—but particularly those by Veritasium, Snazzy Labs and Jason Silva—there’s an interesting assumption which goes completely unacknowledged. We find it in the title of Silva’s video: THE FUTURE OF AUTONOMOUS CARS IS HERE.
And, it’s also baked into the original title of Veritasium’s video: Driverless Cars Are Already Here. It’s the assumption that self-driving cars are the future of transport, with the only uncertainty being when they will become a viable option for replacing human-driven cars as the dominant mode of transit in the Global North. In his 2018 book Winners Take All, Anand Giridharadas suggests that, in recent years, technology companies have often used the circulation of such limited visions of the future as a key rhetorical tactic.
He writes that ‘in [Silicon] Valley, prediction has become a popular way of fighting for a particular future while claiming merely to be describing what has yet to occur’. The advantages of doing so are evident throughout this Veritasium video: by painting driverless cars as the only and inevitable future, you can portray anyone who is even mildly sceptical as simply a ratchety old fuddy duddy who is standing in the way of progress. Even assuming that the present flaws in autonomous vehicle technology can be ironed out (and there’s nothing to say they can’t be), it’s highly unlikely that they would solve many of the problems that Muller suggests they would.
The idea that driverless cars would reduce congestion is a complete non-starter; assuming people otherwise maintain the same habits, a car driven by a machine takes up the same amount of road space as one driven by a person. The suggestion that we could reclaim the “wasted value” residing in the 95% of time that the average car spends parked is another excellent example of lying with statistics: it ignores the fact that much of that 95% is the nighttime, or the middle of the day when people are working; for lots of people, the 5% of the time in which they need their cars is the same 5% of the time that everyone else is using their cars. Finally, whether autonomous cars would make getting around cheaper is just complete guesswork; whilst the lack of a driver might reduce costs for taxi companies such as Waymo, there’s nothing to say that they would pass those savings onto customers. The argument that Veritasium makes for driverless cars here is, then, pretty poor.
If we take off the Waymo-sponsored blinkers which consider autonomous vehicles to be the only possible future of transportation for a moment, however, we find that the problems highlighted make a fantastic argument for another solution: public transport. These problems of congestion, of the “wasted value” both of the time that people spend driving and of everyone having their own personal vehicle that they only use 5% of the time and, finally, of cost; these are all problems which would be much better addressed by well-funded, publicly-owned, integrated mass transit systems. The vehicles could even be autonomous if you like.
That I couldn’t find a single sponsored video similar to those paid for by Waymo but making the case for the introduction of more usable public transport points us towards the larger problem that the existence of videos like this one by Veritasium highlights. On the one hand, what we have here is simply a bad educational video. At every opportunity, Veritasium chooses making Waymo look good over actually teaching us anything useful about driverless cars. In fact, in its multitude misrepresentations and omissions of vital details, I think it’s more than fair to call this misinformation.
As I said at the very beginning of the video, it is a perfect example of why we should be highly sceptical of creators who are willing to hand over their platforms to corporate sponsors in this way. Nevertheless, it’s also worth thinking about what effect these practices—particularly by supposedly educational channels—might have on our public discourse more broadly. Rightly or wrongly, YouTube has become a platform to which people flock not only for entertainment but for education and to seek out potential responses to the various social and political challenges facing our world. As is the case with this set of videos sponsored by Waymo, companies, corporate lobby groups and billionaires are increasingly using sponsorship deals to ensure that the answers people find on the platform are amenable to their interests. Actual solutions which serve everyone—such as, in this case, public transport—tend not to have anyone willing to bankroll sponsored videos on their behalf.
We might not be able to change that, but we can at least become attuned to recognising and critiquing these corporate propaganda pieces. Thank you so much for watching this video, I hope it’s been worthy of your time! If you have got anything out of it at all, then I’d be incredibly grateful if you’d consider sharing it with a friend (either online or off) who you think might also get something out of it. Thanks as ever to Richard, Sindre Nilsen, Kaya Lau, David Brothers, Max DeVos, Allan Gann, Luke Meyer, Gary, Dylan Gordon, Diccon Spain, Greg Miller, Bill Mitchell, Al Sweigart, Z.C. Reese, Brent Cottle, Shab Kumar, Anil, Alexander Blank, Niels Abildgaard, Sophia
R, President Dwayne Elizondo Mountain Dew Herbert Camacho, Sergio Suarez, TwoBr0TwoB, Teddy Zamborsky and Alexandra McGinnis Wortendyke for being signed up to the top tier of my Patreon. If you’d like to join them in getting early access to videos, copies of the scripts to them and more, then you can find out how to do so at patreon.com/tomnicholas. Thanks again for watching and have a great week!
2021-10-23