Bloomberg has just released an embarrassingly bad report about the self-driving space, in which it claimed Tesla has an advantage over Waymo by misrepresenting data.
There are currently many eyes on Tesla’s imminent launch of its “robotaxi” service in Austin, Texas.
We have just published a new report today to highlight how the launch is a game of smoke and mirrors, meant to reframe the optics of Tesla’s self-driving effort following years of broken promises.
At the same time, Bloomberg Intelligence released its own report, claiming that Tesla is ahead in self-driving technology, but the firm misrepresented data to support its claim.
The report compares Tesla’s and Waymo’s self-driving efforts, going so far as to claim that “Tesla is closer to vehicle autonomy than peers.”
Here are the two main charts that Bloomberg circulated from the report:
The problem is that the report is misleading by comparing completely different data.
Steve Man, the Bloomberg Intelligence analyst behind the report, based his report on Tesla’s own quarterly misleading “Autopilot Safety Report.”
The report is widely considered to be unserious for several main reasons:
- Tesla bundles all miles from its vehicles using Autopilot and FSD technology, which are considered level 2 ADAS systems that require driver attention at all times. Drivers consistently correct the systems to avoid accidents.
- Tesla Autopilot, which is standard on all Tesla vehicles, is primarily used on highways, where accidents occur at a significantly lower rate per mile compared to city driving.
- Tesla only counts events that deploy an airbag or a seat-belt pretensioner. Fender-benders, curb strikes, and many ADAS incidents never appear, keeping crash counts artificially low.
- Finally, Tesla’s handpicked data is compared to NHTSA’s much broader statistics that include all collision events, including minor fender benders.
All these facts combined render the comparison between Tesla’s accident rate using “Autopilot technology” and NHTSA’s US average completely useless.
Yet, Bloomberg decided not only to use it but also to compare it to Waymo’s data to claim that “Tesla is 10 times safer”:
The problem with this is similar to the comparison with the US average, as the Waymo data includes all police-reported incidents, which is a much wider net than Tesla’s data, in addition to the previously mentioned issues.
To highlight how big a potential discrepancy there is in the data, NHTSA underscored in a report last year how Tesla is not aware of many crashes involving Autopilot and that only 18% of police-reported crashes involve airbag deployment:
Gaps in Tesla’s telematic data create uncertainty regarding the actual rate at which vehicles operating with Autopilot engaged are involved in crashes. Tesla is not aware of every crash involving Autopilot even for severe crashes because of gaps in telematic reporting. Tesla receives telematic data from its vehicles, when appropriate cellular connectivity exists and the antenna is not damaged during a crash, that support both crash notification and aggregation of fleet vehicle mileage. Tesla largely receives data for crashes only with pyrotechnic deployment, which are a minority of police reported crashes. A review of NHTSA’s 2021 FARS and Crash Report Sampling System (CRSS) finds that only 18 percent of police-reported crashes include airbag deployments.
Knowing full well the comparison is not fair and completely misrepresents the situation, the usual Tesla stock pumpers on X widely shared Bloomberg’s misleading report positively, and even CEO Elon Musk shared the misleading data:
Electrek’s Take
This is embarrassing for Bloomberg. It’s such a blatant error and misrepresentation that it is suspicious. They should issue a correction right away.
Tesla fanboys are now pushing this to try to prove that Tesla’s robotaxi is safe to launch amid Tesla doing everything it can to hide its self-driving crash data ahead of the launch. This is a dangerous report from Bloomberg.
Additionally, it’s not just the primary claim regarding the accident rate that is misleading. The report also contains several glaring errors.
In this chart, Bloomberg claims that Tesla is at “3 billion miles of data collected since launched”:
It looks like they simply use Tesla’s “cumulative miles driven with FSD (Supervised)”, which includes driver supervision, and the driver remains responsible for correcting FSD at all times.
In comparison, they talk about 22 million miles for Waymo. It looks like Bloomberg only used Waymo’s rider-only mileage in San Francisco, which is currently at 22 million miles, but when accounting all markets, Waymo is currently at more than 71 million miles:
It’s not clear why they would only use mileage in San Francisco for Waymo when they used Tesla’s global customer FSD mileage for Tesla.
Again, these are also “rider-only” miles, which means that there are only people riding inside the Waymo vehicles, compared to Tesla’s mileage being completely supervised by customer-drivers at all times.
We simply don’t know how many “rider-only” miles Tesla has, since it only started with one or two cars in Austin over the last few weeks. It is likely to have no more than a few hundred or a few thousand miles.
Regardless, it’s completely nonsensical to claim that Tesla is “ahead of its peers” in self-driving, especially Waymo, based on this report.
Tesla is currently only trying to launch something that Waymo has been doing for years.
The other argument the report attempts to make is that Tesla’s “self-driving” vehicles are approximately 7 times cheaper than Waymo’s.
Again, the problem is that Tesla’s vehicles are not self-driving. Tesla has yet to prove that, and that’s why it is using “plenty of teleoperation” in this launch in Austin. Mapping, optimizing for geo-fenced area, and teleoperations are the real limiting factors here. Not the cost of the vehicles.
Suppose Tesla has anything less than a 100-to-1 vehicle-to-teleoperator ratio. In that case, its system is not profitably scalable and I wouldn’t be surprised if it has a 1-to-1 ratio for the foreseeable future – at least on its current hardware.
FTC: We use income earning auto affiliate links. More.
