Tesla CEO Elon Musk tweeted a few hours ago that Tesla would open up the Full Self-Driving (FSD) Beta to the masses via an option on Tesla vehicle touchscreens in ~10 days. (Well, by “masses,” I mean people who have bought the FSD upgrade, which currently costs $10,000.) It’s unclear if Tesla will take away the “Download Beta” option once a certain number of people have activated it in their cars (to limit/slow the expansion a bit), but the way the tweet reads is that anyone with the FSD package will get access to the updated firmware and Tesla will not limit who can download it.
Due to high levels of demand for FSD Beta, adding “Download Beta” button to Service section of car display in ~10 days https://t.co/D6M3ZiMarG
— Elon Musk (@elonmusk) March 6, 2021
When I wrote yesterday about Tesla increasing access to the FSD Beta by ~10× in a few weeks, one commenter noted, “Often lost in discussions of FSD is how to evaluate improvements in safety vs the average human driver. While it’s obvious that perfect FSD (whatever that means) may be years away, the safety advantages of even the current betas are significant compared to humans.”
Tesla does publish a quarterly “Vehicle Safety Report,” which shows number of miles driven between accidents for Tesla cars and SUVs with Autopilot engaged, for Tesla cars and SUVs without Autopilot engaged but with active safety features, for Tesla cars and SUVs without either of those, and for the US vehicle fleet as a whole. While that is fun to look at and to consider, there are some statistical issues that make any conclusions quite limited. For example, most Autopilot miles are done on the highway and in places where Autopilot works, which are fundamentally easier places to drive. Additionally, comparing to the US fleet as a whole is not the best comparison — it would be better to look at data from other premium-class cars of the same average age. There is surely a difference in accident rates between a 3 year old Mercedes E-Class and a 20 year old Toyota Corolla. I don’t know how those accident rates differ, but I know that I’d rather see Tesla’s stats compared to the E-Class’s.
One more factor: Comparing rate of accidents with Autopilot engaged versus not engaged is interesting, but Autopilot will disengage when faced with sudden difficult scenarios. I’ve had that happen to me a handful of times in the past year and a half. If Autopilot disengages one, two, or three seconds before an accident, does that count as Autopilot not being engaged? I don’t know and I’ve never even seen the question posed to Elon or to anyone else at Tesla.
All of that said, my experience with Autopilot is that it mostly increases safety and prevents accidents. I do think the rates of accident shown above look logical. And I do think that the FSD Beta will increase safety even further overall, despite certain ways in which it will inevitably decrease safety. If so, what are the implications of that improved safety? And when will be have proper scientific studies measuring Tesla FSD safety compared to alternatives in the same market?
First of all, there’s the issue of measurement — as the commenter I mentioned above, Baird Edmonds, pointed out. The limitations of Tesla’s quarterly safety report lead to it mostly being ignored, and I think that’s sensible without a better control group and without more details on how an accident is classified if it occurred a few seconds after Autopilot was automatically disengaged. (If Autopilot disengaged 3 seconds before an accident occurred because it could no longer handle the scenario, in my opinion, that should still count as an Autopilot-engaged accident since the driver had almost no time to adjust after the system disengaged itself. But I would assume it’s not currently classified that way.)
If Tesla can’t show more convincingly that drivers using FSD Beta have lower accident rates, then the implications are going to be more limited than they would be if Tesla could show that — even if FSD Beta does lead to lower accident rates.
If it can be shown that FSD users are definitely seeing lower accident rates, then this tech could be incentivized by governments, insurance companies could lower rates for cars using FSD, other automakers might be more convinced to license the firmware from Tesla (and put the necessary hardware in their vehicles), regulatory agencies could allow FSD in more regions/countries, and accidents could of course be significantly lowered. This could also lead more quickly to regulatory approval of true Level 4 or Level 5 fully autonomous driving, Tesla robotaxi service, and autonomous delivery services.
Naturally, another effect could be that more people buy Teslas and more Tesla buyers purchase the FSD option. And the more drivers have and use FSD, the more data Tesla receives, and the quicker Tesla can improve the Autopilot/FSD firmware and further increase safety.
When I bought our Tesla Model 3 in 2019, one of the two core reasons was for greater safety for my family. The other was that I expected the value of the car to increase as Tesla advanced significantly with semi-autonomous driving tech and eventual Level 4/5 autonomy. Those are still two core reasons I wouldn’t replace our Tesla with a non-Tesla vehicle. Though, it sure would be nice to get more data on the safety improvements from Tesla Autopilot and, soon, from Tesla FSD. Perhaps Tesla could partner with a university (ahem, Stanford) in order to get us some statistically significant and truly useful findings on the relative safety of a Tesla with FSD Beta versus another car in its class.
What are your thoughts on the implications of Tesla FSD Beta rolling out to tens of thousands of Tesla owners? In particular, what are your thoughts on what will happen if Tesla FSD Beta improves safety to a notable degree? And do you have any other ideas on how we could determine whether or not Tesla FSD Beta leads to lower accident rates and fewer severe accidents?
Want to buy a Tesla Model 3, Model Y, Model S, or Model X? Feel free to use my referral code to get some free Supercharging miles with your purchase: https://ts.la/zachary63404. You can also get a $100 discount on Tesla solar with that code. No pressure.
Full disclosure: I own stock in Tesla [NASDAQ:TSLA]. However, I offer no investment advice of any kind and I do not think my ownership of Tesla stock has influenced this article in any way.