Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

I tested FSD for a month. Why would I pay Tesla to train their AI?

This site may earn commission on affiliate links.
I recently went on a multi-state road trip in my Model Y. Decided to test what FSD is all about. I had never done this before in 4+ yrs of Tesla driving. So I purchased a one-month FSD subscription for $200.

On net, it worked rather decently on highway driving, only occasionally trying to suddenly trying to veer left into a turnaround on a Texas road at high-speed, despite navigation indicating continuation on the current route. On balance, the additional ease of driving, lower stress, statistically higher safety, etc. seem to be of some positive value when on a highway without construction. In contrast, driving through some construction zones, and for much of the city driving, FSD was rather janky, and did not provide me any net positive benefit.

But when I am home, which is most of the time, most of my driving is city driving.

So the question is: Why would one want to pay Tesla for FSD, to experience less-positive driving performance, take on personal risk for a loss if FSD messes up (driver opens a door in front of me, etc,. where I would have to pay the deductible, accept higher future premiums, and lose use of the car during the time it gets repaired), and also do free work for Tesla in training their AI?

I get why helping to train the AI is good for Tesla, and perhaps for TSLA stock price. But why would a typical Tesla owner want to shell out $200/mo + tax just to work for Tesla for free? Seems like Tesla ought to pay the best AI-teacher-drivers to help train their AI.
 
But why would a typical Tesla owner want to shell out $200/mo + tax just to work for Tesla for free?
Most wouldn't (or shouldn't). The ones that do (and are happy with it) are the ones that have bought into Elon's hype that the next release is 🔥🔥🔥 (which has never been the case, and likely won't for many years). And those that defend it are merely trying to justify the huge amount of money they spent on it.
 
I purchased it nearly 6 years ago when it was only $3k and that included a FSD computer upgrade (my car had the HW 2.5 at the time). I use FSDb all the time now. You quickly learn where it's good and where it needs work and use it accordingly. When it works well, it's really rather relaxing. I actually enjoy using it. And if I can send feedback to help make it better for me in the future, why not! Now if I just bought a new Tesla, would I pay the current price or $200 per month, absolutely not! Way too expensive for a L2 system. I would advise anybody to avoid purchasing it until it's at least L3. Until then, basic AP still does the majority of driving on the freeway for long trips and rush hour traffic.
 
  • Like
Reactions: A2be
But why would a typical Tesla owner want to shell out $200/mo + tax just to work for Tesla for free? Seems like Tesla ought to pay the best AI-teacher-drivers to help train their AI.

I really doubt people are paying monthly for FSD just out of the goodness of their heart to help Tesla... I've never even seen anyone claim that online.

FSD take rates (buying outright and subscription) aren't that high, so I think most Tesla owners have (correctly, IMO) assessed that FSD isn't a great deal. But for those that do have it there's plenty of plausible reasons:

- maybe they find it helpful and enjoy using it
- maybe they enjoy playing with new technology
- maybe their super rich and it's a negligible amount of money for them

In my opinion, FSD today is intentionally priced for low adoption. Tesla is playing the long game here and their intent isn't to squeeze as much money from the market today as possible (if it was FSD would be priced more like 1/5th of what it is now to dramatically push up take rates). They're trying to fund the ongoing development of FSD, without incurring more risk than they need to.

Seems like Tesla ought to pay the best AI-teacher-drivers to help train their AI.

Funnily enough all the cars out there without FSD are potentially contributing more to development. Assuming V12 "end to end" AI ends up being the future, they need a ton of driving video to train on. Non-FSD cars drive more manually so they'd contribute more (and yes as far as I know Tesla can source telemetry for FSD training from all cars in the fleet, if they want to).
 
Last edited:
  • Like
Reactions: RTPEV