CyberDutchie
Active Member
How is an investor supposed to estimate the FSD timeline, especially considering that we are not on the team developing it and probably don't have experience with AI being used on anything remotely close to FSD in terms of breadth, scale and diversity?
I've been a long-time bull on Full Self Driving and have defended the technology often in the face of naysayers who claim it's not viable (ever) or will take decades to be safe enough to use. But as an investor, I would caution against thinking the latest beta release has moved the timetable for governmental approvals forward or that the market will suddenly "get it". In terms of the market, it will be a gradual and inconsistent factor in TSLA valuation with many fits and starts.
To have a better idea of what the timeline might be, I look to people who are much closer to the effort than I. Like Elon. From parsing his statements about FSD I gather the point at which it will mostly be able to drive itself is going to come quickly, much more quickly than most people suspect (and indeed, the latest demonstrations support this), but Elon has been hesitant to make predictions about government approvals and has been warning about the uncertainty of the "march of 9's" for years now. At some point the challenge of approval moves from a purely technical one to a political one.
My personal belief is that it would be criminal NOT to approve it as soon as it can be shown to be statistically significantly safer than humans even though a skilled and attentive human driver is capable of being better than a machine for perhaps a decade or more. Some of the very qualities that make the average human such a terrible driver overall are the same things that will allow an attentive driver with good eyesight and plenty of experience to outperform the best machines for years to come. Humans can be good drivers, they just don't pay enough attention because they become conditioned to believe they don't have to be hyper-attentive. The fact that a hyper-attentive and skilled human will still be better than a machine is not a reason to not approve FSD. The fact that humans are, statistically speaking, absolutely horrific drivers makes FSD approval a very low hurdle indeed.
First principles thinking says the acid test for FSD approval should be: Does it cause the death/accident rate to go up or down from where it would have been without FSD approval? And Tesla is in an enviable position in terms of the amount of data they will have at their disposal with which to make their case to the regulators. Regardless of when first principles thinking says it SHOULD be approved, Elon knows this is not a given. And while his actions and statements can and will AFFECT exactly when it is approved, ultimately, it's beyond his direct control because regulators don't necessarily use first principles thinking. Humans are not always logical and have many built-in biases. Hence his hedging on the timeline of FSD approval. Even though the market will recognize FSD value in fits and spurts in advance of approval, the valuation will also reflect that the bulk of the value is not unlocked until it's actually approved.
Even Elon doesn't know when regulators will actually approve it and it could drag on longer than one might think is reasonable. If it does drag on, previously recognized value of FSD (in the share price) will fluctuate. All of the above might seem obvious to some but it outlines a number of strong reasons why the best strategy is still "buy and hold".
I don't think we are seeing many challenging situations in any FSD videos yet. Here is a video of a Cruise presentation last year at Google Next that show clips of a few of these challenging situations in San Francisco--one of the more challenging urban environments you can find. Scroll to 3:40 in the video.
Pretty good that Cruise can handle situations with multiple cars blocking the lane, double- or triple-parked cars. Cruise received a permit from the California DMV just last week to remove the human backup drivers from their cars. This allows them to test level 4 autonomy (full self-driving under certain conditions, geo-fenced) on the streets of San Francisco.
Tesla of course has the advantage of much faster learning due to cars on the road, and has a good chance to be first in solving general full autonomy (vs. geo-fenced or other boundary conditions).
Btw. Reason for posting this here is that while to many of us it's clear Tesla is in the lead, the industry does not agree with that. The industry puts Cruise and Waymo in the lead with Tesla a challenger. See the 2019 Navigant Research autonomy leaderboard:
As for Tesla's data/real-world miles advantage: the counter argument from Cruise, NVIDIA, and others is that simulation can make up for a lot of real-world miles. This, plus the fact that Tesla FSD is still at level 3 (limited self-driving: vehicle will inform the driver when he or she must take over), is probably the main reason that you don't see analyst upgrades yet from this FSD beta release. I expect as soon as Tesla moves to level 4, the analysts are going to pay attention.