Morgan Stanley posts optimistic Tesla FSD expectations after V12.3.6 experience

Morgan Stanley analyst Adam Jonas recently shared some thoughts after personally experiencing FSD (Supervised) V12.3.6. Jonas seemed quite impressed with the advanced driver-assist system, noting that it becomes cautious or assertive when it should be. He also noted that the improvements from FSD (Supervised) V12.3.4 to V12.3.6 are quite evident.

Jonas noted that the version of FSD that he recently experienced included minor tweaks, though a more substantial improvement to the system is expected with the FSD (Supervised) V12.4 release. With version V12.4, most training compute limitations have been eliminated, the analyst noted. 

The Morgan Stanley analyst described a situation where cars could learn by learning instead of labeling. 

Morgan Stanley’s Adam Jonas shares his Tesla FSD (Supervised) v12.3.6 experience. His initial impressions:

• Cautious when it needs to be
• Assertive when it should be
• Seems to demonstrate a heightened sensitivity to pedestrians or things associated with pedestrians (i.e.… pic.twitter.com/dZtVti3BdT

— Sawyer Merritt (@SawyerMerritt) May 6, 2024

“Imagine you tell your robot to: ‘Pick up the red pen on the floor and balance it on one end on the table.’ And the robot picks up the pen and balances it on the table. The command and resulting action is possible due to the robot understanding natural language, knowing what a red pen is, what a table is and what it was asked to do with it. 

“Even if it never did this precise combination of movements before, it understood what was asked of it by observing and imitating permutations of such behaviors many, many times before – perhaps millions of times in reality and trillions of times in simulation. Even if it was never explicitly ‘told’ what a red pen was. It just figured it out,” Jonas wrote. 

Interestingly enough, the Morgan Stanley analyst provided a notable estimate of the data that Tesla is currently collecting from its fleet. As per the analyst, Tesla has about 5.5 million cars on the road today, and this number may grow to 25 million units by FY2030 and over 50 million units by 2035. 

Jonas also noted that Tesla’s vehicle fleet should be driving a collective 100 billion miles per year run rate by late 2025. Such a number translates to 11.6 million miles an hour or 3,200 miles per second, or almost 2 billion collective miles per week. Provided that FSD’s take rate does increase, a significant number of these miles could be driven without human input.

For context, Tesla noted during its Q1 2024 Update Letter that its FSD fleet has traveled a cumulative distance of 1.3 billion miles. Musk also noted in his Master Plan Part Deux in July 2016 that worldwide regulatory approval for autonomous cars may be possible with 6 billion cumulative miles. 

Don’t hesitate to contact us with news tips. Just send a message to simon@teslarati.com to give us a heads up.

The post Morgan Stanley posts optimistic Tesla FSD expectations after V12.3.6 experience appeared first on TESLARATI.