Gary Zhou 
- 🤖 Morgan Stanley analyst Adam Jonas shared his experience with Tesla’s FSD (Supervised) V12.3.6, noting it behaves cautiously and assertively when appropriate.
- 📈 Jonas observed evident improvements from V12.3.4 to V12.3.6, with more substantial upgrades expected in V12.4 after training compute limitations are eliminated.
- 🚗 Tesla’s vehicle fleet is projected to reach 25 million units by 2030 and over 50 million by 2035, collectively driving 100 billion miles per year by late 2025.
- 📊 Tesla’s FSD fleet has already traveled a cumulative distance of 1.3 billion miles, and Elon Musk previously stated 6 billion miles could enable worldwide regulatory approval for autonomous cars.
- 🧠 Jonas described a scenario where cars could learn by imitating and observing permutations of behaviors, rather than explicit labeling.
The race towards fully autonomous driving has captivated the automotive industry, and Tesla’s relentless pursuit of its Full Self-Driving (FSD) technology has been a focal point of this pursuit. Recently, Morgan Stanley analyst Adam Jonas shared his firsthand experience with Tesla’s FSD (Supervised) V12.3.6, offering valuable insights into the progress and potential of this cutting-edge system.
Cautious and Assertive: A Balanced Approach
One of the standout observations made by Jonas was the system’s ability to strike a delicate balance between caution and assertiveness. He noted that FSD (Supervised) V12.3.6 behaved cautiously when the situation demanded it, yet remained assertive when appropriate. This balanced approach is crucial for an autonomous driving system, as it must navigate complex environments while prioritizing safety and efficiency.
Incremental Progress and Anticipation
Jonas highlighted the evident improvements from FSD (Supervised) V12.3.4 to V12.3.6, indicating Tesla’s dedication to continuous refinement and optimization. However, the anticipation for the upcoming V12.4 release is palpable, as Morgan Stanley expects it to bring more substantial upgrades. With training compute limitations set to be eliminated, the stage is set for a significant leap forward in the system’s capabilities.
A Rapidly Expanding Fleet
One of the most intriguing aspects of Tesla’s FSD development is the sheer scale of data collection and learning opportunities. Jonas estimates that Tesla’s vehicle fleet will reach a staggering 25 million units by 2030 and over 50 million by 2035. Collectively, these vehicles are projected to drive an astonishing 100 billion miles per year by late 2025, generating an unprecedented amount of real-world data for the FSD system to learn from.
Milestone Achievements and Musk’s Vision
Tesla’s FSD fleet has already achieved a remarkable milestone, having traveled a cumulative distance of 1.3 billion miles. This milestone holds significant importance, as Elon Musk previously stated that 6 billion cumulative miles could potentially enable worldwide regulatory approval for autonomous cars. With Tesla’s rapidly expanding fleet and continuous improvements, this ambitious goal seems increasingly attainable.
Learning Through Imitation and Observation
One of the most intriguing observations made by Jonas was his description of a scenario where cars could learn by imitating and observing permutations of behaviors, rather than relying solely on explicit labeling. This approach aligns with the principles of machine learning and could potentially accelerate the pace of progress, as the system learns from a vast array of real-world situations and adapts accordingly.
As Tesla’s FSD journey continues to unfold, the insights shared by Morgan Stanley’s Adam Jonas provide a glimpse into the remarkable progress being made and the potential for transformative change in the automotive industry. With a balanced approach, continuous refinement, and an ever-growing data pool, Tesla’s pursuit of fully autonomous driving is poised to shape the future of transportation.
