Recently, Morgan Stanley analyst Adam Jonas shared his experience with Tesla’s FSD (Supervised) V12.3.6 and his thoughts on the system. He was impressed with the advanced driver-assist system, noting that it exhibited both caution and assertiveness when required. Jonas observed significant improvements from FSD (Supervised) V12.3.4 to V12.3.6, with expectations of further substantial enhancement with the release of FSD (Supervised) V12.4. He mentioned that version V12.4 eliminates most training compute limitations, leading to a more robust system.
Additionally, Jonas described a scenario where cars could learn by learning instead of labeling. His initial impressions of FSD (Supervised) V12.3.6 were shared on Twitter, highlighting its heightened sensitivity to pedestrians and associated elements. In a detailed account, Jonas provided an example of how a robot following natural language commands demonstrated its understanding of the task, embodying the essence of the system’s capabilities. He stressed the robot’s ability to discern and execute the command without explicit prior instructions, demonstrating a high level of learning and adaptation.
Furthermore, Jonas estimated that Tesla is gathering a significant amount of data from its fleet and predicted that by late 2025, the company’s vehicle fleet should collectively cover a 100 billion miles per year run rate. Overall, Jonas’ experience with FSD (Supervised) V12.3.6 reflected his positive outlook on the advancements in Tesla’s autonomous driving technology.