Tesla FSD v12 Rollout - Initial Impressions and Challenges

What approach has Tesla transitioned to for full self-driving?

—Tesla has transitioned from rule-based coding to an end-to-end neural net approach for full self-driving.
What challenges does Tesla face in ensuring safety?

—Tesla faces challenges in ensuring safety and improving driving abilities with their new approach.
What is the main focus of the video?

—The main focus of the video is Tesla's transition to an end-to-end neural net approach for full self-driving.
What are the benefits of Tesla's new approach?

—The benefits of Tesla's new approach include improved driving abilities and potential for full self-driving capabilities.
What are the potential drawbacks of Tesla's new approach?

—Potential drawbacks of Tesla's new approach include challenges in ensuring safety and reliability.

🚗 Tesla's shift to an end-to-end neuronet approach for FSD version 12 is a monumental rewrite of their full self-driving software, a first at this scale or ambition.
🚗 Tesla's shift from heuristic code to neuron nets for driving is impressive and groundbreaking.
🚗 Tesla's transition from rule-based coding to neural nets for full self-driving is encountering challenges and slower than expected, leading to potential unguarded situations.
🚗 Tesla's decision to use end-to-end neural nets for FSD is a monumental task that requires massive amounts of data and compute power.
🚗 Removing thousands of safety rules for FSD rollout raises concerns about the true driving ability of the AI.
🚗 FSD version 12 needs to figure out driving from millions of video clips without hardcoded rules, a seemingly impossible task.
🛣️ The challenge is to increase FSD version 12's driving ability so it rarely ever drops below the threshold of confidence.
🚗 Tesla's FSD team is laser focused on improving the neural Nets to the point where it's confident in practically every driving situation, even complex ones.

#Tesla #FSD #DaveLee @heydave7

00:00 🚗 Tesla rolled out FSD version 12 software to non-employees, marking a monumental shift to an end-to-end neural net approach for full self-driving.
01:16 🚗 Tesla has shifted from rule-based coding to a neuronet end-to-end approach for driving, with perception code now running on neuron nets and planning being the next big part of driving.
01:48 🚗 Tesla is transitioning from rule-based coding to neural nets for full self-driving, but it's challenging due to the need for rules to prevent dangerous situations.
02:49 🚗 Tesla is using neural nets for complete planning in their FSD v12 rollout, with promising early results and a monumental task of building and training a massive neuron net with the right data.
03:56 🚗 Tesla's FSD v12 rollout is not safe for customers as it removes thousands of safety rules, leaving the true driving ability of the AI in question.
04:25 🚗 Tesla FSD v12 needs to make a huge leap in ability to drive without hardcoded rules, and it will take time to fix current weaknesses and achieve unsupervised FSD and robotaxi.
05:35 🚗 FSD v12 handles situations well but can stop and hand control to human if confidence level drops, aiming to increase driving ability to avoid low confidence situations.
06:29 🚗 Tesla FSD v12 needs to dramatically improve its driving abilities by feeding more data and improving neural nets to confidently handle all driving situations.

------------------------------------- 0:7:31 2024-01-23T01:30:04Z