Tesla Full-Self Driving Takes me out for Coffee #tesla #fsd12 #selfdriving #modely

Показать описание

When Full Self-Driving (Beta) is engaged, the steering wheel icon is blue and a blue line represents your driving path.

FSD v12 introduces single-stack neural networks, trained on millions of clips, marking a significant shift from 300k+ lines of C++ code. This transition enhances urban driving capabilities and human-like driving behaviors, crucial for autonomy.

FSD v12, Tesla's latest version of its Full Self-Driving (FSD) software, is the first to use "end-to-end neural nets". This means that the vehicle's controls are handled by neural nets, rather than being coded by programmers. FSD v12 will also mainly use AI and cameras, instead of relying on a mix of cameras and sensors like radar.

An instantly notable improvement is how FSD Beta v12.1.2 handles urban driving scenarios. Smoother transitions and quicker response times contribute to a more comfortable driving experience, closely mirroring human driving behavior. This shouldn't come as a surprise, as the system now draws inspiration from videos of actual human drivers, rather than relying solely on engineer-written code.

The introduction of features like Park Seek, Automatic Set Speed Offset, and improved U-turn capabilities in v12 showcases Tesla's commitment to refining the driving experience and safety, essential for achieving full autonomy.

Almost everyone who tested the latest version of Tesla's self-driving software noticed the human-like behavior. The software drives smoothly and, intriguingly, finds creative solutions to complex traffic situations. This shows that the end-to-end AI starts to "understand" the world, keeping track of countless vehicles and pedestrians moving around and finding the best course of action like a human would. Consider it a ChatGPT moment, if you will, although many argue that large language models have no clue what they're talking about. They just arrange the words in the most likely order.