Part 4/10:
Edge cases—unexpected or rare driving situations—pose the largest hurdle for FSD. For instance, encountering a moose stepping onto the road or an ambulance running a red light are moments where human intuition and quick judgment are vital. These unpredictables demand a form of general intelligence that current AI systems are still developing to mimic.
Tesla's approach involves machine learning: real-world data collection through over 60,000 beta testers, who are encouraged to record and report difficult situations. This active learning mirrors how humans improve by experiencing and responding to complex scenarios. The idea is that through continuous exposure to challenging situations, Tesla’s FSD system can learn to navigate them effectively.