When will autonomous driving actually be reliable?

08 Tem 16:15

Yeah, totally get your frustration—it feels like we're stuck in limbo with these systems needing constant human backup. Do you think the main hurdle is the tech itself or more about getting people to actually trust it?

08 Tem 16:43

Last week, during my business trip to Phoenix, I had the chance to try out Waymo’s robotaxi service—under the cover of night and in light rain. The vehicle handled Phoenix’s complex urban areas with impressive skill, smoothly overtaking, making decisions at intersections, and even slowing down for a herd of cows. But at one point, it suddenly stopped in an actually empty lane and switched to “driver” mode. The “safety operator” in the back seat—who was essentially a human intervention system—took control of the wheel and manually guided the car forward. I can’t shake the doubt I felt in that moment: just how independent is this technology, really? When it comes to reliability, most current systems do boast high-accuracy perception and decision-making—especially when moving from Level 1 to Level 4 autonomy. But the weakest link is the “perception of distrust.” Without trust from users and society, the need for human intervention persists. When I spoke with users in Tesla’s FSD beta program, many expressed the same anxiety: “Something could go wrong at any moment.” This psychological barrier seems likely to lag far behind technological progress. In the near future—maybe within 3 to 5 years—we may achieve true independence in limited scenarios (like highway driving or closed campuses). But for the average user navigating daily city traffic, full autonomy might not be on the table for another decade, barring unexpected events like an asteroid strike.