Get ready for a reality check on self-driving cars! Despite the hype, we're not there yet. The truth is, these vehicles still need human babysitters.
At a recent industry event, The Autonomous 2025 in Vienna, a former Navy pilot and robotics expert, Mary Cummings, dropped some hard truths. She argued that while self-driving cars are a reality, they're far from fully autonomous. Human supervisors are still required to step in when AI gets confused or makes mistakes.
But here's where it gets controversial... Cummings highlighted a fundamental issue: self-driving cars hallucinate. These hallucinations are statistical errors, causing the vehicles to 'see' things that aren't there. For instance, a Tesla incident on the San Francisco Bay Bridge Tunnel saw a car abruptly stop from 65 mph due to a hallucination, causing an eight-car pileup.
And this is the part most people miss: AI doesn't think like humans. It doesn't imagine or understand, it just follows patterns. So, relying solely on AI for safety cases is a risky move, Cummings warned.
Cummings reserved her strongest criticism for vision-only autonomous systems, stating they'll never work reliably. She emphasized that every self-driving car company needs human oversight, as no vehicle operates fully autonomously.
The industry agrees. TTTech Auto's CEO, Stefan Poledna, echoed Cummings' sentiments, saying some form of human involvement is currently unavoidable. He highlighted the challenges of remote supervision and the risks of offshoring remote operations staff.
So, where does that leave us? Well, the notion of fully self-driving cars is more of a dream than a reality for now. But there's hope. Poledna believes that with redundancy and improved technology, we can get closer to true autonomy. He envisions a future where human involvement is reduced gradually as the technology matures.
However, safety must be designed into these systems, not just claimed. As Cummings put it, safety is about layers of defense, and cutting corners can have fatal consequences. The industry needs to collaborate and share data, especially on AI hallucinations, to bring safer autonomous systems to market.
So, are we ready to embrace the future of self-driving cars? Or do you think we're still a long way off? Let's discuss in the comments!
