Jesse Levinson, cofounder and chief technical officer of Amazon-backed robotaxi company Zoox, has publicly expressed concerns about Tesla’s Full Self-Driving (FSD) technology, stating that the AI-powered system sometimes makes him nervous during use. Speaking at the TechCrunch Disrupt conference in San Francisco, Levinson—who owns a Tesla himself—shared his firsthand observations of the driver assistance software that requires constant human supervision.
The criticism comes at a critical juncture for Tesla’s autonomous vehicle ambitions. CEO Elon Musk recently unveiled the Cybercab robotaxi concept, promising unsupervised fully autonomous vehicles in California and Texas by next year, with production beginning in 2026. FSD represents a cornerstone technology for these plans, using AI and machine learning to navigate roads with minimal human intervention.
Levinson’s primary concern centers on the unpredictability of Tesla’s FSD system. “I do find it a bit stressful because, again, usually it does the right thing and then it sort of lulls you into this false sense of complacency and then it does the wrong thing,” he explained. This observation highlights a critical challenge in AI-driven autonomous systems: maintaining driver vigilance when automation handles most driving tasks correctly.
The debate extends to fundamental technological approaches in autonomous driving AI. Levinson criticized Tesla’s reliance on camera-based vision systems, arguing that current AI technology cannot process camera feeds safely enough for fully autonomous operation. “We just don’t, as a society, have the technology that can take cameras, feed them into a computer, and drive as safely as a human. That doesn’t exist yet, and it’s not close to existing,” he stated.
Tesla’s FSD is currently under regulatory scrutiny, with the National Highway Traffic Safety Administration investigating four crashes involving the system, including one fatal pedestrian incident. Tesla’s website acknowledges that FSD systems “require active driver supervision and do not make the vehicle autonomous.”
The technological divide between competitors is significant. While Zoox and Waymo employ multi-sensor approaches combining LiDAR, radar, and cameras, Tesla relies primarily on cameras and AI neural networks. However, former Tesla AI director Andrej Karpathy defended Tesla’s approach, noting the company uses LiDAR during training phases, creating a more scalable deployment model that could prove advantageous long-term.
Zoox announced plans to begin testing its own AI-powered robotaxi service in San Francisco’s South of Market neighborhood and on the Las Vegas Strip within weeks, intensifying competition in the autonomous vehicle sector.
Key Quotes
I do find it a bit stressful because, again, usually it does the right thing and then it sort of lulls you into this false sense of complacency and then it does the wrong thing
Jesse Levinson, Zoox’s CTO and cofounder, described his experience using Tesla’s FSD system. This observation highlights a critical challenge in AI-assisted driving: the unpredictability that can lead to dangerous complacency among human supervisors.
We just don’t, as a society, have the technology that can take cameras, feed them into a computer, and drive as safely as a human. That doesn’t exist yet, and it’s not close to existing
Levinson challenged Tesla’s camera-only approach to autonomous driving AI, arguing that current computer vision technology is insufficient for safe fully autonomous operation. This statement directly contradicts Tesla’s technological strategy and timeline claims.
So our perspective is you really do need significantly more hardware than Tesla’s putting in their vehicles to build a robotaxi that’s not just as safe but especially safer than a human being
The Zoox CTO explained his company’s philosophy that robotaxis must exceed human safety standards, requiring more sophisticated sensor arrays than Tesla currently employs. This reflects Zoox’s belief that public acceptance of AI-driven vehicles demands superior safety performance.
Our Take
This confrontation reveals a defining moment for AI in autonomous vehicles: the tension between rapid deployment and cautious development. Tesla’s approach embodies the “move fast” Silicon Valley ethos, betting that AI neural networks trained on massive camera data can achieve safe autonomy. Zoox represents the conservative camp, arguing that current AI capabilities require supplementary hardware to compensate for algorithmic limitations.
The real test will be regulatory and public acceptance. Even if Tesla’s AI eventually achieves statistical safety parity with human drivers, high-profile failures could derail public trust and invite restrictive regulation. Conversely, if Tesla’s scalable approach succeeds, competitors investing heavily in LiDAR and multi-sensor systems may find themselves technologically obsolete.
What’s particularly notable is that even Tesla’s competitor acknowledges using their FSD system—suggesting the technology, while imperfect, represents meaningful progress in AI-driven autonomy. The question isn’t whether AI will power autonomous vehicles, but which technological pathway and safety threshold will define the industry’s future.
Why This Matters
This story illuminates the fundamental technological and philosophical debates shaping the future of AI-driven autonomous vehicles, an industry projected to revolutionize transportation and generate hundreds of billions in economic value. The disagreement between Zoox and Tesla represents more than corporate rivalry—it reflects competing visions for how AI systems should achieve safe autonomous driving.
The safety implications are profound. As AI systems take control of vehicles, the margin for error shrinks dramatically. Levinson’s observation about “false complacency” highlights a critical human-AI interaction challenge that extends beyond automotive applications to any domain where AI assists human decision-making. The regulatory scrutiny facing Tesla’s FSD, including fatal incidents, underscores the real-world consequences of deploying AI systems before they achieve sufficient reliability.
For the broader AI industry, this debate raises questions about the trade-offs between scalability and safety. Tesla’s camera-centric approach may be more economically viable and easier to deploy at scale, while multi-sensor systems like Zoox’s may offer superior safety but at higher costs. The outcome of this competition will influence how AI systems are designed, regulated, and deployed across industries, setting precedents for balancing innovation speed with public safety in AI development.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- Tesla Q1 Earnings Preview: What to Expect From Elon Musk’s EV Giant
- Larry Ellison’s Wealth Could Skyrocket Thanks to Tesla Stock and AI Boom
- Elon Musk Drops Lawsuit Against ChatGPT Maker OpenAI, No Explanation
- US to ban exports of car software to China and Russia
Source: https://www.businessinsider.com/tesla-full-self-driving-mode-fsd-zoox-robotaxi-amazon-2024-10