This video features a test drive of Nissan's advanced autonomous driving technology, leveraging the power of Wayve's End-to-End Generative AI, which is likened to the impact of ChatGPT on information processing. Conducted in a Nissan Ariya prototype in complex urban Tokyo environments, the system demonstrates "veteran driver" competence, smoothly executing complex maneuvers, lane changes, and obstacle negotiation without human input (operating at Level 2 capability). The video highlights the revolutionary nature of the E2E AI, noting Wayve’s ability to adapt to Japan’s complex traffic using only 1% new data, built upon its UK foundation. The system, utilizing 11 cameras, 5 radar units, and LiDAR (crucial for high-speed night driving safety), is managed by a generative AI framework that handles its own training curriculum and evaluation scenarios. The host and a Nissan representative discuss the system’s "black box" nature, its astonishing development speed (collaboration started in Spring 2024), and the anticipated commercial launch in 2027 for high-end Nissan vehicles.
Nissan's Autonomous Leap: Experiencing the 2027 ProPILOT with Wayve End-to-End AI
0:00 Next-Generation ProPILOT Introduction: The host introduces a Nissan Ariya prototype equipped with a system using Wayve's End-to-End AI, comparing its revolutionary potential to ChatGPT. Commercialization is anticipated in 2027.
1:46 E2E AI Technology: The test drive begins with Mr. Iijima (Nissan's head of autonomous driving/ADAS) at the wheel, confirming the use of Wayve AI, a UK-based company. The prototype uses 11 cameras and 1 LiDAR.
3:55 Rapid Adaptation Capability: The AI demonstrated unprecedented learning speed, adapting to complex Japanese traffic environments in only four months. Nissan confirms that the AI only required approximately 1% of new, Japan-specific data, building upon 99% of its prior learning, showcasing the power of the E2E architecture.
5:36 Complex Urban Driving: The system is shown operating smoothly and strategically in heavy urban traffic (Shimbashi), handling junctions, obstacles, and heavy pedestrian/cyclist traffic without the safety driver touching the wheel.
8:51 E2E AI Architecture: Unlike conventional AI which processes Perception, Judgment, and Operation in stages (potentially losing significant data), the E2E AI operates as a holistic "black box," integrating all information for superior decision-making.
14:21 Strategic Maneuvering: The car intelligently anticipates upcoming navigation and strategically moves into the appropriate, less congested lane, demonstrating sophisticated strategic planning.
14:55 Commercial Timeline & Challenges: Commercial release is slated for 2027, primarily due to the time required for product packaging (compacting the unit), comprehensive thermal and vibration testing, and developing new safety verification protocols suitable for black-box AI systems.
17:45 Generative AI in Development: Wayve uses Generative AI to manage the entire educational process: creating the driving model, generating diverse test scenarios, and rigorously evaluating the system's performance, enabling growth with less real-world data than competitors.
23:28 Hardware and Safety Rationale: The system relies on 11 cameras, 5 millimeter-wave radar units, and 1 long-range LiDAR. LiDAR is included explicitly to maintain safety and visibility at high speeds (100–120 km/h) on unlit roads at night where cameras are compromised.
30:09 System Naming & Flexibility: Nissan refers to the system as "Chōba" (surpassing driving system). The host questions the AI's flexibility concerning driving slightly above the legal speed limit (a common human behavior). Nissan indicates initial programming must adhere to legal limits, but future designs may allow driver-selected optimal speed.
37:40 Technical Specifications: The Wayve AI uses approximately 10 billion parameters (compared to competitors using millions). Crucially, the system maintains a continuous operating frequency of 10Hz (10 Perception-Judgment-Operation cycles per second), significantly higher than typical standard AI operation (1Hz).
39:01 Key Takeaway: The technology is significantly advanced beyond current expectations, providing an experience that feels like Level 5 capability (though currently certified as Level 2). The system is expected to debut on high-end Nissan models like the Ariya or Elgrand in 2027.