Rivian recently offered a glimpse into its ambitious plans for AI-powered self-driving technology at its "Autonomy & AI Day." While the event showcased significant strides, it also underscored the complex challenges inherent in achieving true vehicle autonomy, a point humorously highlighted by a minor mishap involving a cafeteria robot that flashed "I'm stuck" on its screen.
The demonstration of Rivian's new "Large Driving Model" (LDM) in a 2025 R1S SUV provided a firsthand look at its capabilities. During a 15-minute drive on a winding route near the company's campus, the EV, equipped with the automated driving software, navigated turns, stopped at traffic lights, and slowed for speed bumps without relying on pre-programmed rules. This end-to-end AI approach, similar to how Tesla developed its Full Self-Driving (Supervised) system, marks a significant departure for Rivian.
However, the demo wasn't without its imperfections. The vehicle braked hard when detecting a Tesla Model S turning ahead, requiring a near-intervention from the safety driver. There was also one actual disengagement when the driver took over in a single-lane section due to tree-trimming. Such disengagements were observed in other demo rides as well, reminding observers that while promising, the software is still under development and not yet ready for widespread deployment.
A Strategic Shift to AI in 2021
Rivian CEO RJ Scaringe revealed that the company made a quiet but pivotal shift in its approach to autonomous driving in 2021. Previously, Rivian's driver assistance system was "deterministic" and "structured," with every vehicle action dictated by human-written control strategies. Recognizing the rapid advancements in transformer-based artificial intelligence, Scaringe "reconstituted the team and started with a clean sheet," redesigning their self-driving platform for an AI-centric world.
After extensive development, this new ground-up driving software was launched in 2024 on Rivian's second-generation R1 vehicles, which utilize Nvidia's Orin processors. Scaringe noted that dramatic progress only began to materialize "once the data started really pouring in," indicating the crucial role of real-world data in training their LDM.
Ambitious Rollout and Future Challenges
Rivian is betting on the rapid training of its Large Driving Model using fleet data to enable a swift rollout of new features. Later this month, the company plans to introduce "Universal Hands-Free," allowing Rivian owners to remove their hands from the wheel on 3.5 million miles of roads across the U.S. and Canada, provided visible lane lines are present. By the second half of 2026, Rivian aims to enable "point-to-point" driving, offering consumers the full experience demonstrated to journalists.
Looking further ahead, Rivian's smaller, more affordable R2 SUVs, expected to ship by the end of 2026, will feature a new custom autonomy computer and a lidar sensor, replacing the Nvidia chips. This advanced hardware is designed to eventually support "eyes-off" driving, where drivers can disengage from monitoring the road. True Level 4 autonomy, where a driver doesn't need to re-take control, will depend heavily on how quickly Rivian can train its LDM.
This aggressive rollout presents a near-term challenge: the custom autonomy computer and lidar won't be ready until months after the R2 goes on sale. Early R2 buyers will receive the "point-to-point" hands-off driving feature but will need to wait for the full "eyes-off" capabilities. Scaringe addressed this transparently, acknowledging that while not perfectly aligned, the company's honesty allows customers to make informed decisions, especially given the significant demand backlog for the R2.
Scaringe also revisited a long-held vision from a 2018 interview: a Rivian vehicle capable of meeting its owner at the end of a hiking trail. He believes this aspirational adventure-focused use case is achievable within the next few years, particularly as the LDM is trained on trickier roads without guiding features like lane lines, moving closer to Level 4 autonomy. While autonomous rock crawling up challenging terrains like Hell's Gate in Moab isn't a focus, getting to the trailhead autonomously certainly is.
