is transportation editor with 10+ years of experience who covers EVs, public transportation, and aviation. His work has appeared in The New York Daily News and City & State.
Posts from this author will be added to your daily email digest and your homepage feed.
RJ Scaringe is sitting in Rivian’s Palo Alto offices, explaining why the adventure-themed EV company suddenly decided to build its own self-driving cars, when an unexpected guest glides by the window outside: Waymo.
A robotaxi from the Alphabet-owned company pulls up outside the office. The passenger, an analyst from Goldman Sachs, briefly takes a selfie before climbing inside. The Rivian founder and CEO chuckles at the scene.
“That’s amazing,” he laughs. “So perfect.”
The arrival of the Waymo helps clarify the challenge that lies ahead for Rivian. A few hours earlier, Scaringe was onstage in front of an audience of hundreds of investors, reporters, and influencers gathered for the company’s announcement of a huge, expensive, and undeniably risky bet on autonomy and AI. The goal, he said, is for Rivian to design its own AI chips that can help power higher levels of autonomy, eventually leading to Level 4 — no human supervision required within certain limits. Could Rivian achieve what has taken Waymo decades to accomplish, but on a much shorter timeline?
More importantly, can it do it safer and more effectively than Tesla, the other EV-only company that is attempting its own messy transformation into an AI and robotics company? Rivian was founded as the outdoor lover’s answer to Elon Musk’s company: rugged, off-road worthy, and fully electric. Now it seems like it’s trying to chase the mercurial billionaire down an AI rabbit hole.
Scaringe insists this isn’t anything motivated by Tesla, but a recognition that the auto industry is staring up its own cliff. Advances in transformer-based encoding and large-parameter models prompted a fundamental shift in Rivian’s thinking of the “physical AI” of autonomous driving, he says. As such, Rivian began a clean-sheet redesign of its autonomy platform in early 2022. With the introduction of its Gen 2 platform, the company began building a “data flywheel” to train a large driving model (sort of like an LLM, but for driving) using real-world driving data from its fleet. And because the model is trained end-to-end, Rivian says that improvements in sensors or compute directly enhance its capabilities, allowing the system to continuously improve as hardware advances.
Either Rivian pursues its own self-produced, vertically integrated autonomy strategy, or it risks being left in the dust by Waymo, Tesla, and others. Several times during the presentation, Scaringe notes that this isn’t some bandwagon chasing, but rather a move born out of years of hard work and thoughtful design.
“I spent some time last night with the team talking about this right before we’re about to show it today,” Scaringe says onstage. “And one of the lead engineers looked at me and said, ‘Boy, we’ve been working on this for years, and I haven’t been able to talk about it. It’s so cool. Tomorrow I can start to talk about what I do every day, all day long.’”
... continue reading