Tesla’s Full (supervised) self-driving may not actually be self-driving, but it’s still unique in the US even years after its initial launch. for now.
While advanced highway driver assistance systems will cost a dime in 2025, FSD Tesla It aims to address all the complexities of driving from point A to point B: traffic lights, roundabouts, highway intersections, four-way stops and everything in between. Whether it’s because Tesla CEO Elon Musk is convinced that self-driving personal cars are almost here, or because of the company’s notorious tolerance for risk (or a little of both), FSD is the only system like it in America.
But the competition is heating up to build the best car you don’t actually have to drive. in Rivian Autonomy and AI Day Earlier this month, company CEO and founder RJ Scaringe said point-to-point automated driving would be coming to Rivians sometime in 2026. (A different new feature, Universal Hands-Free, launched in December. This brings adaptive cruise control and lane keeping to 3.5 million miles of road.)
At the event, I got a roughly 20-minute ride in an R1S that was running unreleased driver-assist technology that previews future point-to-point software. The experience was serene in a good way. The drive was mostly smooth except for a few minor mishaps which I will describe.
It’s hard to rule out any real insights from this kind of controlled demo, unless things go very badly or spectacularly. However, the trip and chat with a Rivian self-driving engineer was a fascinating insight into the challenges of training a car to drive itself. Watch the video above to see the full drive, or read on for highlights.
Overall, things went well. The R1S stopped smoothly behind other cars, obeyed red lights, successfully executed turns and changed lanes like a human would. At the same time, nothing out of the ordinary happened that the Rivian had to react to. I probably felt a little more vibration in the steering wheel than I saw in Tesla’s FSD system when I used it recently. But again, this was a short test ride so it’s hard to draw conclusions.
The driver kept his hands off the steering wheel most of the time. The only unplanned interventions were to push the envelope when the regime was most hesitant. First, it happened after a speed bump, when the SUV inexplicably slowed to a crawl, possibly due to pedestrians on the nearby sidewalk. Later, the same thing happened at the green light. And again while turning right with some people walking around the nearest corner.
Nick Carlevaris-Bianco, senior director of perception on Rivian’s autonomy team, sat next to me in the back seat and explained how the system works and how to train it. What I’ve been testing is Rivian’s new “big driving model,” which the EV startup says works much like the big language models we’ve all come to know from chatbots like ChatGPT.
Rivian feeds the model huge amounts of real-world driving data, primarily from customers, and tries to make it replicate only good driving behavior. While previous generations of self-driving systems relied on several built-in rules (if you see a stop sign, stop), Rivian’s grand driving model has been trained end-to-end. It absorbs driving data on the one hand and outputs driving outputs on the other (with some regulated guardrails as well). This is also the approach Tesla is using for full self-driving.
The idea is to create a system that is more generalizable, one that you don’t have to explicitly teach to deal with every situation. For example, Carlevaris-Bianco said speed bumps were not something the team explicitly told the model about. Rivian only slowed down for those who were driving me because that’s what happened in the training data.
The Rivian R1S charges at Ionna Rechargery.
Photography: Iona
The same goes for traffic lights, which is interesting. He said there was no “explicit logic” in the system that said “if the traffic light is there and it’s red, don’t go.” The output depends on what the model decides.
This approach also brings challenges. Because humans are not perfect drivers, the raw data is full of bad habits. Carlevaris-Bianco said the team noticed early on that the model would blip the throttle on an open road. Rivian also had to practice the tendency to roll through stop signs and stop partially in crosswalks.
Especially for a feature that is still several months away from general use, Rivian’s software has performed admirably. Accounts of other drives at the event were largely positive as well, although some people experienced some hard braking and other disengagements.
Over time, Rivian says its self-driving system will get much better, especially once Lidar-equipped R2 crossovers It will launch in late 2026, providing better awareness and the ability to collect higher-quality data for LDM.
Rivian has big plans beyond point-to-point too, with ambitions to get into blind driving in some situations eventually followed by full Level 4 autonomy one day. For now, all I can say is that it looks like Rivian is off to a good start.
Contact the author: Tim.Levin@InsideEVs.com
