Watch What Happens When an FSD Beta-Equipped Tesla Spots an Ambulance at a Green Light

A 2023 Tesla Model Y running FSD Beta v11.4.4 shows just how predictable traffic could become if all cars were to be autonomous. The EV sits at an intersection, waiting for the green light with the ADAS turned on. The software has complete control over the car until the driver intervenes.

The traffic light turns green right about the same time an ambulance with sirens and emergency lights can be heard approaching. The cars on the oncoming lane make room for the rescue truck and allow it to become visible. The Model Y does not move.

Proving that the ADAS understood the situation correctly and acted faster than most people would have in such a scenario is a human driver in a Toyota who was waiting at the same traffic light. They proceeded forward once it turned green, even though the ambulance was getting closer and closer and was indicating turning left.

After the emergency vehicle cleared the path ahead, the Model Y continued with its route.

Despite this impeccable behavior, it's worth noting that FSD Beta is known for acting up at times. It has made quite a few erroneous decisions in the past, and it will most likely keep making them until the software can fully understand all kinds of traffic scenarios like or better than a human.

Even the out-of-beta V12 ADAS that was demoed live by Elon Musk on Twitter (X) attempted to run a red light because it misunderstood that different lanes had different traffic lights.

However, the footage below proves another important thing – Tesla fine-tuned the software to run on Hardware 4-equipped vehicles. That's the name given to the new computer installed on the brand's latest EVs.

Even though it's a bit behind other cars with Hardware 3 that got the v11.4.7 build of the ADAS, this is restoring confidence in FSD Beta's capabilities.

One could argue that it was necessary since the popular Texas-based marque started claiming its FSD-Beta-equipped cars "will be able to drive themselves almost anywhere."

The next stop for Tesla's ADAS is switching entirely to a neural net-based software with FSD V12. That means the system gets fed videos from customers who paid $15,000 or $12,000 for the suite.

They allow the automaker to use the footage to train FSD Beta. Instead of hardcoding every single scenario, the system sees what one or more people did right in a specific situation and emulates their behavior.

Tesla's minimalistic ADAS solution uses only cameras. Its cars don't even have ultrasonic sensors for parking, let alone radar or LiDAR. Other industry players use multiple hardware solutions to ensure the cars equipped with them can drive themselves.

Cruise and Waymo, for example, champion the self-driving movement. But others are coming to market with more ingenious solutions. Polestar, for example, will use Mobileye's suite from 2025. It encompasses a lot more than just cameras.

It's clearer than ever that Tesla will push forward with the vision-based FSD Beta, which could soon be just FSD if the V12 build goes out to everyone. It's just another promising product the brand might bring to market before anyone else.

But there are still some challenges ahead. Ford and General Motors have already deployed the BlueCruise and Super Cruise systems, which allow customers to take their hands off the wheel when navigating pre-approved roads (mainly highways). Mercedes-Benz also received California's approval for its SAE J3016 Level 3 Drive Pilot.

Tesla's FSD Beta is still considered a Level 2 ADAS, which is why it clearly warns users to never put too much trust in it because it could "do the wrong thing at the worst time."