Tesla has gotten “Full Self Driving (Supervised)” approved in another European country. Following the Netherlands, Lithuania, Estonia, and Denmark, the driver-assist system is now allowed to be deployed in Belgium. Well, Tesla still has to pass some tests with it, and then it can be rolled out to customers, but that shouldn’t be a problem for the technology. The transport minister of the region of Flanders signed the approval on Wednesday, he announced on X (formerly Twitter). That’s basically the full news on this matter, but I published an article a couple of days ago about a Wall Street analyst claiming Full Self Driving (FSD) has “effectively achieved Level 4 autonomy (at least in most conditions).” There were many good comments from readers on the topic. However, this long comment from Steve Shaw does a superb job of characterizing the current state of FSD appropriately, so I’m taking the opportunity to share it more broadly here. “The analyst’s points are a great description of what is colloquially referred to as ‘L2++’: Driver assist that is often able (in good weather) to drive the car, BUT the human driver must remain alert, ready to intervene at any instant, sometimes without warning. Perhaps is close to L3 in limited conditions, such as a divided highway (in good weather). “The other points relate to whether Tesla is close to ‘safe enough for a robotaxi fleet’, in some cities. Maybe. A fleet is a controlled situation: we know from recent exposé that Tesla had ‘data labelers’ mark every important feature in Austin for months. This was then fed into AV training data, to increase ability of robotaxis to recognize what they need to recognize. I consider that a substitute for MAPPING a city: prior information about what is where. “It is true that Tesla’s general model means that data helps everywhere, not just in Austin. That’s likely part of why FSD (Supervised) in consumer cars has improved. And it should mean they can roll out robotaxi service faster over time. AFTER we see hundreds of cars making many trips per day in several Texas cities. So far, it is still unproven. “IMPORTANT: We’ll likely soon see boasting about how quickly Tesla has gone from a small number of robotaxi miles, to 10 million miles — WITHOUT a mention that most of those miles are from the 500 cars in SF area, that operate with a human driver. Registered as a normal (non robo-) human taxi service. Proves nothing. Each quarterly report and tweet continues this MISLEADING mingling of driverless and drivered ‘robotaxi’ miles! “HOWEVER, it is a long ways from robotaxi services to consumer L4. Too many variables once it is in people’s hands. I personally won’t trust an ‘almost L4’ that lacks LOCAL MAPS and RADAR (and/or LiDAR). No reason for an AV to be as blind as a person in bad weather! “If I was in government and had the authority, I would require all consumer self-driving modes to include direct reporting to the government agency of such incidents. That is, built into the dash, would be an option, after the car is stopped, for the person to describe what happened such that they had to intervene. If we had those reports, our varied opinions about what Tesla has achieved would not be needed. We’d know what Tesla knows, about how far they are from safe autonomy.” Well said.
