“It’s a demo or test using safety drivers – it is not [autonomous vehicle] District, “says Bryant Walker Smith, a professor of law at the University of South Carolina, who studies autonomous vehicles.
Bloopers and sensors
Tesla was quiet about the many details of her technology. And it is hard to draw specific conclusions about his technology from social media posts sent by riders. But some of these posts seem to show less than velvety rides. In one movie, robotaksi trying to turn left It seems to exceed the double yellow line to the oncoming movement. In another robotaksi Apparently does not detect the UPS truck stopping and reversing to parkingAnd the front seat safety monitor must intervene to stop the car.
One YouTuber sent a video depicting “Phantom braking” Robotaxi-Szybko stops without a clear reason-the phenomenon, which was also marked by hundreds of users of less advanced full self-government (supervised). examined by the federal government. Unlike the actual technology of self -conduct, full local government (supervised) requires users to look at the road.
According to Tesla’s website, the service stops in bad weather. One youtuber I had a downpour ride; Robotaksi dropped the rider in the Austin park when the wind began to surround himself. A few minutes later, according to videoThe same Robotaksi raised the creator to continue driving. However, denying the above, one poster he reported Cars occur “flawlessly” in weighty rain.
Experts say that early blocks of flats are not surprising. Full self -summary (supervised) requires an intervene from the human driver if necessary, and it seems that Robotaxi is the same now, says Philip Koopman, professor at the University of Carnegie Mellon, who studies the safety of autonomous vehicles. He says that the slip -ups that robotaxia made do not differ from what human drivers do on the road. But the value of autonomy is to be security, so it makes sense that movies – and “rough edges” of technology – annoy people.
Camera
The launch has recovered public debates on the basic principle of Tesla technology: its employ of cameras themselves to perceive and “make decisions” while driving. Musk and his company have long claim that artificial intelligence, supplemented with data collected by cameras, is sufficient to handle a protected car without a driver. The general director promised that all his cars could become autonomous without any modifications, with a straightforward push of updated software (although Tesla He also quietly gave up on this claim). Other companies perceive more pricey sensors, including Radar and Lidar, as essential verification and support. (Lidar price has dropped dramatically; many Chinese producers are now a sensor Every car they sell.)
Progress in large language models I convinced some in the automotive industry that Musk’s approach is appropriate. In an interview with a podcast published this week Kyle Vogt, former General CEO of General Motors AV Unit Cruise, he argued These paintings from many cameras mounted on vehicles and advanced models can be “really accurate”. (VOGT gave way to Cruise after one of the vehicles without a driver hit and dragged the pedestrian. The company was not limpid with the regulatory authorities about the incident, later a report was found.)
In the case of Cummings, Austin reports confirmed their beliefs that the cameras themselves are not enough for autonomously operating the car. “There is no robotic system that is crucial for safety – people can die [using it]”This has always been successful on one relief of the sensor,” he says. “It is not clear why Tesla thinks they can do what has never been done before.”
One record that can reveal Tesla’s internal success: how quickly it expands. Musk Boldly said in May that Tesla would have hundreds of thousands – and maybe to a million – autonomous vehicles next year. The company seems to be motivated. According to Publishing workTesla employs additional vehicle operators who are paid for driving cars around Austin to collect data. But of course Musk is not alien to unmet dates.