Tesla, known for its innovation in electric vehicles and autonomous driving, has taken an interesting direction with its autopilot system. In an attempt to make cars drive more like humans, the company has trained its artificial intelligence to deliberately ignore certain traffic signals. According to employee statements, the AI in Tesla vehicles is designed to adapt to various driving situations, which includes bypassing rules like "no right turn on red" or "no U-turns" under specific circumstances.
Training in Data Centers
Tesla has centralized the training of its AI at facilities such as the plant in Buffalo, New York. In these centers, employees analyze thousands of hours of videos recorded by Tesla vehicles worldwide. The process involves labeling the situations captured by the cameras, allowing the AI to learn how to interpret and respond to different driving scenarios. According to several former employees, the idea is to "teach" the system to behave like a human driver, who sometimes makes decisions based on experience rather than strictly following rules.
However, this approach has sparked controversy. Seven employees revealed that they were instructed to ignore certain traffic signals, raising questions about the safety and ethics of this training. In situations like traffic jams, for example, some Teslas have been observed mimicking human behavior by crossing into prohibited lanes to move faster.
Privacy and Monitoring Issues
In addition to training the system, Tesla has faced criticism over the handling of data captured by its vehicles’ cameras. In 2022, private images recorded by the cars were leaked, sparking concerns over the privacy of owners. These incidents led the company to implement watermarks on the videos to track down the employees responsible for the leaks.
Tesla has also been accused of closely monitoring its workers, which led to tensions at the Buffalo plant, where employees launched a union campaign.
