Tesla's Self-Driving Cars Still Need More Training

Tesla's boss Elon Musk has promised for years that his company's cars will drive themselves. But workers who help train Tesla's driving computers say the technology still makes dangerous mistakes. These employees spend their days watching videos of Tesla cars hitting animals, speeding through traffic, and coming too close to children in the street.

The workers are called "data labelers" and they work in an office in Utah. Their job is to teach Tesla's Full Self-Driving software what good and bad driving looks like. They watch thousands of videos every day showing Tesla cars making mistakes on the road. Many of these workers say they would not trust the technology to drive them around.

Tesla is worth over $1.6 trillion because investors believe the company will create self-driving cars before other companies do. But the reality inside Tesla is very different from what Musk tells the public. The self-driving software still struggles with basic driving situations that human drivers handle easily.

Tesla has always said its self-driving software is much safer than human drivers. The company points to data that seems to prove this claim. But traffic safety experts who looked at Tesla's numbers say the comparisons are misleading. They don't meet the standards that scientists expect for good safety research.