Teddy bears also fool the self-driving system of electric cars

This system is inherently equipped to ensure that while the vehicle is moving the driver is alert and aware of the dangers around.

Specifically, Dan O'Ward, the man behind the Dawn Project channel, in conjunction with YouTuber AI Addict conducted this test. The team tricked the system by placing a small weight on the electric car's steering wheel to make it look like the driver was resting his hand on it. Then, in turn, they put various objects such as teddy bears, giant beer bottles, balloons . on the driver's seat to pretend to be the driver, then activate the self-driving mode and see what happens.

First, with a large teddy bear in the driver's seat of the tram, the monitoring system didn't give any warnings and the car continued to drive on its own. The test team placed a small doll across the road to test the operation of the emergency braking system. But the car crashed directly into the doll because it did not have time to brake because the system took a while to recognize that there was a "person" crossing the road.

 

The second test was a unicorn and the driver monitoring system proved useless. As a result, the doll was also stabbed, even worse, did not stop at all.

The last test, even without any heavy objects being placed on the driver's seat, the car still ran. It crashed into the doll across the street, then stopped for a moment and then decided to keep moving.

Meanwhile, YouTuber Whole Mars Catalog also conducted another test and as a result, the driver monitoring system worked quite well. In this test, the driver was constantly looking at his phone, so the system issued a warning to pay attention to the road. Because the driver ignored this warning four times, the self-driving system was eventually disabled.

It is not clear whether the vehicle that Dawn Project and AI Addict used for testing had an update error that caused the driver monitoring system to malfunction. But it can be seen that this system is still not stable enough, so self-driving and driver monitoring still have a long way to go.