Tesla's Neural Network is Receiving a Massive Amount of Data from Cars

Yesterday, courtesy of Reddit user kutrod, we received the first images that show the change since Tesla started gathering data from the 50,000 customer-owned vehicles around the U.S. — although the actual change in policy occurred last month. The manifestation of this is that the vehicles send the company photos from its cameras seemingly at random. Kutrod's image showed huge spikes in the amount of data a Tesla vehicle has uploaded since the beginning of May.

According to the company, Tesla's neural network is then applied to the massive collection of data, which will allow it to build a 3D virtual world of numerous cityscapes, as well as learn constantly and exponentially about real-world environments. This is pivotal for safety because it allows Tesla to get feedback from cars already in the hands of customers and apply this information to updates.

 

The decision marks a departure from the strategy of other companies, such as GM and Waymo, who are using test fleets to collect data before the vehicles are released onto the market.

Tesla asked for permission to use the recorded clips during an upgrade for Autopilot 2.0. Tesla stated in a message that accompanied the upgrade that, "In order to protect your privacy, we have ensured that there is no way to search our system for clips that are associated with a specific car.”