Who is best fit to build a self-driving car; Tesla or Google? My answer is Tesla, and the reason is their access to Data.
You have to ”teach” the car how to drive by itself, and for this you need data. Lot’s of it. Google doesn’t lack data, but search queries, gps coordinates and ad clicks simply doesn’t´ help a car understand a sudden bump in the road.
As of Q3 2016 ”(Tesla has) collected more than 1.3 billion miles of data from Autopilot-equipped vehicles operating under diverse road and weather conditions around the world.” The cars are constantly sending data to Tesla, data that is used to improve the Autopilot-algorithm. In May 2017 Tesla updated it’s data sharing policy, to be able to collect even more data (in this case video) from all Tesla cars already out there on the roads.
At this point in time Google has collected 3 million miles of data from cars in self-driving mode. The same number for Tesla, almost one year ago, was 222 million miles.
The thing is that driving conditions vary all the time. Roadsides differ, signs might be covered, the sun hits differently, snow, darkness, moving objects, etc. A safe drive in a self-driving car means the car has to recognise, understand and take the correct action 100% of the time. Even if a kangaroo suddenly appears on the road:
According to Morgan Stanley, Google’s self driving cars (now developed in the new company Waymo), is still the way of the future: ”Forget Tesla as the future of road transport, Google´s Way is the way to go, according to Morgan Stanley”. Let´s hope they refer to self-driving cars as a means of public transportation, and not to the safety aspects of self-driving cars. Personally I will leave it to people at Morgan Stanley to go with Google-cars, while I go for public transport powered by Tesla Autopilot.