One of the more intriguing developments in the autonomous cars space over the last few months has been the Baidu initiative, Apollo. Apollo is an open platform designed to facilitate collaboration between different companies operating in the autonomous hardware and software space. Its central aim is to accelerate the development of autonomous vehicle technology by allowing its partner companies to share IP and collected data. The model is simple; the more data you contribute to the platform, the greater your access to the data, services and algorithm models stored on the platform.
Datasets play an essential role in refining machine learning algorithms. At a basic level, the larger the dataset, the more the algorithm has to learn from, and thus the more polished and robust that algorithm becomes. As suggested in our piece, The Race to Full Autonomy, Waymo and Tesla are leagues ahead of competition on the data collection front. This is particularly true in terms of real-world data.
Collecting data takes time. It took Waymo seven months of testing to complete their third million miles of on-road, autonomous driving, completed by May last year. By comparison, the company’s first million miles took almost six years to complete! And this is a team with significant resources.
So how can the smaller companies compete? Apollo presents a solution; it allows a company to share its data on the platform and, in return, leverage the data made available by partner companies. Multiple companies collecting data instead of just one. This gives those companies who are unable to commit the necessary resources an opportunity to compete with the incumbents.
To get the ball rolling, Baidu have elected to share their software stack for free. This decision came after internal modelling estimated that the software would only generate an IRR in the high single-digits if customers were charged. Instead, Baidu’s business model centres around monetising the data generated by their autonomous vehicles.
One of the fundamental differences between autonomous vehicles and the cars of today, is that when a car is sold today, the manufacturer loses communication with the vehicle until it is returned for repairs or maintenance. They don’t know where it is or what it is being used for. Autonomous cars, on the other hand, will be connected, producing immense amounts of data. Baidu plans to monetise this data by offering over-the-top services such as cloud, emergency and security systems. A report by McKinsey & Company estimates that the combined value pool from car data and shared mobility could equal more than US$1.5 trillion. This model is perfect for the Chinese market in which Baidu operates, as there is less of a cultural resistance to the idea of personal privacy.
Apollo boasts 50 partner companies on the platform already, including heavyweights like Bosch, Microsoft, Velodyne, Nvidia and Delphi. Note that these aren’t exclusive partnerships, and don’t represent a complete sharing of resources. The platform has already fostered some interesting projects. As reported by TechCrunch, US start up AutonomouStuff created two autonomous cars on Apollo’s 1.0 software release in just 3 days! This is just one example of the enormous potential offered by the platform. In creating Apollo, we believe Baidu has positioned itself as a serious contender in the race to full autonomy.