By Kirsten Rulf, Harvard Kennedy School Autonomous Vehicles Policy Initiative
When Tesla a few weeks ago experienced its second fatal crash with one of its cars in self-declared “Autopilot” mode, the company quickly pushed the blame on the driver’s inattentiveness. A reaction, which has caused consumer-safety advocates and autonomous vehicle experts to criticize Tesla sharply:
“This is another potential illustration of the mushy middle of automation,” Bryant Walker Smith, a University of South Carolina law professor who studies self-driving cars, said in an email interview with Bloomberg Technology.
It is not the first time that Tesla is snubbed for naming a feature in its Level-3 car “Autopilot”. In fact, sophisticated and misleading marketing by several AV companies around the technical capabilities of their semi-autonomous vehicles has been in the focus of the NHTSA for a long time.
Labeling a car feature evasively “Autopilot” poses not only an urgent safety problem for drivers and pedestrians, but increasingly for policy- and lawmakers, for insurance companies, and regulatory bodies. With companies fiercely guarding their work on autonomous vehicles, and often their engineering teams from competitors (just recollect the Uber vs. Waymo lawsuit), the only truly marketing-free data, that policymakers can base decisions on, is locked up in patent documents.
For my capstone project at the Harvard Kennedy School, I have used machine learning to first assemble and then in a second step reveal some of the critical and powerful information that patents, which companies have filed on crucial autonomous vehicle components, can disclose. Often the discovered data matches exactly some of the more urgent questions that policy- and lawmakers would pose to companies, if they had the chance.
Amongst other issues, the algorithms I have developed can deliver answers to these four key questions from investors and decision makers:
1) Who are the leading inventors in the Autonomous Vehicles space?
Close to 90 Percent of Policymakers and investors said in a Harvard Kennedy School survey that they want to know the leading talents in the AV space. I have elaborated on this first question and on the methodology in another blog post.
2) What is the speed of the technology’s development? How long does it take for an invention to go to market?
3) Which inventions are economically most valuable and will therefore be pushed more by companies?
4) What companies emerge as leaders in Autonomous Vehicles?
Of course, by far not every invention in the AV realm is filed as a patent. Many companies keep their AV work in trade secrets to protect it from public filing and the eyes of competitors. Besides, there is a time lag of 18 months between filing and publication, so the picture in the data set is not a real-time representation.
Nevertheless, marketing-free patent data helps to sharpen and complement the descriptions given by AV companies. It has the potential to become an enhancing factor in the policy and regulatory decision-making process. Moreover, with appropriate data visualization, it gives decision makers the chance to keep up with the rapid technological development that can currently be overwhelming. Lastly, with the right prediction algorithm, patent documents can become a leading (versus a lagging) indicator and predict future technology developments.
Using Machine Learning To Make Patent Data Accessible For The AV Regulation Process was originally published in Harvard Kennedy School Autonomous Vehicles Policy Initiative on Medium, where people are continuing the conversation by highlighting and responding to this story.