One must understand that driverless vehicles will come sooner than one may think possibly by 2020-2022. Computers sitting in autonomous vehicles (AV) are essentially in the business of learning and improving on what a good human driver would do, writes Ajay Agraual in the new book, “Prediction Machine.” The more data they have, the better they become at predicting that the blur ahead is a pedestrian rather than sunlight reflecting off the road and reacting accordingly. The consequence is that the more miles under an AV test, the more unusual events (ie: a child biking on the road) the system faces and learns.
Such lessons once learned by computers are not forgotten and can be drawn upon by every vehicle using the same software.
AVs never fall asleep at the wheel or pull their eyes from the road to check their phone; this would make driverless cars much safer then human driven ones, which contribute to roughly 1.25 million road deaths each year worldwide. (The Economist, June 9th, 2018).
Google has tested driverless cars the longest through WAYMO which has reported in 2017 three collisions in 350,000 miles of driving in California while GM had 22 accidents in 132,000 miles. Neither was involved in a fatal accident such as TESLA (4 fatalities) and UBER (2 fatalities).
A recent blog posted by WAYMO mentions how, “with machine learning, we can navigate nuanced and difficult situations maneuvering construction zones, yielding to emergency vehicles and giving room to cars that are parallel parking. Our image processing sensors use machine learning to filter weather destinations, like snow and rain.”
The best software such as the one at WAYMO, could create a monopoly situation where in the future only the best software would be certified by governmental agencies as the only acceptable one.