Saturday, June 17, 2017

Lipson, Hod (2016): Driverless - Intelligent Cars and the Road Ahead

What is it about?

This book is about autonomous cars - their past, present and future.

However, the book not only discusses autonomous cars themselves, but to quite a notable degree covers the underlying technologies as well. Perhaps the most notable such examples are artificial intelligence and especially "deep learning", machine vision (with optical, laser and radio technology) and computer and software technology in general.

In addition, the authors very welcomingly discuss also the derivative social effects which would result from widespread adoption of autonomous cars (e.g. loss of certain jobs, changes in urban landscape, economies of urban, sub-urban and rural areas etc. Furthermore, the book does a good job in discussing some of the most obvious ethical questions such as how to value human life, because such a value is needed when an autonomous vehicle is in an emergency situation forced to choose between two or more courses of action all of which involve fatalities and/or damage to property.

Was it good?

The book is very interesting and reads quite well. At places, the authors get close to that stylistic line which irritates me in contemporary non-fiction books (overly colourful language, hyperbolic analogues etc.), but don't get there.

I especially appreciated the extended discussion on underlying or enabling technologies (e.g. how a neural network works, what affects its performance and how this have developed over the past decades, and how neural networks can be and are being employed in machine vision/sensing such as in autonomous vehicles), as this gives one a substantially deeper understanding about the current state and foreseeable future of autonomous cars.

Moreover, I equally liked the discussion concerning societal and ethical issues. This, actually, sets this book apart from may other contemporary non-fiction books especially on technical subjects, because the authors explicitly admit that there currently seems to be to much simplistic hype around autonomous vehicles.

The main take-away for me?

The main take-away for me certainly was an increased understanding about the technical complexity of making autonomous cars reliable and eventually "mainstream". For example, detecting a human progressing slowly while carrying a large dense object as human is very, very difficult to pull off with machine sensing. Yet, this must be routine with 99.9999% accuracy if autonomous vehicles are to become ubiquitous.

Who should read the book?

The book certainly requires some interest in the subject and an engineering mindset (neural networks, laser distance detection...), but anyone at all wondering about how advanced autonomous cars currently are and how (or whether) they become commonplace, should certainly read the book.

The book on  Driverless

No comments:

Post a Comment