ZINE: Driving Headless into Autonomy

What is the true environmental impact of the intensive computation and data processing required to develop and use autonomous vehicles?

The tech and automotive industry pitch that autonomous vehicles deliver a safer, more efficient future than current cars is so pervasive that it has become an unquestioned touchstone for policymakers.

But what is the true environmental impact of the intensive computation and data processing required to develop and use autonomous vehicles?

When fleets of autonomous vehicle are trialled over countless miles to earn our trust, what is the scale of the computational demands and ongoing strains on digital infrastructure?

How does this compare with a human driver’s emissions? And, most importantly, how is this impact weighed against the ever-alluring goals of increased vehicle safety and driver convenience, promoted by tech firms and endorsed politically?

Let’s do a quick addition on just some of the hidden environmental costs of autonomous convenience:

-Energy and infrastructure needed to sustain network connectivity, transmit data, and render maps and offer computer vision across vast geographies. One of the most interesting and neglected dimensions of this is the subsea telecommunications cable network (see work by Anne Pasek (Trent University) and Hunter Vaughan (University of Colorado));

-Energy needed to run data centres (see work by Mél Hogan (University of Calgary) and Julia Rone (University of Cambridge));

-Water-intensive needs in order to build the various computer chips in both vehicles and data centres (see work by Janna Z Huang (University of California, Berkeley))

-Metal extraction costs of the elements needed to build electronic circuits, leading to destruction of tropical forests  (see work by Joana Varon (Coding Rights) and Camila Nobrega (Free University of Berlin));

-Carbon emission costs to get components from where they are extracted to where they will be used. Similarly, the absence of rights of repair drives waste and further extraction (see work by Steve Jackson (Cornell University), Maio Lu (The Chinese University of Hong Kong), and Jack Qui and Ziyi Wang (National University of Singapore)).

Now complicate the picture further.

What if our democratic requirements for autonomous vehicles don’t reduce these demands on the Environment, but instead compound them?

What if it turns out that data retention is necessary in order to investigate and determine issues of fault and liability when autonomous vehicles inevitably crash, and the uncertainty around applicable rules incentives a culture of ‘collecting everything’?

Some technologies, most particularly autonomous weapon systems, provoke a sharp normative question: should we build them?

A quick addition on the environmental consequences of autonomous vehicles suggests that if we want better futures, we should at least ask the same question.

Should we?

Helen Stamp, University of Western Australia, Julia Powles, University of Western Australia and Audrey Guinchard, University of Essex

This provocation is from the new ZINE: The Cost of Convenience