Governing AI: The Importance of Environmentally Sustainable and Equitable Innovation

• Date: 01 November 2020

50 ELR 10888 ENVIRONMENTAL LAW REPORTER 112020

by Henry Gunther and Julietta Rose

GOVERNING AI: THE IMPORTANCE

OF ENVIRONMENTALLY

SUSTAINABLE AND EQUITABLE

INNOVATION

Julietta Rose graduated from Berkeley Law in May 2020 and plans to work in f‌inance,

cleantech, and renewable energy. Henry Gunther is a third-year law student at Washington

University in St. Louis and plans to work in environmental law and policy.

Articial intelligence (AI) and complex machine

learning algorithms have come to play a profound

role in many of our day-to-day activities. A quick

Google search on the role of AI in everyday life (itself a

complex algorithm that decides on the most relevant

result s)¹ will yield somewhere around 150 million results

in about 0.88 seconds. AI can unlock your smartphone,²

curate your social media feeds,³ and predict your online

purchases to improve shipping and delivery speeds. And

all of this happens without your awareness or enga gement.

We have nally landed squarely in the age of ubiqui-

tous computing—a stage of computer-society integration

rst predicted in 1988 by Mark Weiser at the Xerox Palo

Alto Research Center, in which computer systems would

“vanish into the background,” weaving “themselves into

the fabric of everyday life until they a re indistinguishable

fr om it.” Only a small fraction of total computer process-

ing power is actually in t he computers we use. Most is now

networked with billions of sensors that surround us —from

refrigerators to hair dryers, scales to garage door openers,

bikes to watches, and of course the smart home systems

that link all of your sma rt appliances together—providing

increasingly complete pictures of our everyday lives, act ivi-

ties, and propensities.

1. Google, How Search Algorithms Work, https://www.google.com/search/

howsearchworks/algorithms/ (last visited Sept. 15, 2020).

2. Apple, About Face ID Advanced Technolog y, https://support.apple.com/en-

us/HT208108 (last visited Sept. 15, 2020).

3. Paige Cooper, How the Facebook Algorithm Works in 2020 and How to Make

It Work for You, H, Jan. 27, 2020, https://blog.hootsuite.com/

facebook-algorithm/.

4. Alina Selyukh, Optimized Prime: How AI and Anticipation Power

Amazon’s 1-Hour Deliveries, NPR, Nov. 21, 2018, https://www.npr.

org/2018/11/21/660168325/optimized-prime-how-ai-and-anticipation-

power-amazons-1-hour-deliveries.

5. Stanford University, Ubiquitous Computing, https://web.stanford.edu/dept/

SUL/library/extra4/weiser/ubiq.html (last updated Sept. 15, 2020).

6. John R. Delaney & Angela Moscaritolo, What Is a Smart Home Hub (And

Do You Need One)?, PCM, July 13, 2020, https://www.pcmag.com/news/

what-is-a-smart-home-hub-and-do-you-need-one.

7. Matt Burgess, What Is the Internet of ings? WIRED Explains, W,

Feb. 16, 2018, https://www.wired.co.uk/article/internet-of-things-what-is-

explained-iot.

is fusion of digital technologies and blurring of the

human and digital boundaries is a new form of indus-

trialism. Simila r to all other industrial revolutions, these

advancements and rapid shifts in productivity are outpac-

ing our understanding of the potential costs and benets.

As AI becomes more prevalent in all areas of life, we need

to turn our attention to the interactions between AI a nd

our physical environment, to harness the potential of this

technolog y while avoid ing environmental and societa l

harms. Technological revolutions may fail to materialize,

but when they do, they may have unforeseen consequences

that leave us little time to prepare.

I. Def‌initions—Waves of AI

When we think about AI in relation to environmental

impacts, it is easy to recall the Volkswagen (V W) emis-

sions scandal and the “defeat devices” programmed to

allow their cars to cheat emi ssions tests. It is hard to forget

such an intentionally designed violation of the law. What

makes the V W emissions scandal so captivating is that

human programmers are easy to blame. We can almost

picture the engineers working in secret to design t he device

with m alintent.

However, the VW devices are far from the sophisti-

cated algorithms and machine learning systems t hat have

become common in the eld of AI. ese devices would

fall under the “rst wave” of AI. e truth is that the con-

sequences and potential harms from A I are far more com-

plicated than the V W scandal, and cu lpability for harm is

much more dicult to assign.

An example of the widespread unintended environmen-

tal impacts of algorith mic failure dates back to the study of

ozone depletion that led to the Montreal Protocol. Scien-

tists had designed algorithms to monitor and record ozone

8. Guilbert Gates et al., How Volkswagen’s “Defeat Devices” Worked, N.Y. T,

Mar. 16, 2017, https://www.nytimes.com/interactive/2015/business/inter-

national/vw-diesel-emissions-scandal-explained.html.

9. Andrew Tutt, An FDA for Algorithms, 69 A. L. R. 83 (2017), avail-

able at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2747994.

Copyright © 2020 Environmental Law Institute®, Washington, DC. Reprinted with permission from ELR®, http://www.eli.org, 1-800-433-5120.