Page 143 - ITU Journal - ICT Discoveries - Volume 1, No. 2, December 2018 - Second special issue on Data for Good

,78 -2851$/ ,&7 'LVFRYHULHV 9RO 'HFHPEHU

TOWARDS A HUMAN ARTIFICIAL INTELLIGENCE FOR HUMAN DEVELOPMENT
Emmanuel Letouzé , Alex Pentland
2
1
2
1 Data-Pop Alliance, MIT Media Lab, and OPAL, MIT and Data-Pop Alliance, and OPAL

Abstract – This paper discusses the possibility of applying the key principles and tools of current artificial
intelligence (AI) to design future human systems in ways that could make them more efficient, fair,
responsive, and inclusive.

Keywords – Artificial intelligence, big data, human development, open algorithms, fourth industrial
revolution

principles to build a ‘Human AI” that reflects and
1. MOTIVATION serves the objectives and drivers of human
development in the data era [3].
The rise of “Big Data” over the past decade and the
more recent emergence of artificial intelligence (AI) 2. THE GIST AND “GOOD MAGIC” OF
have stirred many hopes and, increasingly, fears, CURRENT AI
about the fate of humankind in the “fourth
industrial revolution”. Are we heading towards By and large, the “Data Revolution” [4], big data [5],
brighter or darker times? Do big data and AI pose and current AIs runs on personal data emitted by
existential threats to democracy [1][2], or do they people using digital devices and services for their
offer the possibility of building a future where daily actions and interactions; yielding digital
decisions will be more rational, policies more signatures or “data breadcrumbs” in the forms of
efficient, processes fairer, politicians more cell phone records, bank transactions, web and
accountable? social media content, geolocation data, pictures,
videos, etc. These yield large data sets which can
As with past techno-political revolutions, the then be analyzed by algorithms to unveil patterns
trajectories experienced by different groups will and correlations and make estimations, projections,
primarily depend on the decisions made by humans, predictions, and prescriptions, among others.
the most powerful or those who, sadly, do not
typically care very much about how others are Most of us already rely on these tools to decide
affected. To avoid a dystopian future shaped by and which roads to drive on, articles to read, clothes to
for elites and machines, a growing number of buy, content to like, flights to book, or people to
citizens, organizations, and governments, feel a connect with. Our doctors will soon use the same
sense of urgency to act, but are unsure as to how. types of tools to diagnose cancer and suggest
treatment plans [6]. There are, and will be, actual
A first obstacle is a lack of clear understanding of “robots” running our factories, doing home chores,
what is really happening and looming with big data and entertaining our kids, but, generally, current
and AI. Another is a lack of long-term vision of how and future AIs are what were called big data a few
humans and machines may cooperate, and what the years ago: computational analytics models fed and
corresponding processes and ‘building blocks’ trained on large quantities of data crunched by
ought to be. Yet another hurdle is a lack of a clear machines (computers) to reach an objective and in
roadmap for mobilizing and coordinating scarce some cases power sophisticated machines (robots)
resources including human and technological, to implement decisions in a more or less
towards that end. A last barrier is personal agendas autonomous manner.
favoring a naïve embrace or systematic
fearmongering of all things AI. When a driverless car is on the road, the computer
that steers it looks at its surroundings, asking itself
In this paper, we aim to sketch an ambitious and whether what is in front of it is pavement or people
optimistic vision and offer some reflections on how and acts accordingly, in ways it was trained and
human societies could “leverage” AI, not just by taught itself to do through millions of past
using it but also by applying some of its key

‹ ,QWHUQDWLRQDO 7HOHFRPPXQLFDWLRQ 8QLRQ