Building the algorithm commons: Who discovered the algorithms that underpin computing in the modern enterprise?

• DOI: http://doi.org/10.1002/gsj.1393
• Author: Yash M. Sherry,Shuning Ge,Neil C. Thompson
• Published date: 01 February 2021
• Date: 01 February 2021

SPECIAL ISSUE ARTICLE

Building the algorithm commons: Who

discovered the algorithms that underpin

computing in the modern enterprise?

Neil C. Thompson

1

| Shuning Ge

2

| Yash M. Sherry

1

1

MIT, Cambridge, Massachusetts

2

University of Pennsylvania, Philadelphia, Pennsylvania

Correspondence

Neil C. Thompson, MIT, Cambridge,

MA 02139.

Email: neil_t@mit.edu

Funding information

Tides Foundation, Grant/Award Number:

1903-57432

Abstract

Research Summary: The reach of the modern enter-

prise relies on the power of information technology

(IT) tools such as sensors, databases, and machine learn-

ing. But tool improvements must be fueled by increased

computing power (e.g., faster hardware) or getting more

productivity from existing systems (e.g., through better

computer algorithms). New research has uncovered that

this second source, algorithm progress, is more impor-

tant than previously realized—sometimes orders of

magnitude more important than hardware—and thus

could be an important technological stepping-stone to

give competitive advantage to a country's firms. Analyz-

ing this “Algorithm Commons”reveals that the United

States has been the largest contributor to algorithm pro-

gress, with universities and large private labs (e.g., IBM)

leading the way, but that U.S. leadership has faded in

recent decades.

Managerial Summary: Companies are increasingly

tackling problems with big data and sophisticated analy-

sis techniques (e.g., Machine Learning). To meet the

increased computational demands of these approaches,

the capability of computers must improve. One important

Received: 16 June 2020 Revised: 29 August 2020 Accepted: 1 September 2020

DOI: 10.1002/gsj.1393

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and

reproduction in any medium, provided the original work is properly cited.

© 2020 The Authors. Global Strategy Journal published by John Wiley & Sons Ltd on behalf of Strategic Management Society.

Global Strategy Journal. 2021;11:17–33. wileyonlinelibrary.com/journal/gsj 17

technique for doing this is to redesign algorithms, the rec-

ipes that computers follow to perform calculations. A s a

result, firms that develop better algorithms, or get acce ss

to them first, can get important advantages over their

competitors. This paper shows that it is U.S. corporations

and U.S. universities that have produced most of the

important algorithm improvements, and thus suggests

that better algorithms may have been a source of advan-

tage for U.S. multinationals.

KEYWORDS

algorithms, computing, digital economy, information technology,

public goods

1|INTRODUCTION

Modern ente rprises rely s trongly on in formation technology to organize their data and analytics

(Brynjolfsson & McElheran, 2019), improve productivity (Brynjolfsson & Hitt, 1996; Thompson,

2017), and manage large multinational operations (Bloom, Sadun, & Van Reenen, 2012; Mauri &

Neiva de Figueiredo, 2017; McDermott, Mudambi, & Parente, 2013; Stallkamp & Schotter, 2019).

But staying at the forefront of IT requires exponential increases in computing power to feed new

functionality, big data, and analytical tools (Thompson, Greenewald, Lee, & Manso, 2020; Thomp-

son, Ge, & Manso, 2020; Thompson & Spanuth, forthcoming). Some of this increase comes from

improved computer chips (although less as Moore's Law slows (Thompson & Spanuth, forthcom-

ing)), but new evidence suggests that algorithmic improvement has made a larger contribution

than was previously realized (Sherry & Thompson, 2020).

Algorithms are one of the most important parts of computing, alongside hardware and soft-

ware (Leiserson et al., 2020). Algorithms organize computations, converting a goal (e.g., sorting

a list) into a series of steps that the computer will perform to reach that goal. Many algorithms

are so embedded into computer systems that users and programmers never notice them. For

example, a coder in a programming language like Python or R might call a standard function

(e.g., “sort”or “matrix multiplication”) without realizing that, behind the scenes, the designer

of that function built it using an algorithm that ensured that the calculation would be done effi-

ciently.

1

Despite going unseen, algorithmic improvements can have a transformative impact on

calculations, for example by taking a computation that was so computationally expensive that it

was impossible on even the largest supercomputers and making it tractable. This is, for exam-

ple, what quantum computing algorithms are expected to do with code-breaking (Ekert, 1997;

Roetteler & Svore, 2018).

Algorithm improvement is analogous to the economic concept of productivity improvement.

Just as a productivity improvement allows a firm to produce more output for a given set of

inputs, an algorithmic improvement allows it to tackle a bigger, harder problems for the same

computational budget. Recent work by Sherry and Thompson (2020) has shown that improve-

ments in algorithms have been large: rivaling or exceeding the decades-long improvements in

computer hardware coming from Moore's Law for many problem types.

18 THOMPSON ET AL.