DOCSLIB.ORG

Ai & the Sustainable Development Goals

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Ai & the Sustainable Development Goals 1 AI & THE SUSTAINABLE DEVELOPMENT GOALS: THE STATE OF PLAY Global Goals Technology Forum 2 3 INTRODUCTION In 2015, 193 countries agreed to the United Nations (UN) 2030 Agenda for Sustainable Development, which provides a shared blueprint for peace and prosperity for people and the planet, now and into the future. At its heart are the 17 Sustainable Development Goals (SDGs), which are an urgent call for action by all countries – developed and developing – in a global partnership. Achieving the SDGs is not just a moral imperative, but an economic one. While the world is making progress in some areas, we are falling behind in delivering the SDGs overall. We need all actors – businesses, governments, academia, multilateral institutions, NGOs, and others – to accelerate and scale their efforts to deliver the SDGs, using every tool at their disposal, including artificial intelligence (AI). In December 2017, 2030Vision 2030Vision published its first report, Uniting to “WE ARE AT A PIVOTAL Deliver Technology for the Global ABOUT 2030VISION Goals, which addressed the role MOMENT. AS ARTIFICIAL of digital technology – big data, AI & The Sustainable Development Goals: The State of Play State Goals: The Development Sustainable AI & The Founded in 2017, 2030Vision INTELLIGENCE BECOMES robotics, internet of things, AI, and is a partnership of businesses, MORE WIDELY ADOPTED, other technologies – in achieving NGOs, and academia that aims to the SDGs. transform the use of technology to WE HAVE A TREMENDOUS support the delivery of the SDGs. OPPORTUNITY TO In this paper, we focus on AI for 2030Vision serves as a platform REEVALUATE WHAT WE’VE the SDGs. AI extends and amplifies to convene cross-sector leaders CREATED AND ENSURE the capacity of human beings to understand and solve complex, to raise awareness of the SDGs, to THAT AI IS DEVELOPED showcase thought-leadership on dynamic, and interconnected the role of technology in addressing COLLABORATIVELY AND systems challenges like the SDGs. the SDGs, and to stimulate APPLIED TO THE UN Our main objective was to survey partnerships for action. SUSTAINABLE DEVELOPMENT the landscape of research and GOALS TO ACHIEVE A BETTER initiatives on AI and the SDGs to identify key themes and questions AND MORE SUSTAINABLE in need of further exploration. We FUTURE FOR ALL.” also reviewed the state of AI and Tabitha Goldstaub the SDGs in two sectors – food and agriculture and healthcare – to Co-Founder at CognitionX & Chair understand if and how AI is being of the UK Government’s AI Council deployed to address the SDGs and the challenges and opportunities in doing so. 4 5 AI AND THE SUSTAINABLE DEVELOPMENT GOALS: As illustrated by the quotes opposite, there are a variety of DEFINING ARTIFICIAL perspectives about AI and its 1 potential impact on sustainable INTELLIGENCE development. For some, AI is a Whether we know it or not, AI is job killer, a tool that will benefit THE now part of the fabric of our daily wealthier nations and citizens. lives. For example, AI helps power Worse, AI could pose existential predictive Google searches, Spotify threats, for example Stephen music recommendations, Waze Hawking warned that AI could driving directions, and Facebook lead to the end of humanity. POTENTIAL facial recognition. However, AI promises – and is There are a variety of definitions of starting to deliver – benefits AI, which we paraphrase as follows: across sectors and geographies, and holds great potential to help solve complex and interconnected AI IS AN FOR GOOD sustainable development challenges such as climate change, access to health care, OVERARCHING TERM and inequality. While 2030Vision “PERHAPS THE MOST recognizes the potential risks of FOR A COLLECTION IMPORTANT QUESTION AI, we believe AI will be a positive OF TECHNOLOGY “...CLIMATE CHANGE IS A WE HAVE LOOKED AT IS force for global sustainable development. Guided by the 2030Vision MASSIVE PROBLEM ACROSS WHETHER AI WILL POSE SDGs, we must now work together ALGORITHMS AND NEARLY EVERY SECTOR A THREAT – OR PROVIDE to accelerate and scale the AND MEASURE OF HUMAN NEW OPPORTUNITIES – FOR development and use of AI. APPROACHES THAT DEVELOPMENT. TO ADDRESS DEVELOPING REGIONS SUCH AI & The Sustainable Development Goals: The State of Play State Goals: The Development Sustainable AI & The IT AT THE SPEED AND SCALE AS AFRICA. OPTIMISTS SAY ALLOW MACHINES TO THAT CURRENT CONDITIONS THAT SUCH PLACES COULD PERFORM HUMAN- REQUIRE, WE’LL NEED USE RAPIDLY ADVANCING TO TAKE A MORE DATA- AI SYSTEMS TO BOOST LIKE COGNITIVE DRIVEN APPROACH – ONE PRODUCTIVITY AND THAT HARNESSES THE FULL LEAPFROG AHEAD. BUT I AM FUNCTIONS SUCH POWER OF ARTIFICIAL BECOMING INCREASINGLY INTELLIGENCE AND OTHER CONCERNED THAT AI AS REASONING AND ADVANCED TECHNOLOGIES WILL, IN FACT, BLOCK THE LEARNING. TO ACCELERATE DISCOVERY TRADITIONAL GROWTH PATH AND INNOVATION AT A BY REPLACING LOW-WAGE AI is also being used to address TRULY PLANETARY SCALE.” JOBS WITH ROBOTS.” sustainable development challenges: improving the Lucas Joppa Ian Goldin diagnosis of various diseases, Chief Environmental Scientist, Professor of Globalisation and fighting wildlife poaching, and Microsoft Development, Oxford University improving crop yields to name a few. Other examples of SDG- related applications of AI are included throughout this report. 6 7 PROGRESS ON THE SDGS AI AND THE SUSTAINABLE In its most recent progress report However, progress has been on the SDGs, the UN cited success insufficient in other areas, DEVELOPMENT GOALS: in some areas, for example: for example: GOAL 1: NO POVERTY GOAL 4: QUALITY GOAL 14: LIFE BELOW EDUCATION WATER 2 Since 1990, the percentage of people living in extreme poverty GOAL 15: LIFE ON LAND CAPTURING (living on less than $1.90 per 617m UN scientists recently day) has declined from children of primary and warned that secondary school age do not 33% TO 9% meet minimum proficiency in million mathematics and reading. 1 THE STATE out of the world’s eight million GOAL 3: GOOD HEALTH species are at risk of extinction AND WELL-BEING GOAL 5: GENDER EQUALITY due to human activities. Since 2000, maternal Globally, women represent just OF PLAY mortality has declined by 23% OF SEATS 37% in single or lower houses of national parliaments. and under-five mortality has declined by Organizations such as McKinsey Over the last two and PwC have published reports 2030Vision GOAL 13: CLIMATE ACTION years, we have seen a with a wealth of information – 47% including use cases, economic In May 2019, the concentration considerable increase impacts, risks, enabling factors, of CO2 in the atmosphere and more. We’ve seen a number reached in discussion about AI & The Sustainable Development Goals: The State of Play State Goals: The Development Sustainable AI & The GOAL 7: AFFORDABLE With 10 years of non-profit organizations and AND CLEAN ENERGY remaining to achieve the role that AI can events emerge at the intersection 415ppm play in sustainable of AI and the SDGs, for example the ambitions of the AI for Good Foundation and 87% for the first time in human development issues. ITU’s AI for Good Summit. And history, and it continues to rise. the 2030 agenda, every day there are multiple media of the global population has pieces on the role of AI and some access to electricity, up from it is up to all of sustainability topic. 78% in 2000. us – businesses, governments, academia, multilateral institutions, NGOs, and others – to strengthen and quicken our efforts to build a better future for everyone. 8 9 DIAGNOSIS AND TREATMENT FINANCIAL CRIME A growing number of researchers Financial institutions are using AI are exploring the use of AI in to analyze ever larger and more the diagnosis and treatment of varied data sets to spot fraud and various diseases and the design of money laundering. Banks spend health interventions. For example, £5 billion per year on fighting researchers in Germany, the US, financial crime in the UK, and AI and France found that AI is on can reduce the cost and increase par or better than dermatologists the effectiveness of such efforts. at diagnosing skin cancer. Researchers from IBM and New York University recently published a paper on how AI can be used to detect glaucoma, the second leading cause of blindness in the world. Such AI-based systems hold THE HIGH COST OF ILLEGAL promise to make diagnostic and TRADE IN WILDLIFE Given how much positive work has been other health services more readily Illegal and unsustainable available in regions that have a lack completed and is ongoing in this space, wildlife trade is a threat to the of health care workers. we reviewed this work to understand: existence of certain species COMBATTING FRAUD AND and human livelihoods. The UN OTHER FINANCIAL CRIMES estimates that illegal wildlife What ground has been covered with respect to AI and the SDGs? trade worldwide is worth $8 HSBC is using AI to help spot What questions are in need of further exploration? billion to $10 billion per year, money laundering, fraud, and the value of the ivory trade and terrorist funding. AI What frameworks are useful for connecting AI to the SDGs? alone is roughly $1 billion. Non- enables the bank to screen the vast amounts of data it Can AI be used to better understand the interaction between the SDGs? profit RESOLVE is deploying cameras enhanced with image holds on customers and their How is AI being used to address the SDGs in both the food and agriculture processing and deep neural transactions against publicly and health sectors? network algorithms across a available data, in the search for suspicious activity. What should different stakeholders do to foster the use of AI for the SDGs? REDUCING MATERNAL range of reserves in Africa to MORTALITY IN UTTAR allow rangers to detect and PRADESH, INDIA respond to poachers in near real-time.
Load more

Recommended publications
  • Artificial Intelligence in Health Care: the Hope, the Hype, the Promise, the Peril
    Artificial Intelligence in Health Care: The Hope, the Hype, the Promise, the Peril Michael Matheny, Sonoo Thadaney Israni, Mahnoor Ahmed, and Danielle Whicher, Editors WASHINGTON, DC NAM.EDU PREPUBLICATION COPY - Uncorrected Proofs NATIONAL ACADEMY OF MEDICINE • 500 Fifth Street, NW • WASHINGTON, DC 20001 NOTICE: This publication has undergone peer review according to procedures established by the National Academy of Medicine (NAM). Publication by the NAM worthy of public attention, but does not constitute endorsement of conclusions and recommendationssignifies that it is the by productthe NAM. of The a carefully views presented considered in processthis publication and is a contributionare those of individual contributors and do not represent formal consensus positions of the authors’ organizations; the NAM; or the National Academies of Sciences, Engineering, and Medicine. Library of Congress Cataloging-in-Publication Data to Come Copyright 2019 by the National Academy of Sciences. All rights reserved. Printed in the United States of America. Suggested citation: Matheny, M., S. Thadaney Israni, M. Ahmed, and D. Whicher, Editors. 2019. Artificial Intelligence in Health Care: The Hope, the Hype, the Promise, the Peril. NAM Special Publication. Washington, DC: National Academy of Medicine. PREPUBLICATION COPY - Uncorrected Proofs “Knowing is not enough; we must apply. Willing is not enough; we must do.” --GOETHE PREPUBLICATION COPY - Uncorrected Proofs ABOUT THE NATIONAL ACADEMY OF MEDICINE The National Academy of Medicine is one of three Academies constituting the Nation- al Academies of Sciences, Engineering, and Medicine (the National Academies). The Na- tional Academies provide independent, objective analysis and advice to the nation and conduct other activities to solve complex problems and inform public policy decisions.
    [Show full text]
  • Artificial Intelligence
    TechnoVision The Impact of AI 20 18 CONTENTS Foreword 3 Introduction 4 TechnoVision 2018 and Artificial Intelligence 5 Overview of TechnoVision 2018 14 Design for Digital 17 Invisible Infostructure 26 Applications Unleashed 33 Thriving on Data 40 Process on the Fly 47 You Experience 54 We Collaborate 61 Applying TechnoVision 68 Conclusion 77 2 TechnoVision 2018 The Impact of AI FOREWORD We introduce TechnoVision 2018, now in its eleventh year, with pride in reaching a second decade of providing a proven and relevant source of technology guidance and thought leadership to help enterprises navigate the compelling and yet complex opportunities for business. The longevity of this publication has been achieved through the insight of our colleagues, clients, and partners in applying TechnoVision as a technology guide and coupling that insight with the expert input of our authors and contributors who are highly engaged in monitoring and synthesizing the shift and emergence of technologies and the impact they have on enterprises globally. In this edition, we continue to build on the We believe that with TechnoVision 2018, you can framework that TechnoVision has established further crystalize your plans and bet on the right for several years with updates on last years’ technology disruptions, to continue to map and content, reflecting new insights, use cases, and traverse a successful digital journey. A journey technology solutions. which is not about a single destination, but rather a single mission to thrive in the digital epoch The featured main theme for TechnoVision 2018 through agile cycles of transformation delivering is AI, due to breakthroughs burgeoning the business outcomes.
    [Show full text]
  • The Importance of Generation Order in Language Modeling
    The Importance of Generation Order in Language Modeling Nicolas Ford∗ Daniel Duckworth Mohammad Norouzi George E. Dahl Google Brain fnicf,duckworthd,mnorouzi,[email protected] Abstract There has been interest in moving beyond the left-to-right generation order by developing alter- Neural language models are a critical compo- native multi-stage strategies such as syntax-aware nent of state-of-the-art systems for machine neural language models (Bowman et al., 2016) translation, summarization, audio transcrip- and latent variable models of text (Wood et al., tion, and other tasks. These language models are almost universally autoregressive in nature, 2011). Before embarking on a long-term research generating sentences one token at a time from program to find better generation strategies that left to right. This paper studies the influence of improve modern neural networks, one needs ev- token generation order on model quality via a idence that the generation strategy can make a novel two-pass language model that produces large difference. This paper presents one way of partially-filled sentence “templates” and then isolating the generation strategy from the general fills in missing tokens. We compare various neural network design problem. Our key techni- strategies for structuring these two passes and cal contribution involves developing a flexible and observe a surprisingly large variation in model quality. We find the most effective strategy tractable architecture that incorporates different generates function words in the first pass fol- generation orders, while enabling exact computa- lowed by content words in the second. We be- tion of the log-probabilities of a sentence.
    [Show full text]
  • The Deep Learning Revolution and Its Implications for Computer Architecture and Chip Design
    The Deep Learning Revolution and Its Implications for Computer Architecture and Chip Design Jeffrey Dean Google Research [email protected] Abstract The past decade has seen a remarkable series of advances in machine learning, and in particular deep learning approaches based on artificial neural networks, to improve our abilities to build more accurate systems across a broad range of areas, including computer vision, speech recognition, language translation, and natural language understanding tasks. This paper is a companion paper to a keynote talk at the 2020 International Solid-State Circuits Conference (ISSCC) discussing some of the advances in machine learning, and their implications on the kinds of computational devices we need to build, especially in the post-Moore’s Law-era. It also discusses some of the ways that machine learning may also be able to help with some aspects of the circuit design process. Finally, it provides a sketch of at least one interesting direction towards much larger-scale multi-task models that are sparsely activated and employ much more dynamic, example- and task-based routing than the machine learning models of today. Introduction The past decade has seen a remarkable series of advances in machine learning (ML), and in particular deep learning approaches based on artificial neural networks, to improve our abilities to build more accurate systems across a broad range of areas [LeCun et al. 2015]. Major areas of significant advances ​ ​ include computer vision [Krizhevsky et al. 2012, Szegedy et al. 2015, He et al. 2016, Real et al. 2017, Tan ​ ​ ​ ​ ​ ​ and Le 2019], speech recognition [Hinton et al.
    [Show full text]
  • AI Principles 2020 Progress Update
    AI Principles 2020 Progress update AI Principles 2020 Progress update Table of contents Overview ........................................................................................................................................... 2 Culture, education, and participation ....................................................................4 Technical progress ................................................................................................................... 5 Internal processes ..................................................................................................................... 8 Community outreach and exchange .....................................................................12 Conclusion .....................................................................................................................................17 Appendix: Research publications and tools ....................................................18 Endnotes .........................................................................................................................................20 1 AI Principles 2020 Progress update Overview Google’s AI Principles were published in June 2018 as a charter to guide how we develop AI responsibly and the types of applications we will pursue. This report highlights recent progress in AI Principles implementation across Google, including technical tools, educational programs and governance processes. Of particular note in 2020, the AI Principles have supported our ongoing work to address
    [Show full text]
  • User Experience + Research Scientist Collaboration in a Generative Machine Learning Interface
    CHI 2019 Case Study CHI 2019, May 4–9, 2019, Glasgow, Scotland, UK Identifying the Intersections: User Figure 1: Google AI. http://ai.google Experience + Research Scientist Collaboration in a Generative Machine Learning Interface Claire Kayacik, Sherol Chen, Signe Noerly, Jess Holbrook, Adam Roberts, Douglas Eck Google, Inc. Mountain View, CA , 94043 USA (cbalgemann,sherol,noerly,jessh,adarob,deck)@google.com ABSTRACT Creative generative machine learning interfaces are stronger when multiple actors bearing different points of view actively contribute to them. User experience (UX) research and design involvement in the creation of machine learning (ML) models help ML research scientists to more effectively identify human needs that ML models will fulfill. The People and AI Research (PAIR) group within Google developed a novel program method in which UXers are embedded into an ML research group for three Figure 2: Magenta is a project within Google Brain that focuses on training ma- months to provide a human-centered perspective on the creation of ML models. The first full-time chines for creative expressivity, genera- cohort of UXers were embedded in a team of ML research scientists focused on deep generative models tive personalization, and intuitive inter- to assist in music composition. Here, we discuss the structure and goals of the program, challenges we faces. https://g.co/magenta faced during execution, and insights gained as a result of the process. We offer practical suggestions Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.
    [Show full text]
  • Google Cloud Platform: Healthcare Solutions Playbook
    Google Cloud Platform: Healthcare Solutions Playbook Cloud training, tools, and reference architectures for researchers - Prepared for the National Institutes of Health (NIH) STRIDES. Author(s):​ Adrienne Wilkins, Trenton Leslie Title:​ Technical Account Manager (TAM), Intern PROFESSIONAL SERVICES Contents Introduction 2 Learning 2 Data Management Solutions 4 Reference Architectures 5 Genomics and Secondary Analysis 5 Patient Monitoring 9 Variant Analysis 11 Healthcare API for Machine Learning and Analytics 14 Radiological Image Extraction 18 Appendix 21 Page 1 PROFESSIONAL SERVICES Introduction The National Institutes of Health (NIH) established The Science and Technology Research Infrastructure for Discovery, Experimentation, and Sustainability (STRIDES) initiative to provide biomedical researchers with access to advanced, cost-effective, cloud-based computational infrastructure, tools, and services. Through STRIDES, researchers can take advantage of emerging data management methodologies, technological expertise, computational platforms, and tools to support cutting-edge experimentation and innovation. NIH has partnered with Google Cloud to support the STRIDES initiative through cloud services. In support of STRIDES, we’ve developed sets of playbooks to help enable researchers to build healthcare and life sciences solutions on Google Cloud Platform (GCP). The goal of this playbook it to aid researchers in developing healthcare systems and workflows on GCP. This playbook will provide GCP architecture and solutions examples for implementing
    [Show full text]
  • Microsoft Outlook
    From: Scott Wilson <[email protected]> Sent: Wednesday, November 13, 2019 5:18 PM To: aipartnership Subject: Re: Request for Comments on IP protection for AI innovation Docket No. PTO- C-2019-0038 It appears that companies such as Google and others are using copyrighted material for machine learning training datasets and creating for-profit products and patents that are partially dependent on copyrighted works. For instance, Google AI is using YouTube videos for machine learning products, and they have posted this information on their GoogleAI blog on multiple occasions. There are Google Books ML training datasets available in multiple locations on the internet, including at Amazon.com. There are two types of machine learning processes that I am aware of - expressive and non-expressive. A non- expressive ML training dataset would catalogue how many times a keyword appears in a copyrighted text, and appears to be permitted by fair use. However, expressive ML training analyzing artistic characteristics of copyrighted works to create patented products and/or processes does not appear to be exempted by fair use copyright laws. The whole issue revolves around the differences between research and product/profit creation in a corporate context. I am including an article below that examines these legal issues in more detail. https://lawandarts.org/wp-content/uploads/sites/14/2017/12/41.1_Sobel-FINAL.pdf or www.bensobel.org - first article on page links to the link above. As a copyright holder, I am concerned that this is a widespread industry practice of using copyrighted material without the knowledge or consent of copyright holders to create for-profit AI products that not only exploits copyright creators, but increasingly separates them from the ability to profit from their own work.
    [Show full text]
  • Google Goes All in on Artificial Intelligence As 'Google Research' Becomes 'Google AI'
    AiA Art News-service Google Goes All In On Artificial Intelligence As 'Google Research' Becomes 'Google AI' Sam Shead , CONTRIBUTORI cover artificial intelligence and Google DeepMind. Opinions expressed by Forbes Contributors are their own. MOUNTAIN VIEW, CA - MAY 08: Google CEO Sundar Pichai delivers the keynote address at the Google I/O 2018 Conference at Shoreline Amphitheater on May 8, 2018 in Mountain View, California. Google's two day developer conference runs through Wednesday May 9. (Photo by Justin Sullivan/Getty Images) Google has rebranded the whole of its "Google Research" division to Google AI as the company aggressively pursues developments in the burgeoning field of artificial intelligence. The change, announced ahead of its Google I/O developers conference this week, shows just how serious Google is when it comes to AI, which encompasses technologies such as computer vision, deep learning, and speech recognition. Announcing the news in a blog post, Google said it has been implementing machine learning into nearly everything it does over the last few years. "To better reflect this commitment, we're unifying our efforts under "Google AI", which encompasses all the state-of-the-art research happening across Google," wrote Christian Howard, editor-in-chief of Google AI Communications. Google has changed the homepage for Google AI so that it features published research on a variety of subject matters, stories of AI in action, and open source material and tools that others can use. The old Google Research site redirects to the new Google AI site. Google has also renamed the Google Research Twitter and Google+ channels to Google AI as part of the change.
    [Show full text]
  • Presentation by Amr Gaber
    https://m.facebook.com/story.php?story_fbid=2191943484425083&id=231112080362037 Transcript of UN remarks by Amr Gaber at CCW 2018/28/8 [Starts around 17:40] Hello my name is Amr Gaber. I am a very low level Google software engineer and I'm here speaking on behalf of myself and not my employer. I am also a volunteer with the Tech Workers Coalition. We’re a labor organization of tech workers, community members and other groups that are focused on bringing tech workers together from across industries, companies and backgrounds in order to make technology and the tech industry more equitable and just. I'm here to talk to you today because I took action with several of my co-workers at Google to stop Google from pursuing artificial intelligence in military applications. So there is a contract called Project Maven which is also a larger Department of Defense project and Google's involvement was to use artificial intelligence to enhance drone surveillance and targeting through image recognition. So I heard about this project from my co-workers internally through whistleblowers who were sharing it in the company forums. I co-wrote an open letter with several other Google employees to the CEO Sundar Pichai asking Google to cancel the project. Over 4,000 of my co-workers signed it, but that alone was not enough. It was actually a combination of workers, academics, students and outside groups that eventually got Google to step back from the contract. Organizations like the Tech Workers Coalition, Coworker.org, the Campaign To Stop Killer Robots, and the International Committee for Robot Arms Control which wrote a petition that over 1,000 academics signed including Noam Chomsky supporting me and my co-workers and our call to cancel the contract.
    [Show full text]
  • AI Explainability Whitepaper
    AI Explainability Whitepaper About This Whitepaper This whitepaper is a technical reference accompanying Google Cloud's AI Explanations product. It is targeted at model developers & data scientists responsible for designing and delivering ML models. Our goal is for them to leverage AI Explanations to simplify model development as well as explain the model’s behavior to key stakeholders. Product managers, business leaders and end users may also find parts of this whitepaper relevant, particularly around use cases enabled by AI Explanations and, crucially, considerations around its proper usage as well as its current limitations. Specifically, we refer these readers to the "​Usage Examples​" as well as the "​Attribution Limitations and Usage Considerations​" sections. Overview Evolution of Machine Learning In the quest for more accurate AI, the availability of compute resources coupled with increasing dataset sizes have fueled a trend towards more complex non-linear models. Indeed, we've seen a progression from handcrafted rules and heuristics, to linear models and decision trees, ensembles and deep models to, most recently, meta-learning or models that create other models. 1 These more accurate models have resulted in a paradigm shift along multiple dimensions: - Expressiveness ​enables fitting a wide range of functions in an increasing number of domains like forecasting, ranking, autonomous driving, particle physics, drug discovery, etc. - Versatility​ unlocks data modalities (image, audio, speech, text, tabular, time series, etc.) and enables joint/multi-modal applications. - Adaptability ​to small data regimes through transfer and multi-task learning. - Efficiency​ through custom optimized hardware like GPUs and TPUs enabling practitioners to train bigger models faster and cheaper.
    [Show full text]
  • Data-Driven Mergers and Personalization∗
    Data-Driven Mergers and Personalization∗ Zhijun Chen,y Chongwoo Choe,z Jiajia Cong,x and Noriaki Matsushima{ October 27, 2020 First draft. Comments welcome. Abstract Recent years have seen growing cases of data-driven tech mergers such as Google/Fitbit, in which a dominant digital platform acquires a relatively small firm possessing a large volume of consumer data. The digital platform can consolidate the consumer data with its existing data set from other services and use it for personalization in related markets. We develop a theoretical model to examine the impact of such mergers across the two markets that are related through a consumption synergy. The merger links the markets for data collection and data application, through which the digital platform can leverage its market power and hurt competitors in both markets. Personalization can lead to exploitation of some consumers in the market for data application. But insofar as competitors remain active, the merger increases total consumer surplus in both markets by intensifying competition. When the consumption synergy is large enough, the merger can result in monopolization of both markets, leading to further consumer harm when stand-alone competitors exit in the long run. Thus there is a trade-off where potential dynamic costs can outweigh static benefits. We also discuss policy implications by considering various merger remedies. Key words: big data, personalization, tech mergers JEL Classification: D43, L13, L41, K21 ∗We are grateful to Arthur Campbell, Greg Crawford, Jacques Cr´emer,Doh-Shin Jeon, Bruno Jullien, Paul Heidhues, Vai-Lam Mui, Patrick Rey, Chengsi Wang, and Xiaojian Zhao for many helpful comments.
    [Show full text]