Artificial Intelligence And The Modern Productivity .

2y ago
793.33 KB
46 Pages
Last View : 17d ago
Last Download : 5m ago
Upload by : Averie Goad

NBER WORKING PAPER SERIESARTIFICIAL INTELLIGENCE AND THE MODERN PRODUCTIVITY PARADOX:A CLASH OF EXPECTATIONS AND STATISTICSErik BrynjolfssonDaniel RockChad SyversonWorking Paper 24001 BUREAU OF ECONOMIC RESEARCH1050 Massachusetts AvenueCambridge, MA 02138November 2017We thank Eliot Abrams, Ajay Agrawal, David Autor, Seth Benzell, Joshua Gans, Avi Goldfarb,Austan Goolsbee, Guillaume Saint-Jacques, Andrea Meyer, Manuel Tratjenberg, and numerousparticipants at the NBER Workshop on AI and Economics in September, 2017. In particular,Rebecca Henderson provided detailed and very helpful comments on an earlier draft and LarrySummers suggested the analogy to the J-Curve. Generous funding for this research was providedin part by the MIT Initiative on the Digital Economy. The views expressed herein are those of theauthors and do not necessarily reflect the views of the National Bureau of Economic Research.NBER working papers are circulated for discussion and comment purposes. They have not beenpeer-reviewed or been subject to the review by the NBER Board of Directors that accompaniesofficial NBER publications. 2017 by Erik Brynjolfsson, Daniel Rock, and Chad Syverson. All rights reserved. Shortsections of text, not to exceed two paragraphs, may be quoted without explicit permissionprovided that full credit, including notice, is given to the source.

Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations andStatisticsErik Brynjolfsson, Daniel Rock, and Chad SyversonNBER Working Paper No. 24001November 2017JEL No. D2,O3,O4ABSTRACTWe live in an age of paradox. Systems using artificial intelligence match or surpass human levelperformance in more and more domains, leveraging rapid advances in other technologies anddriving soaring stock prices. Yet measured productivity growth has declined by half over the pastdecade, and real income has stagnated since the late 1990s for a majority of Americans. Wedescribe four potential explanations for this clash of expectations and statistics: false hopes,mismeasurement, redistribution, and implementation lags. While a case can be made for each, weargue that lags have likely been the biggest contributor to the paradox. The most impressivecapabilities of AI, particularly those based on machine learning, have not yet diffused widely.More importantly, like other general purpose technologies, their full effects won’t be realizeduntil waves of complementary innovations are developed and implemented. The requiredadjustment costs, organizational changes, and new skills can be modeled as a kind of intangiblecapital. A portion of the value of this intangible capital is already reflected in the market value offirms. However, going forward, national statistics could fail to measure the full benefits of thenew technologies and some may even have the wrong sign.Erik BrynjolfssonMIT Sloan School of Management100 Main Street, E62-414Cambridge, MA 02142and NBERerikb@mit.eduDaniel RockMIT Sloan School of Management100 Main Street, E62-365Cambridge, MA 02142drock@mit.eduChad SyversonUniversity of ChicagoBooth School of Business5807 S. Woodlawn Ave.Chicago, IL 60637and

The discussion around the recent patterns in aggregate productivity growthhighlights a seeming contradiction. On the one hand, there are astonishing examples ofpotentially transformative new technologies that could greatly increase productivity andeconomic welfare (see e.g. Brynjolfsson and McAfee, 2014). There are some early concretesigns of these technologies’ promise, the recent leaps in artificial intelligence (AI)performance being the most prominent example. However, at the same time, measuredproductivity growth over the past decade has slowed significantly. This deceleration islarge, cutting productivity growth by half or more of its level in the decade preceding theslowdown. It is also widespread, having occurred throughout the OECD and, more recently,among many large emerging economies as well (Syverson, 2017). 1We thus appear to be facing a redux of the Solow (1987) Paradox: we seetransformative new technologies everywhere but in the productivity statistics.In this paper, we review the evidence and explanations for the modern productivityparadox and propose a resolution. Namely, there is no inherent inconsistency betweenforward-looking technological optimism and backward-looking disappointment. Both cansimultaneously exist. Indeed, there are good conceptual reasons to expect them tosimultaneously exist when the economy undergoes the kind of restructuring associatedwith transformative technologies. In essence, the forecasters of future company wealth andthe measurers of historical economic performance show the greatest disagreement duringtimes of technological change. In this paper we argue and present some evidence that theeconomy is in such a period now.Sources of Technological OptimismPaul Polman, Unilever’s CEO, recently claimed that “The speed of innovation hasnever been faster.” Similarly, Bill Gates, Microsoft’s co-founder, observes that “Innovation ismoving at a scarily fast pace.” Vinod Khosla of Khosla Ventures sees “the beginnings of. [a]rapid acceleration in the next 10, 15, 20 years.” Eric Schmidt, Executive Chairman ofAlphabet Inc., believes “we’re entering the age of abundance [and] during the age ofA parallel yet more pessimistically oriented debate about potential technological progress is the activediscussion about robots taking jobs from more and more workers (e.g., Brynjolfsson and McAfee, 2011;Acemoglu and Restrepo, 2017; Bessen, 2017; Autor and Salomons, 2017).11

abundance, we’re going to see a new age the age of intelligence.” Ray Kurzweil famouslypredicts that The Singularity – when AI surpasses humans – will occur sometime around2045. 2 Assertions like these are especially common among technology leaders and venturecapitalists.In part, these assertions reflect the continuing progress of IT in many areas, fromcore technology advances like further doublings of basic computer power (but from everlarger bases) to successful investment in the essential complementary innovations likecloud infrastructure, and new service-based business models. But the bigger source ofoptimism is the wave of recent improvements in AI, especially machine learning. Machinelearning represents a fundamental change from the first wave of computerization.Historically, most computer programs were created by meticulously codifying humanknowledge, step-by-step, mapping inputs to outputs as prescribed by the programmers. Incontrast, machine learning systems use categories of general algorithms (e.g., neuralnetworks) to figure out the relevant mapping on their own, typically by being fed very largedata sets of examples. By using these machine learning methods leveraging the growth intotal data and data processing resources, machines have made impressive gains inperception and cognition, two essential skills for most types of human work. For instance,error rates in labeling the content of photos on ImageNet, a dataset of over 10 millionimages, have fallen from over 30% in 2010 to less than 5% in 2016 and most recently aslow as 2.2% with SE-ResNet152 in the ILSVRC2017 competition (see Figure 1). 3 Error ratesin voice recognition on the Switchboard speech recording corpus, often used to measureprogress in speech recognition, have improved from 8.5% to 5.5% over the past year (Saonet al., 2017). The five percent threshold is important, because that is roughly theperformance of humans on each of these tasks on the same test made by Ray Kurzweil#2045: The s. ImageNet includes labels for each image, originallyprovided by humans. For instance, there are 339,000 labeled as flowers, 1,001,000 as food, 188,000 as fruit,137,000 as fungus, and so on.32

Although not at the level of professional human performance yet, Facebook’s AIResearch team recently improved upon the best machine language translation algorithmsavailable using convolutional neural net sequence prediction techniques (Gehring et al.,2017). Deep learning techniques have also been combined with reinforcement learning, apowerful set of techniques used to generate control and action systems wherebyautonomous agents are trained to take actions given an environment state to maximizefuture rewards. Though nascent, advances in this field are impressive. In addition to itsvictories in the game of Go, Google DeepMind has achieved superhuman performance inmany Atari games (Fortunato et al., 2017).These are notable technological milestones. But they can also change the economiclandscape, creating new opportunities for business value creation and cost reduction. Forexample, a system using deep neural networks was tested against 21 board certifieddermatologists and matched their performance in diagnosing skin cancer (Esteva et al.,2017). Facebook uses neural networks for over 4.5 billion translations each day. 4Figure 1. AI vs. Human Image Recognition Error tion/43

An increasing number of companies have responded to these opportunities. Googlenow describes its focus as “AI first,” while Microsoft’s CEO Satya Nadella says AI is the“ultimate breakthrough” in technology. Their optimism about AI is not just cheap talk. Theyare making heavy investments in AI, as are Apple, Facebook, and Amazon. As of September2017, these companies comprise the five most valuable companies in the world.Meanwhile, the tech-heavy Nasdaq composite stock index more than doubled between2012 and 2017. According to CBInsights, global investment in private companies focusedon AI has grown even faster, increasing from 589 million in 2012 to over 5 billion in2016. 5The Disappointing Recent RealityAlthough the technologies discussed above hold great potential, there is little signthat they have yet affected aggregate productivity statistics. Labor productivity growthrates in a broad swath of developed economies fell in the mid-2000s and have stayed lowsince then. For example, aggregate labor productivity growth in the U.S. averaged only1.3% per year from 2005 to 2016, less than half of the 2.8% annual growth rate sustainedfrom 1995 to 2004. Fully 28 of 29 other countries for which the OECD has compiledproductivity growth data saw similar decelerations. The unweighted average annual laborproductivity growth rates across these countries was 2.3% from 1995 to 2004 but only1.1% from 2005 to 2015. 6 What’s more, real median income has stagnated since the late1990s and non-economic measures of well-being, like life expectancy, have fallen for somegroups (Case and Deaton, 2017).Figure 2 replicates the Conference Board’s analysis of its country-level TotalEconomy Database (Conference Board, 2016). It plots highly smoothed annual productivitygrowth rate series for the U.S., other mature economies (which combined match much ofAnd the number of deals increased from 160 to 658. See lligence-startup-funding/6 These slowdowns are statistically significant. For the U.S., where the slowdown is measured using quarterlydata, equality of the two periods’ growth rates is rejected with a t-statistic of 2.9. The OECD numbers comefrom annual data across the 30 countries. Here, the null hypothesis of equality is rejected with a t-statistic of7.2.54

the OECD sample cited above), emerging and developing economies, and the world overall.The aforementioned slowdowns in the U.S. and other mature economies are clear in thefigure. The figure also reveals that the productivity growth acceleration in emerging anddeveloping economies during the 2000s ended around the time of the Great Recession,causing a recent decline in productivity growth rates in these countries too.These slowdowns do not appear to simply reflect effects of the Great Recession. Inthe OECD data, 28 of the 30 countries still exhibit productivity decelerations if 2008-09growth rates are excluded from the totals. Cette, Fernald, and Mojon (2016), using otherdata, also find substantial evidence that the slowdowns began before the Great Recession.Figure 2. Smoothed Average Annual Labor Productivity Growth (Percent) by RegionBoth capital deepening and total factor productivity (TFP) growth lead to laborproductivity growth, and both seem to be playing a role in the slowdown (e.g., Fernald,2014; OECD, 2015). Disappointing technological progress can be tied to each of thesecomponents. TFP directly reflects such progress. Capital deepening is indirectly influencedby technological change because firms’ investment decisions respond to improvements incapital’s current or expected marginal product.These facts have been read by some as reasons for pessimism about the ability ofnew technologies like AI to greatly affect productivity and income. Gordon (2014, 2015)5

argues that productivity growth has been in long-run decline, with the IT-drivenacceleration of 1995 to 2004 being a one-off aberration. While not claiming technologicalprogress will be nil in the coming decades, Gordon essentially argues that we have beenexperiencing the new, low-growth normal and should expect to continue to do so goingforward. Cowen (2011) similarly offers multiple reasons why innovation may be slow atleast for the foreseeable future. Bloom et al. (2017) document that in many fields oftechnological progress, research productivity has been falling, while Nordhaus (2015) findsthat the hypothesis of an acceleration of technology-driven growth fails a variety of tests.This pessimistic view of future technological progress has entered into long-rangepolicy planning. The Congressional Budget Office, for instance, reduced its 10-year forecastfor average U.S. annual labor productivity growth from 1.8 percent in 2016 (CBO, 2016) to1.5 percent in 2017 (CBO, 2017). Although perhaps modest on its surface, that drop impliesU.S. GDP will be considerably smaller 10 years from now than it would in the moreoptimistic scenario—a difference equivalent to almost 600 billion in 2017.Potential Explanations for the ParadoxThere are four principal candidate explanations for the current confluence oftechnological optimism and poor productivity performance: 1) false hopes, 2)mismeasurement, 3) concentrated distribution and rent dissipation, and 4) implementationand restructuring lags. 7False HopesThe simplest possibility is that the optimism about the potential technologies ismisplaced and unfounded. Perhaps these technologies won’t be as transformative as manyexpect, and although they might have modest and noteworthy effects on specific sectors,their aggregate impact might be small. In this case, the paradox will be resolved in thefuture because realized productivity growth never escapes its current doldrums, which willforce the optimists to mark their beliefs to market.7To some extent, these explanations parallel the explanations for the Solow Paradox (Brynjolfsson, 1993).6

History and some current examples offer a quantum of credence to this possibility.Certainly one can point to many prior exciting technologies that did not live up to initiallyoptimistic expectations. Nuclear power never became too cheap to meter, and fusionenergy has been 20 years away for 60 years. Mars may still beckon, but it’s been more than40 years since Eugene Cernan was the last person to walk on the moon. Flying cars nevergot off the ground, 8 and passenger jets no longer fly at supersonic speeds. Even AI, perhapsthe most promising technology of our era, is well behind Marvin Minsky’s 1967 predictionthat “Within a generation the problem of creating ‘artificial intelligence’ will besubstantially solved” (Minsky, 1967, p. 2).On the other hand, there remains a compelling case for optimism. As we outlinebelow, it is not difficult to construct back-of-the-envelope scenarios in which even a modestnumber of currently existing technologies could combine to substantially raise productivitygrowth and societal welfare. Indeed, knowledgeable investors and researchers are bettingtheir money and time on exactly such outcomes. Thus, while we recognize the potential forover-optimism—and the experience with early predictions for AI makes an especiallyrelevant reminder for us to be somewhat circumspect in this essay—we judge that it wouldbe highly preliminary to dismiss optimism at this point.MismeasurementAnother potential explanation for the paradox is mismeasurement of output andproductivity. In this case, it is the pessimistic reading of the empirical past, not theoptimism about the future, that is mistaken. Indeed, this explanation implies that theproductivity benefits of the new wave of technologies are already being enjoyed but haveyet to be accurately measured. Under this explanation, the slowdown of the past decadeillusory. This “mismeasurement hypothesis” has been put forth in several works (e.g.,Mokyr, 2014; Alloway, 2015; Feldstein, 2015; Hatzius and Dawsey, 2015; Smith, 2015).There is a prima facie case for the mismeasurement hypothesis. Many newtechnologies, like smartphones, online social networks, and downloadable media involvelittle monetary cost, yet consumers spend large amounts of time with these technologies.8At least not yet:

Thus, the technologies might deliver substantial utility even if they account for a smallshare of GDP due to their low relative price. Guvenen, Mataloni, Rassier, and Ruhl (2017)also show how growing offshore profit shifting can be another source of mismeasurement.However, a set of recent studies provide good reason to think that mismeasurementis not the entire, or even a substantial, explanation for the slowdown. Cardarelli andLusinyan (2015); Byrne, Fernald, and Reinsdorf (2016); Nakamura and Soloveichik (2015);and Syverson (2017), each using different methodologies and data, present evidence thatmismeasurement is not the primary explanation for the productivity slowdown. After all,while there is convincing evidence that many of the benefits of today’s technologies are notreflected in GDP and therefore productivity statistics, the same was undoubtedly true inearlier eras as well.Concentrated Distribution and Rent DissipationA third possibility is that the gains of the new technologies are already attainable,but that through a combination of concentrated distribution of those gains and dissipativeefforts to attain or preserve them (assuming the technologies are at least partiallyrivalrous), their effect on average productivity growth is modest overall, and is virtually nilfor the median worker. For instance, two of the most profitable uses of AI to date have beenfor targeting and pricing online ads, and for automated trading of financial instruments,both applications with many zero-sum aspects.One version of this story asserts that the benefits of the new technologies are beingenjoyed by a relatively small fraction of the economy, but the technologies’ narrowlyscoped and rivalrous nature creates wasteful “gold rush”-type activities. Both those seekingto be one of the few beneficiaries, as well as those who have attained some gains and seekto block access to others, engage in these dissipative efforts, destroying many of thebenefits of the new technologies. 9Recent research offers some indirect support for elements of this story. Productivitydifferences between frontier firms and average firms in the same industry have beenStiglitz (2014) offers a different mechanism where technological progr

Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics Erik Brynjolfsson, Daniel Rock, and Chad Syverson NBER Working Paper No. 24001 November 2017 JEL No. D2,O3,O4 ABSTRACT We live in an age of paradox. Systems using artificial intelligence match or surpass human level performance in more and more domains, leveraging rapid advances in other .

Related Documents:

Silat is a combative art of self-defense and survival rooted from Matay archipelago. It was traced at thé early of Langkasuka Kingdom (2nd century CE) till thé reign of Melaka (Malaysia) Sultanate era (13th century). Silat has now evolved to become part of social culture and tradition with thé appearance of a fine physical and spiritual .

May 02, 2018 · D. Program Evaluation ͟The organization has provided a description of the framework for how each program will be evaluated. The framework should include all the elements below: ͟The evaluation methods are cost-effective for the organization ͟Quantitative and qualitative data is being collected (at Basics tier, data collection must have begun)

̶The leading indicator of employee engagement is based on the quality of the relationship between employee and supervisor Empower your managers! ̶Help them understand the impact on the organization ̶Share important changes, plan options, tasks, and deadlines ̶Provide key messages and talking points ̶Prepare them to answer employee questions

Dr. Sunita Bharatwal** Dr. Pawan Garga*** Abstract Customer satisfaction is derived from thè functionalities and values, a product or Service can provide. The current study aims to segregate thè dimensions of ordine Service quality and gather insights on its impact on web shopping. The trends of purchases have

On an exceptional basis, Member States may request UNESCO to provide thé candidates with access to thé platform so they can complète thé form by themselves. Thèse requests must be addressed to esd rize unesco. or by 15 A ril 2021 UNESCO will provide thé nomineewith accessto thé platform via their émail address.

Chính Văn.- Còn đức Thế tôn thì tuệ giác cực kỳ trong sạch 8: hiện hành bất nhị 9, đạt đến vô tướng 10, đứng vào chỗ đứng của các đức Thế tôn 11, thể hiện tính bình đẳng của các Ngài, đến chỗ không còn chướng ngại 12, giáo pháp không thể khuynh đảo, tâm thức không bị cản trở, cái được

IN ARTIFICIAL INTELLIGENCE Stuart Russell and Peter Norvig, Editors FORSYTH & PONCE Computer Vision: A Modern Approach GRAHAM ANSI Common Lisp JURAFSKY & MARTIN Speech and Language Processing, 2nd ed. NEAPOLITAN Learning Bayesian Networks RUSSELL & NORVIG Artificial Intelligence: A Modern Approach, 3rd ed. Artificial Intelligence A Modern Approach Third Edition Stuart J. Russell and Peter .

BCS Foundation Certificate in Artificial Intelligence V1.1 Oct 2020 Syllabus Learning Objectives 1. Ethical and Sustainable Human and Artificial Intelligence (20%) Candidates will be able to: 1.1. Recall the general definition of Human and Artificial Intelligence (AI). 1.1.1. Describe the concept of intelligent agents. 1.1.2. Describe a modern .