Irving Wladawsky-Berger

Why Some AI Efforts Succeed While Many Fail

Winning with AI, - a 2019 report based on a survey jointly conducted by the MIT Sloan Management Review and the Boston Consulting Group, - found that 90% of respondents agree that AI represents a business opportunity for their company.  The global survey attracted over 2,500 respondents from 29 industries and 97 countries, and conducted interviews with 17 executives leaders of AI initiatives in large organizations.

The report classified the total survey population into four subgroups based on their understanding of AI tools and concepts and their levels of adoption of AI applications: Pioneers (20%) are leading-edge organizations that both understand and have widely adopted AI; Investigators (30%) understand AI but have not deployed applications beyond the pilot stage; Experimenters (18%) are learning by doing, conducting pilots without a deep understanding of AI; and Passives (32%) have not adopted AI and have little understanding of the technology.

“Many AI initiatives fail,” was the report’s overriding finding.  “Seven out of 10 companies surveyed report minimal or no impact from AI so far.  Among the 90% of companies that have made at least some investment in AI, fewer than 2 out of 5 report obtaining any business gains from AI in the past three years. This number improves to 3 out of 5 when we include companies that have made significant investments in AI.  Even so, this means 40% of organizations making significant investments do not report business gains from AI.”

Innovation and National Security in the 21st Century

“Countries that can harness the current wave of innovation, mitigate its potential disruptions, and capitalize on its transformative power will gain economic and military advantages over potential rivals,” was the top finding of the Innovation and National Security Task Force which was commissioned by the Council of Foreign Relations (CFR) to assess the current state of US technological innovation.  The Task Force noted that leadership in innovation, research and technology since World War II has made the US the most secure and economically prosperous nation on earth.

“Today, this leadership position is at risk,” the Task Force warned.  Federal support and funding for R&D has stagnated over the past two decades.  “Washington has failed to maintain adequate levels of public support and funding for basic science.  Federal investment in R&D as a percentage of GDP peaked at 1.86 percent in 1964 but has declined from a little over 1 percent in 1990 to 0.66 percent in 2016.”

The US has successfully responded to technological competition in the past.  Sputnik was a tipping point in the space race with the Soviet Union during the cold war, leading to, among other things, a significant increase in the number of graduate student fellowships in STEM disciplines, of which I was personally a beneficiary.  Another prominent example is the strong economic competition from Japan in the 1980s, which the US fended off a decade later with our leadership in the Internet and other major digital innovations.

The Internet of Things is Changing the World

“How the world will change as computers spread into everyday objects,” is the title of the lead article in a comprehensive review of the Internet of Things (IoT) in a recent issue of The Economist.  Like many technological advances, IoT has been long in coming.  Ubiquitous or Pervasive Computing dates back to the 1990s, when neither the necessary low-cost devices and wireless networks were anywhere near ready.  “But the transformation is about to go into overdrive, notes The Economist.  “One forecast is that by 2035 the world will have a trillion connected computers, built into everything from food packaging to bridges and clothes.”

IoT promises to bring many benefits, including a whole new generation of innovative products.  We’ve long thought of products as physical entities built with a combination of mechanical and electrical components, - e.g., appliances, cars, agricultural machines, industrial equipment, - some quite simple and some highly complex.  IoT has been ushering a whole new class of smart connected products.  In addition to mechanical and electrical components, these new products use digital components like microprocessors, sensors, data storage, software, and connectivity in a wide variety of ways.

As the world’s digital and physical infrastructures converge, digital technologies are being designed right into all kinds of consumer and industrial products.  Internet-connected smart doorbells, for example, include motion sensors and video cameras that notify a home owner when someone arrives at the door.  Using a smartphone app, the home owner can watch and talk to the visitor, while a video of their interaction is saved for an added level of security.

“The magic of computers is that they provide in a machine an ability - to calculate, to process information, to decide - that used to be the sole preserve of biological brains,” writes The Economist.  “The IoT foresees a world in which this magic becomes ubiquitous.  Countless tiny chips will be woven into buildings, cities, clothes and human bodies, all linked by the internet… Over the past century electricity has allowed consumers and businesses at least in the rich world, access to a fundamental, universally useful good - energy - when and where they needed it.  The IoT aims to do for information what electricity did for energy.”

The Puzzling Economic Impact of Transformative Technologies

“General purpose technologies (GPTs) are engines for growth,…” wrote Erik Brynjolfsson, Daniel Rock, and Chad Syverson in The Productivity J-Curve, a working paper recently published by the National Bureau of Economic Research (NBER).  “These are the defining technologies of their times and can radically change the economic environment.  They have great potential from the outset, but realizing that potential requires larger intangible and often unmeasured investments and a fundamental rethinking of the organization of production itself.”

As we’ve learned over the past two centuries, there’s generally been a significant time lag between the broad acceptance of a major new transformative technology and its ensuing impact on companies, governments and other institutions.  Even after reaching a tipping point of market acceptance, it takes considerable time, - often decades, - for these new technologies and business models to be widely embraced across economies and societies, - and for their full benefits to be realized.

In her 2002 influential book, Technological Revolutions and Financial Capital, economic historian Carlota Perez wrote that since the advent of the Industrial Revolution, we’ve had a major technological revolution every 60 years or so.  First was the age of machines and factories in the latter part of the 18th century.  This was followed by the age of steam, coal and iron in the early to mid 19th century; electricity and steel around the 1870s-1880s; and automobiles, oil and mass production in the early decades of the 20th century.  Then came the computer and communications revolution in the latter part of the 20th century, ushering the transition from the industrial economy of the past two centuries to our present digital economy.

According to Perez, the economic transformations accompanying these technologies are composed of two distinct periods, each lasting roughly 20 to 30 years.  First comes the installation period when the new technologies emerge into the marketplace, entrepreneurs launch many new startups, and venture capitalists encourage experimentation with new business models. This is then followed by the deployment period, when the now well accepted technologies and business models become the norm, leading to long-term economic and productivity growth.

Can AI Help Develop and Execute a Competitive Business Strategy?

Strategy For and With AI, a recent article in the MIT Sloan Management Review by David Kiron and Michael Schrage, argues that while formulating a comprehensive strategy for the use of AI technologies is absolutely necessary, it’s not sufficient.  “Creating strategy with AI matters as much - or even more - in terms of exploring and exploiting strategic opportunity.”  The article is based on a survey of over 3,000 executives, managers and analysts from companies in over 100 countries and 20 industries, as well as on interviews with executives and academics.

AI is expected to be the biggest commercial opportunity for companies and industries over the next couple of decades.  Two 2018 reports, one by PwC and one by McKinsey, estimated that AI has the potential to incrementally add around $13-$15 trillion to global economic output by 2030.  The growth will come from productivity gains, - e,g, the continuing automation of routine tasks, and AI-based tools to augment human capabilities, - and from increasingly sophisticated, AI-enhanced products, services, and system-wide applications.

Machine learning advances, like deep learning, have played a central role in AI’s recent achievements, giving computers the ability to be trained by ingesting and analyzing large amounts of data instead of being explicitly programmed.  Machine learning methods for solving problems, - subtle adjustments to the numerical weights that interconnect huge number of artificial neurons, - are radically different and complement the way humans solve problems.  But, while being a very powerful tool, machine learning, and AI in general, lack all-important human qualities like common sense and empathy, so their decisions and suggestions should be carefully reviewed by humans.  People and technology must closely collaborate, with each playing the particular role they are best at.

Every day we can read about the latest AI advances from research labs, startups and large companies.  AI technologies are approaching or surpassing human levels of performance in vision, speech recognition, language translation, playing championship-level Go, and the early detection and diagnosis of various forms of cancer.  But, can AI help address broad, open-ended and ambiguous problems like developing and executing a competitive business strategy?

The Increasing Demand for Hybrid, “T-Shaped” Workers

A recent article in The Atlantic used the USS Gabrielle Giffords to illustrate the important changes taking place in the US Navy, - and in the world of work in general.  After discussing various features of its advanced design, the article noted that the ship’s most futuristic aspect is its crew.  “It was designed to operate with a mere 40 souls on board - one-fifth the number aboard comparably sized legacy ships and a far cry from the 350 aboard a World War II destroyer.  The small size of the crew means that each sailor must be like the ship itself: a jack of many trades and not, as 240 years of tradition have prescribed, a master of just one,” noted the article.  Due to increasingly sophisticated technologies and the high cost of personnel, the USS Gabrielle Giffords is among a new class of Navy ships that’s turned away from specialists in favor of so-called hybrid sailors who are able to handle multiple tasks and rapidly acquire new skills.

The replacement of highly specialized workers with problem-solving generalists who are able to handle and quickly learn new tasks has been taking place across a wide variety of  occupations and industries.  Increased productivity and lower costs, i.e., doing more with less, are some of the reasons for this change.  But, the increased complexity of products, applications and systems, and the unanticipated problems that often follow are other important reasons.

The concept of hybrid or T-shaped workers was first introduced around 25 years ago as a metaphor to describe the kind of individuals that the Navy and many other organizations are looking for.  The vertical stroke in the letter T represents a depth of skills and expertise in one or more specific fields; the horizontal bar implies broad multidisciplinary and social skills, as well as the ability to collaborate with experts across disciplines to jointly solve complex problems.

Impact of AI on Government - Opportunities and Challenges

A few weeks ago I wrote about Leveraging AI to Transform Government based on a 2018 report by the Partnership for Public Service and the IBM Center for the Business of Government.  The report illustrates how government is using AI by describing four use cases being pursued by different agencies.  In 2019, the Partnership and IBM collaborated on two additional reports, Assessing the Impact of AI on the Work of Government, and Building Trust, Managing Risks.

This first report is focused on the impact of AI on the federal workforce, based on an analysis of the jobs and tasks most susceptible to automation, as well as on roundtables and interviews with over 40 AI experts.  The overriding conclusion is that, over time, AI will have a deep impact on how government works.  More than 80 different federal occupations, employing over 130,000 workers, are likely be significantly transformed, including 30,000 in the Treasury Department, 16,000 in Defense, and 5,000 in Transportation, as well as a variety of occupations in independent agencies like the Securities and Exchange Commission and the Government Publishing Office.

In particular, AI will have a major impact in three key areas: the transformation of the federal workday, the potential for personalized customer service and the increased importance of technical and data skills.  Let me summarize the findings and recommendations in each of these areas.

Leveraging AI to Transform Government

“AI has the potential to help address many of our country’s pervasive problems and advance our safety, health and well-being,” said The Future Has Begun, a report on the impact of AI on government by the Partnership for Public Service and the IBM Center for the Business of Government.  “That promise has great potential to transform government.”

 “The most challenging problems AI may help us solve - from fighting terrorists to serving vulnerable populations - will involve government.  More immediately, though not less consequentially, AI will change the way public servants do their jobs.  Because of the impact AI could have on how governments ensure our safety and further our well-being, government leaders are likely to play a significant role in dictating future directions for AI.”

To illustrate how government is using AI, the report is focused on four concrete use cases being pursued by different agencies.  In-depth interviews were conducted with 14 people from 10 organizations, including the various government agencies involved, as well as their collaborators in non profits and universities.  Let me briefly summarize each of these four case studies.

The MIT Work of the Future Report

“The world now stands on the cusp of a technological revolution in artificial intelligence and robotics that may prove as transformative for economic growth and human potential as were electrification, mass production, and electronic telecommunications in their eras,” said the MIT Task Force on the Work of the Future in its recently released interim report.  The final report will be issued after conducting additional research over the next year.

The Task Force was convened in the spring of 2018 by MIT President Rafael Reif to address what may well be the most critical question of the digital economy: as emerging technologies raise aggregate economic output and the wealth of nations, will they also enable people to attain higher living standards, better working conditions, greater economic security, and improved health and longevity?

The report’s overriding conclusion is that the likelihood that AI and automation will wipe out major workforce sectors in the near future is exaggerated.  However, we’ve already seen important reasons for concern, especially the rising polarization of employment and wage distribution over the past few decades, which has disproportionately benefited high-skilled professionals while reducing opportunities for mid- and low-skilled workers.

“[A] critical challenge is not necessarily a lack of jobs, but the low quality of many jobs and the resulting lack of viable careers for many people, particularly workers without college degrees.  With this in mind, the work of the future can be shaped beneficially by new policies, renewed support for labor, and reformed institutions, not just new technologies. Broadly, the task force concludes, capitalism in the U.S. must address the interests of workers as well as shareholders.”

How to Support the Widespread Adoption of AI

“Artificial intelligence is reshaping business - though not at the blistering pace many assume,” notes Building the AI-Powered Organization, an article in Harvard Business Review by McKinsey partners Tim Fountaine, Brian McCarthy, and Tamim Saleh in its opening sentence.  A number of studies agree that AI is the biggest commercial opportunity for companies and industries over the next 10-15 years.  Two recent reports by McKinsey and PwC respectively concluded that AI has the potential to incrementally increase global economic output between now and 2030 by an additional $13 to $15 trillion.

“Yet, despite the promise of AI, many organizations’ efforts with it are falling short,” adds the HBR article.  Surveys with thousands of executives, showed that most firms are only using AI in ad hoc pilots or applying it to a single business process.  Only 8% of firms are engaged in practices that support widespread adoption.

Why the slow progress?  “At the highest level, it’s a reflection of a failure to rewire the organization.  In our surveys and our work with hundreds of clients, we’ve seen that AI initiatives face formidable cultural and organizational barriers…  at most businesses that aren’t born digital, traditional mindsets and ways of working run counter to those needed for AI.”

To support the widespread adoption of AI, companies must make three fundamental shifts.

The State of AI in the Enterprise

A few months ago, Babson College professor Tom Davenport gave a talk on the state of AI in the enterprise at the annual conference of MIT’s Initiative on the Digital Economy.  His talk was based on two recent US surveys conducted by Deloitte, the first one in 2017 followed by a second in 2018.  Davenport was a co-author of both reports.

The 2017 survey was focused on the responses of 250 US executives who were leading the applications of AI in their companies.  The larger 2018 survey reached out to 1,100 IT (46%) and line-of-business (54%) executives from US-based companies (64% at the C-level) and 10 different industries.  All of these respondents were early AI adopters compared with their counterparts in an average company, - 90% were directly involved in their company’s AI projects, and 75% said that they had an excellent understanding of AI.

Davenport started his talk by summarizing the key findings in the Deloitte surveys:

• 20-30% of enterprises are early adopters, having implemented at least one AI prototype or production application;
  
• Many projects are in pilots but some are already in production;
• Relatively simple low hanging fruit projects prevail over more ambitious and complex moon shots;
  
• Only 24% cited “reducing headcount through automation” as one of their top AI priorities;
• The great majority of respondents believe that AI leads to moderate or substantial changes in job roles and skills;
• Implementation, integration, data issues and talent top the list of challenges faced by early adopters;
  
• Further AI growth is inevitable.

Using Agile Processes to Develop AI-Based Solutions

In the waterfall development model, a project is broken down into a set of sequential phases, - e.g., conception, analysis, design, construction, testing, deployment, maintenance, - each dependent on the completion of the previous phase.  Such a model is most appropriate for complex engineering projects, like bridges, airplanes and computer hardware.  It’s also quite applicable for mission critical software and infrastructure projects where security, availability, scalability and other ilities are paramount.

But, a serial, orderly approach isn’t appropriate for projects where time-to-market, usability, and flexibility are the highest priorities.  For such projects, an agile development approach is much more suitable.  Key to the agile approach is the involvement of the product’s target end-users early in the process to make sure that the product under development is responsive to their requirements. 

Agile development is evolutionary in nature, based on the early availability of experimental prototypes, user feedback, continuous improvements, and rapid responses to technology, market and user requirements.  It’s not surprising that agile software development took off in the 1990s, when the explosive rise of the Internet brought millions and millions of users online, lots of new user-oriented applications, and short time-to-market cycles in response to fierce competition.

Our increasingly sophisticated smart machines bring with them a whole new set of development challenges.  These AI-based systems are generally software-intensive, user-centric, and data-driven.  Their complexity goes way up the smarter we want the systems to be.  The behavior of an AI system based on machine learning depends on the information we use to train its algorithms rather than on the precise set of software instructions that tell the computer what to do.

AI systems are designed to sense, respond and adapt to a changing environment.  Their development feels sort of like bringing up and sending a teen-ager into the world.  Given that we want them to be able to handle unanticipated situations, they must have a fair degree of flexibility and autonomy.  But, it’s hard to be sure that the system (or the teen-ager) will always behave the way we’d like them to.

A Quarter Century into the Digital Age

The digital age was born around 25 years ago with the public release of the Netscape browser in 1994.  The browser made it much, much easier for the average person to access information over the Internet, sparking the explosive growth of users, websites and online applications.  This was a truly historical point, marking the transition from the industrial economy of the past couple of centuries to a new kind of digital economy.   Ever since those days, the digital age has continued to advance driven by a stream of new technologies, and business models, from smartphones and cloud to big data and machine learning.

A few months ago, the McKinsey Global Institute published Twenty-five Years of Digitization, a report offering ten insights to help business leaders navigate this still relatively new digital world.  Let me first list the ten insights, and then briefly comment on a few of them.

• Large economic potential is linked to digitization - and much of it is yet to be captured;
• Digital superstars are rising far beyond the US big four and China’s big three;
• Digital natives are calling the shots;
• Digital changes everything - even industry boundaries;
• Agile is the new way to compete;
• Playing the platform economy is an in-the-money option;
• Self-cannibalization and innovation are a necessity for digital reinvention;
• Going after the right M&A is key;
• Effective management of digital transformation is vital - but challenging; and
  
• Leveraging, and transitioning from, digital to new frontier technologies is an imperative.

The Challenges of Automation in a Fast Changing Economy

“Technological innovation should be embraced,” notes Automation and a Changing Economy, a recently published report by The Aspen Institute.  “Automation has been a largely positive economic and social force, and looking forward,   automation will be necessary to feed, house, and raise the living standards of a growing and aging population.” 

But, “While it is important to note that automation’s overall, long-run effect on the economy has been positive - more jobs, more growth, higher living standards - this does not negate the disruptive impact of automation on individuals and communities, which results from displacements, changing skill needs, and income inequality.”

The report is divided into two sections: The Case for Action, which explores how automation impacts the economic security and opportunity of American workers, and Policies for Shared Prosperity, which outlines a policy agenda for addressing automation’s challenges and opportunities.  Let me summarize the key conclusions and recommendations in each of the report’s two sections.

AI and the Evolution of History

Several weeks ago, I wrote about the event I attended on February 28 to celebrate the launch of MIT’s new Schwarzman College of Computing, - MIT’s strategic response to the rise of artificial intelligence, - a technology that will reshape “geopolitics, our economy, our daily lives and the very definition of work” in the decades to come.  The all-day celebration featured talks and panels on a wide variety of topics, some focused on innovative applications of AI technologies, others on the challenging issues raised by these powerful technologies.

A few weeks later I wrote about one of these challenging issues, the impact of AI on our social interactions, based on the talk by MIT professor Sherry Turkle on Rethinking Friction in Digital Culture, and a related article by Yale professor Nicholas Christakis on How AI Will Rewire US.

I now want to discuss the interview conducted by NY Times columnist Thomas Friedman with former US Secretary of State Dr. Henry Kissinger at the February 28 event.  The interview, - which can be seen in this video, - was based on a June, 2018 article by Dr. Kissinger in The Atlantic, - How the Enlightenment Ends: “Philosophically, intellectually - in every way - human society is unprepared for the rise of artificial intelligence.”

The Third Wave of the Digital Economy - Promises and Challenges

“The next wave of digital innovation is coming.  Countries can welcome it, prepare for it, and ride it to new heights of innovation and prosperity, or they can ignore the changing tide and miss the wave,” writes Robert Atkinson in The Task Ahead of Us.  Atkinson is founder and president of the Information Technology and Innovation Foundation (ITIF), a think tank focused on science and technology policy.

We’re now entering the third wave of the digital economy, says Atkinson.  The first was based on personal computing, the Internet, Web 1.0, and e-commerce.  The second brought us Web 2.0, big data, smartphones and cloud computing.  The emerging third wave promises to be significantly more connected, - including higher bandwidth and a wide variety of devices; more automated, - with more work being done by machines while integrating the physical and digital worlds; and more intelligent, - leveraging huge volumes of data and advanced algorithms to help us understand and deal with our increasingly complex world.

“Building and adopting the new connected, automated, and intelligent technology system will lead to enormous benefits globally, not least of which will be robust rates of productivity growth and improvements in living standards.  Moreover, these technologies will help address pressing global challenges related to the environment, public health, and transportation, among others.”

We’re in the early stages of this third wave.  5G, IoT, robotics, AI and other promising technologies are being embraced by early marketplace adopters, but their full-scale impact is still 5 to 10 years away.  We’re in a period not unlike the late 1980s, when it was clear that IT was on the brink of a major transition, but the Internet revolution didn’t arrive until the mid-1990s.

The Top Ten Emerging Technologies of 2018

For the past several years, the World Economic Forum (WEF) has published an annual list of the Top Ten Emerging Technologies that would be potentially disruptive over the next three to five years while also providing significant benefits to economies and societies.  Starting in 2017, the Top Ten list has been chosen in collaboration with Scientific American and published in an issue of the magazine.

The Top Ten Emerging Technologies 2018 was released in September, 2018.  The final list was selected by a steering group of experts from an initial list with more than 50 submissions.  Here are the 10 technologies comprising the 2018 list, along with the reasons cited in the report for their selection:

What Might Work Look Like in 2030?

A number of recent studies have taken a close look at the the future of work over the next 10 to 15 years.  For example, a December, 2017 report by McKinsey examined which jobs will likely be displaced by automation through 2030, as well as which jobs are likely to be created over the same period, based on data from 46 countries.

McKinsey’s overall conclusion was that a growing technology-based economy will create a significant number of new occupations which will more than offset declines in occupations displaced by automation.  However, “while there may be enough work to maintain full employment to 2030 under most scenarios, the transitions will be very challenging - matching or even exceeding the scale of shifts out of agriculture and manufacturing we have seen in the past.”

Another report, The Changing Nature of Work, - published by the World Bank in October, 2018, - concluded that our problem is not that there won’t be enough work in the future.  The key problem is that, in many countries, the workforce isn’t prepared for a fast unfolding future. Neglecting investments in human capital, - i.e., the sum total of a population’s health, skills, knowledge, and experience, - will dramatically weaken a country’s competitiveness.

Let me now discuss Eight Futures of Work Scenarios and their Implications, a white paper published in January, 2018 by the World Economic Forum (WEF) in collaboration with the Boston Consulting Group (BCG).  The paper considered eight distinct scenarios on what work might be like by 2030, based on different combinations of three of the most impactful and uncertain variables affecting the future of work: the rate of technological change; the evolution of learning in the workforce; and the magnitude of talent mobility across geographies.

AI and our Social Interactions

I recently wrote about the event I attended on February 28 to celebrate the launch of MIT’s Schwarzman College of Computing.  This new interdisciplinary College is MIT’s response to the rise of artificial intelligence, - a profoundly powerful technology that will likely reshape our economy, society and personal lives in the decades to come.  But, attaining AI’s broad potential will require not only a continuing slew of technological innovations, but equally important research on the challenges our society must get ready for, including ethical issues, workplace disruptions, and human interactions.

The event included keynotes and panels on a wide variety of topics. One of the talks I found most interesting was Rethinking Friction in Digital Culture by Sherry Turkle, MIT Professor of the Social Studies of Science and Technology.

In her many writings and seminars, - including her 2011 best-seller Alone Together, -  Professor Turkle has long been warning about the impact of technology on our social interactions.  “Most of us here today were introduced to the idea of friction free as a really good thing,” she said at the start of her keynote.  “It’s an aesthetic of engineering efficiency, so why shouldn’t it be a really good thing?”  This idea that technical things should be friction free and smooth easily bleeds into other domains. “Efficiency becomes aspirational - in politics, in business, in education, and in our thinking about relationships.”

She cited a few examples of the kinds of friction free interactions she’s come across in her research on technology and social relationships.  People of all ages have told her that they prefer to text rather than talk, - even when interacting with a colleague in the next cubicle or office.  People have said to her that they prefer to text their spouse rather than have a face-to-face conversation.  Friction free, she explained, “is usually tied up with a hope for greater efficiency and less vulnerability.”

The Productivity Paradox: Digital Abundance and Scarce Genius

Despite the relentless advances of digital technologies, productivity growth has been declining over the past decade.  Investment and interest rates have remained low, and income has continued to stagnate for the majority of workers in the US and other developed economies. The world seems to be stuck in a period of slow growth and no one is quite sure what’s causing this apparent contradiction.

Opinions abound.  Harvard economist Larry Summers has argued that secular stagnation is the reason behind this unusual situation, caused primarily by a reluctance of companies to invest and of consumers to spend, with the ensuing excessive savings acting as a drag on economic growth.

Some contend, - most prominently Northwestern University economist Robert Gordon, - that over the past few decades there’s been a fundamental decline in innovation and productivity.  Perhaps our current technologies, advanced and exciting as they might be, aren’t as transformative as the technologies from the period between 1870 and 1970, when we experienced high productivity growth and a rising standard of living. 

Demographic change, namely, the declining population and labor force growth around the world, is another potential cause for slow economic growth.  The global labor force grew at an average of 1.8% per year between 1960 and 2005, but since then it’s been growing at just 1.1% per year, and it’s already shrinking in a number of countries, including Japan, Germany and China.

Another possible explanation, - advanced by Erik Brynjolfsson, Daniel Rock, and Chad Syverson in AI and the Modern Productivity Paradox, - is that the paradox is primarily due to a time lag between technology advances and their impact on the economy.  We’ve been living in a time of major transformative technologies, - the Internet, smartphones, IoT, big data, AI - whose deployment and impact on productivity growth are still lagging.

While technologies may advance rapidly, humans and our institutions change slowly.  Moreover, the more transformative the technologies, the longer it takes for them to be embraced by companies and industries across the economy.  Leading edge firms are already benefiting from these advances, but most are still in the early learning stages.  Translating technological advances into productivity gains requires major transformations in business processes, organization and culture, - and these take time.

Impact of an Aging Population on Economic Growth and Automation

The world has been stuck in an era of slow economic growth over the past decade.  Top economists and policy makers have proposed a number of explanations for the economic slowdown, but in the end, there’s no consensus on the reasons, on how long the slowdown will likely last, or on what to do about it.

Economic growth has two main components, productivity growth and the growth of the labor force.  One potential theory for our stagnant economic growth is demographics, namely, the rapidly declining population and labor force growth around the world.

The current world’s population is 7.7 billion.  The UN Population Division estimates that global population will reach 9.8 billion with a growth rate of 0.5% in 2050, and will peak around 2100 at roughly 11.2 billion with a growth rate of 0.1%.  Other projections estimate that the population will grow slower, peak at 9.4 billion around 2070 and then decline to around 9 billion in 2100.  Some project that after reaching a peak of 9 billion, the world’s population will decline back to today’s levels, - around 7 billion, - by 2100.

Japan and a few other countries are already experiencing population declines.  The US population is currently around 330 million, and is projected to reach 390 million in 2050 and 450 million in 2100.  While the US birthrate hit a historic low in 2017, the population has continued to grow due to immigration.  The US growth rate is 0.71% including immigration and 0.43% without.

The global labor force grew at an average of 1.8% per year between 1960 and 2005, but since then it’s been growing at just 1.1% per year.  The labor force is still growing in some developing countries like India, Nigeria and the Philippines, but it’s already shrinking in China, Japan and Germany.  In the US, the labor force is growing very slowly, - 0.5 percent per year over the past decade, compared with 1.7 percent from 1960 to 2005.  Given the continuing decline in fertility rates in most parts of the world, the global labor force is expected to slip further in the coming decades.

The New, Interdisciplinary, MIT Schwarzman College of Computing

A few months ago, MIT president Rafael Reif announced the launch of the Stephen A. Schwarzman College of Computing in an e-mail to the MIT community.  “This new College is our strategic response to a global phenomenon – the ubiquity of computing and the rise of AI…” wrote President Reif.  “To state the obvious, AI in particular is reshaping geopolitics, our economy, our daily lives and the very definition of work.  It is rapidly enabling new research in every discipline and new solutions to daunting problems.  At the same time, it is creating ethical strains and human consequences our society is not yet equipped to control or withstand.”

The new College of Computing is made possible by a $350 million foundational gift from Stephen Schwarzman, - chairman, CEO and co-founder of Blackstone, one of the world’s leading investment firms.  The gift is part of a  $1 billion commitment to reshape MIT to help it better address the continuing advances in technology in our 21st century digital economy, as well as their many applications , - both the exciting opportunities and the difficult challenges.

The College is envisioned as an interdisciplinary hub across all five of MIT’s existing schools: Engineering; Science; Humanities, Arts, and Social Sciences; Architecture and Planning; and the Sloan School of Management.  This MIT News announcement succinctly summarized the key objectives of the new College:

• reorient MIT to bring the power of computing and AI to all fields of study at MIT, allowing the future of computing and AI to be shaped by insights from all other disciplines

• create 50 new faculty positions that will be located both within the College and jointly with other departments across MIT - nearly doubling MIT’s academic capability in computing and AI

• give MIT’s five schools a shared structure for collaborative education, research, and innovation in computing and AI

• educate students in every discipline to responsibly use and develop AI and computing technologies to help make a better world; and

• transform education and research in public policy and ethical considerations relevant to computing and AI.

The Current State of AI Adoption

AI is seemingly everywhere.  In the past few years, the necessary ingredients have finally come together to propel AI beyond early adopters to a broader marketplace: powerful, inexpensive computer technologies; advanced algorithms; and huge amounts of data on almost any subject.  Newspapers and magazines are full of articles about the latest advances in machine learning and related AI technologies.

Two recent reports concluded that, over the next few decades, AI will be the biggest commercial opportunity for companies and nations.  AI advances have the potential to increase global GDP by up to 14% between now and 2030, the equivalent of an additional $14-15 trillion contribution to the world’s economy, and an annual average contribution to productivity growth of about 1.2 percent.

Over time, AI could become a transformative, general purpose technology like the steam engine, electricity, and the Internet. AI marketplace adoption will likely follow a typical S curve pattern, - that is, a relatively slow start in the early years, followed by a steep acceleration as the technology matures and firms learn how to best leverage AI for business value.

To get a better sense of the current state of AI adoption, McKinsey recently conducted a global online survey on the topic, garnering responses from over 2,000 participants across 10 industry sectors, 8 business functions and a wide range of regions and company sizes.  The survey asked about their progress in deploying nine major AI capabilities, including machine learning, computer vision, natural language text and speech processing, and robotic process automation.

Will Humans Be Better Off in an Increasingly AI-Based Future?

After decades of promise and hype, artificial intelligence is finally becoming one of the most important technologies of our era.  As AI continues to both spread and advance, will it enhance human capacities or will it lessen human autonomy and agency?  By 2030, will most people be better or worse off than they’re today?

To shed light on these very important questions, the Pew Research Center canvassed AI experts this past summer, including technologists developers, and business and policy leaders.  It received almost 1,000 responses, which formed the basis for its recently published report Artificial Intelligence and the Future of Humans.

Overall, 63% of respondents were optimistic, predicting that most individuals will mostly be better off, while 37% thought otherwise.  But optimistic or not, most experts expressed concerns about the long-term impact of AI on “the essential elements of being human.”

The survey further asked the experts to explain their hopes and fears, to sketch out how they envision human-machine interactions over the coming decade, and what actions should be taken to assure the best possible future. Many answered, and their detailed comments are organized into the report’s three main sections: concerns; suggested solutions; and expectations for the future.  Let me briefly discuss the key points in each section.

Beyond Machine Learning: Capturing Cause-and-Effect Relationships

Artificial intelligence is rapidly becoming one of the most important technologies of our era.  Every day we can read about the latest AI advances from startups and large companies.  AI technologies  are approaching or surpassing human levels of performance in vision, speech recognition, language translation, and other human domains. Machine learning advances, like deep learning, have played a central role in AI’s recent achievements, giving computers the ability to be trained by ingesting and analyzing large amounts of data instead of being explicitly programmed.

Deep learning is a powerful statistical technique for classifying patterns using large training data sets and multi-layer AI neural networks.  It’s essentially a method for machines to learn from all kinds of data, whether structured or unstructured, that’s loosely modeled on the way a biological brain learns new capabilities.  Each artificial neural unit is connected to many other such units, and the links can be statistically strengthened or decreased based on the data used to train the system.  Each successive layer in a multi-layer network uses the output from the previous layer as input.

Machine learning can be applied to just about any domain of knowledge given our ability to gather valuable data in almost any area of interest.  But, machine learning methods are significantly narrower and more specialized than humans.  There are many tasks for which they’re not effective given the current state-of-the-art.  In an article recently published in Science, professors Erik Brynjolfsson and Tom Mitchell identified the key criteria that help distinguish tasks that are particularly suitable for machine learning from those that are not.  These include:

• Tasks that map well-defined inputs to well-defined outputs, - e.g., labeling images of specific animals, the probability of cancer in medical record, the likelihood of defaulting on a loan application;
• Large data sets exist or can be created containing such input-output pairs, - the bigger the training data sets the more accurate the learning;
• The capability being learned should be relatively static, - If the function changes rapidly, retraining is typically required, including the acquisition of new training data; and
• No need for detailed explanation of how the decision was made, - the methods behind a machine  learning recommendation, - subtle adjustments to the numerical weights that interconnect its huge number of artificial neurons, - are difficult to explain because they’re so different from those used by humans.

Redefining Work - Leveraging Human Capabilities in a Future of Expanding Automation

How will labor markets evolve in our 21st century digital economy?  What’s the likely future of jobs, given that our increasingly smart machines are now being applied to activities requiring intelligence and cognitive capabilities that not long ago were viewed as the exclusive domain of humans?  How will AI, robotics and other advanced technologies transform the very nature of work?

Over the past few years, a number of papers, reports and books have addressed these very important questions.  They generally conclude that AI will have a major impact on jobs and the very nature of work.  For the most part,  they view AI as mostly augmenting rather than replacing human capabilities, automating the more routine parts of a job and increasing the productivity and quality of workers, so they can focus on those aspect of the job that most require human attention.  Overall, few jobs will be entirely automated, but automation will likely transform the vast majority of occupations. 

A recent McKinsey report noted while there will likely be enough work to maintain full employment by 2030, the transition will be very challenging, “on a scale not seen since the transition of the labor force out of agriculture in the early 1900s in the United States and Europe, and more recently in in China.”  The report estimated that “up to 375 million workers, or 14 percent of the global workforce, may need to change occupations - and virtually all workers may need to adapt to work alongside machines in new ways.”

Given these predictions about the changing nature of work, what should companies do?  How should firms prepare for a brave new world where we can expect major economic dislocations along with the creation of new jobs, new business models and whole new industries, and where many people will be working alongside smart machines in whole new ways?

The Jobs Outlook for 2022

A few weeks ago, the World Economic Forum (WEF) released The Future of Jobs Report 2018.  The report takes an in-depth look at the world of work in what the WEF calls the Fourth Industrial Revolution. The report is based on a survey of over 300 chief human resource officers and top strategy executives from large global companies across 12 industry sectors.  These companies collectively represent more than 15 million employees, and conduct business in 20 developed and emerging economies which collectively account for about 70% of global GDP.

First introduced at its 2016 annual Davos Forum, the WEF positions the Fourth Industrial Revolution within the historical context of three previous industrial revolutions.  The First, - in the last third of the 18th century, - ushered the transition from hand-made goods to mechanized, machine-based production based on based on steam and water power.  The Second, - a century later, - revolved around steel, railroads, cars, chemicals, petroleum, electricity, the telephone and radio, leading to the age of mass production.  The Third, - starting in the 1960s, - saw the advent of digital technologies, computers, the IT industry, and the automation of process in just about all industries.

“Now a Fourth Industrial Revolution is building on the Third…” wrote WEF founder and chairman Klaus Schwab in a December, 2015 Foreign Affairs article.  “The possibilities of billions of people connected by mobile devices, with unprecedented processing power, storage capacity, and access to knowledge, are unlimited.  And these possibilities will be multiplied by emerging technology breakthroughs in fields such as artificial intelligence, robotics, the Internet of Things, autonomous vehicles, 3-D printing, nanotechnology, biotechnology, materials science, energy storage, and quantum computing.”

The Future of Jobs 2018 examines the potential of these advanced technologies to eliminate and create new jobs, as well as to improve the current workplace and prepare people for emerging jobs. Rather than looking at the future of work over the longer term, the report chose to examine the 2018-2022 period, which is within the planning horizon of the companies and executives that participated in the survey.

The Impact of AI on the World Economy

A few weeks ago I wrote about the economic value of AI based on a recently published report by PwC.  The report’s overriding finding was that AI technologies and applications will increase global GDP by up to 14% between now and 2030, the equivalent of an additional $15.7 trillion contribution to the world’s economy.  According to PwC, AI is the biggest economic opportunity over the next 10-15 years.

I now want to discuss a September, 2018 report by the McKinsey Global Institute on the impact of AI on the world economy.

The McKinsey report is based on simulation models of the impact of AI at the country, sector, company and worker levels.  It looked at their adoption of five broad categories of AI technologies: computer vision, natural language, virtual assistants, robotic process automation, and advanced machine learning.  Its data sources included survey data from approximately 3,000 firms in 14 different sectors; around 400 potential AI use cases across a variety of industries and functions; AI’s potential to automate and transform 800 existing occupations in 46 countries; and economic data from a number of organizations including the United Nations, the World Bank, the Organization for Economic Co-operation and Development (OECD), and the World Economic Forum.

Below is a summary of the report’s key findings. 

The Economic Value of Artificial Intelligence

 PwC recently released a report on the potential economic value of AI to different regions and industry sectors around the world.  The report defined AI, “as a collective term for computer systems that can sense their environment, think, learn, and take action in response to what they’re sensing and their objectives.”  By this broad definition, AI includes the automation of physical and cognitive tasks; assisting people to perform tasks better and faster; helping humans make better decisions; and automating decision making with no human intervention.

The report’s overriding finding is that AI is the biggest commercial opportunity for companies, industries and nations over the next few decades.  PwC estimates that AI advances will increase global GDP by up to 14% between now and 2030, the equivalent of an additional $15.7 trillion contribution to the world’s economy.

Around $6.6 trillion of the expected GDP growth will come from productivity gains, especially in the near term.  These include the continued automation of routine tasks, and the development of increasingly sophisticated tools to augment human capabilities.  Companies that are slow to adopt these AI-based productivity improvements will find themselves at a serious competitive disadvantage.

But, over time, increased consumer demand for AI-enhanced offerings will overtake productivity gains and result in an additional $9.1 trillion of GDP growth by 2030.  Moreover,  network effects, - that is, the more data and better insights companies are able to gather, the more appealing the products and services they’ll be able to develop, - will further increase consumer demand.  AI front-runners will gain an enormous competitive advantage through their ability to leverage this rich supply of customer data to shape product developments and business models, making it harder for slower moving competitors to catch up. 

Social Physics and Cybercrime Detection

A few weeks ago I wrote about social physics, a new discipline that aims to help us better understand and predict the behavior of human groups.  Social physics is based on the premise that all event-data representing human activity, - e,g,, phone call records, credit card purchases, taxi rides, web activity, - contain a special set of group behavior patterns.  As long as the data involves human activity, - regardless of the type of data, the demographic of the users or the size of the data sets, - similar behavioral dynamics apply.  These patterns can be used to detect emerging behavioral trends before they can be observed by other data analytics techniques.

Physics, biology and other natural sciences have long relied on universal patterns or principles to detect a faint signal within a large data set, - i.e., the proverbial needle in a haystack.  It’s what has enabled the discovery of very short lived elementary particles in physics, - like the Higgs boson  in 2013, - amidst the huge amounts of data generated by high energy particle accelerators.  In biology, it’s given rise to DNA sequencing and its growing list of applications in medicine, biotechnology and other disciplines. 

It’s not surprising that evolutionary biology and natural selection have led to similar universal patterns in the behavior of human crowds.  Humans and our ancestors have evolved with the drive to learn from each other because it’s been a major part of our survival over millions of years.  And if a new behavior, - whether the result of an innovative idea like the discovery of tools, or a mutation like a larger brain size, - helps a human group better adapt to a changing environment, natural selection will favor the survival of that group over others.

Social physics originated in MIT’s Human Dynamics Lab based on research by professor Alex (Sandy) Pentland, his then postdoctoral associate Yaniv Altshuler and their various collaborators.  In 2014, Pentland and Altshuler co-founded Endor, an Israeli-based startup that leverages social physics methods to make fast accurate predictions by analyzing data derived from human behavior. 

Social Physics - Reinventing Analytics to Better Predict Human Behaviors

Social physics first emerged over 200 years ago as an attempt to understand society and human behavior using laws similar to those of the physical sciences.  But, it wasn’t until the past two decades that we finally had enough data, powerful computers and sophisticated mathematical algorithms to develop quantitative theories of human social interactions.  We were now able to reliably predict how large groups of people make decisions by analyzing how information and ideas flow from person to person.

“The engine that drives social physics is big data: the newly ubiquitous digital data now available about all aspects of human life,” wrote MIT professor Alex (Sandy) Pentland in his 2014 book Social Physics: How Good Ideas Spread - The Lessons from a New Science.  “Social physics functions by analyzing patterns of human experience and idea exchange within the digital bread crumbs we all leave behind as we move through the world, - call records, credit card transactions, and GPS location fixes, among others.  These data tell the story of everyday life by recording what each of us has chosen to do… Who we actually are is more accurately determined by where we spend our time and which things we buy, not just by what we say we do.”

The book was the result of over a decade’s worth of research in the Human Dynamics group at MIT’s Media Lab.  Pentland, along with his graduate students and research associates, partnered with a variety of companies to obtain and analyze real-world data, properly anonymized to protect the privacy of the individuals whose behavior was reflected in the data.  They eventually discovered that all event-data representing human activity contain a special set of social activity patterns regardless of what the data is about.  These patterns are common across all human activities and demographics, and can be used to detect emerging behavioral trends before they can be observed by any other technique. 

Deep Learning: Is it Approaching a Wall?

Deep learning and related machine learning advances have played a central role in AI’s recent achievements, giving computers the ability to be trained by ingesting and analyzing large amounts of data instead of being explicitly programmed.  In just the past two years, Google’s deep-learning-based AlphaGo defeated the world’s top Go players, surprising most AI experts who thought that it would take another 5 to 10 years to achieve such a milestone.  Similarly, when Google switched to its new deep learning AI system in late 2016, it achieved an overnight improvement in the quality of its machine translations roughly equal to the total gains that the previous program had accrued over its 10 year lifetime.

As is typically the case with major technology achievements, - e.g., the dot-com bubble, - deep learning has quickly climbed to the top of Gartner’s hype cycle, where all the excitement and publicity accompanying new, promising technologies often leads to inflated expectations, followed by disillusionment if the technology fails to deliver.  AI may be particularly prone to such hype cycles, as the notion of machines achieving or surpassing human levels of intelligence leads to feelings of wonder as well as fear.  Over the past several decades, AI has gone through a few such hype cycles, including the so-called AI winter in the 1980s that nearly killed the field.

In a recent article, Deep Learning: A Critical Appraisal, author and NYU professor Gary Marcus offers a serious assessment of deep learning.  He argues that, despite its considerable achievements over the past 5 years, deep learning may well be approaching a wall, an opinion apparently shared by University of Toronto professor Geoffrey Hinton, the so-called Godfather of Deep Learning.  

AI - the Creation of a Human-Centric Engineering Discipline

AI is rapidly becoming one of the most important technologies of our era.  Every day we can read about the latest AI advances from startups and large companies, including applications that not long ago were viewed as the exclusive domain of humans.  Over the past few years, the necessary ingredients have finally come together to propel AI beyond the research labs into the marketplace: powerful, inexpensive computer technologies; huge amounts of data; and advanced algorithms.

Machine learning  and related algorithms like deep learning, have played a major role in AI’s recent achievements.  Machine learning gives computers the ability to learn by ingesting and analyzing large amounts of data instead of being explicitly programmed.  It’s enabled the construction of AI algorithms that can be trained with lots and lots of sample inputs, which are subsequently applied to difficult AI problems like language translation, natural language processing, and playing championship-level Go.

Over the past several decades, AI has been besieged by rounds of hype which over-promised, under-delivered, and nearly killed the field.  Once more, a recent Gartner article positioned machine and deep learning at the top of their hype cycle, - when all the excitement, publicity and promising potential often leads to a peak of inflated expectations, risking falling into the trough of disillusionment if the technology fails to deliver.  Now that AI is finally reaching a tipping point of market acceptance, it’s particularly important to be cautious and not repeat past mistakes.

In Artificial Intelligence - The Revolution Hasn’t Happened Yet, UC Berkeley professor Michael I. Jordan aims to inject such a note of caution.  “AI has now become the mantra of our current era…  The idea that our era is somehow seeing the emergence of an intelligence in silicon that rivals our own entertains all of us  -  enthralling us and frightening us in equal measure.  And, unfortunately, it distracts us…  Whether or not we come to understand intelligence any time soon, we do have a major challenge on our hands in bringing together computers and humans in ways that enhance human life.”

An Agile, Incremental Approach to Business Process Automation

I’ve long believed that customer self-service was key to the success of e-business, - the IBM Internet strategy I was closely involved with in the mid-late 1990s.  It was quite revolutionary how easy it was to now do for yourself so many ordinary activities that previously required a phone call during office hours or a trip to a store or office.

Moreover, such e-business applications were relatively easy to develop.  By integrating their existing transaction and data base applications with a web front-end, - a strategy we succinctly described as Web + IT, - any business could now be in touch with its customers, employees, suppliers and partners at any time of the day or night, no matter where they were.  Companies were able to engage in their core activities in a more productive way by web-enabling their back-end systems, that is, linking their web front-ends to the presentation services of their back-end applications.

I was reminded of IBM’s e-business strategy when recently reading about Robotic Process Automation (RPA), a technology for automating business processes based on emulating the manual actions of a human at a keyboard.  RPA aims to improve the operational efficiency of office and service workers by automating tedious, repetitive, tasks, such as those associated with widely used horizontal processes in HR, Finance & Accounting and IT services.

RPA enables process automation at a fraction of the cost and time of classic software development.  Rather than automating a process by redesigning the overall back-end application, RPA interfaces with the back-end system by performing the same actions that a human does via the application’s user interface.  RPAs thus creates kind of software robots, or bots for short, that work alongside the humans.

What Machine Learning Can and Cannot Do

Artificial intelligence is rapidly becoming one of the most important technologies of our era.  Every day we can read about the latest AI advances from startups and large companies.  Over the past few years, the necessary ingredients have come together to take AI across the threshold: powerful, inexpensive computer technologies; huge amounts of data; and advanced algorithms, especially machine learning.  Machine learning has enabled AI to get around one of its biggest obstacles, - the so-called Polanyi’s paradox.

Explicit knowledge is formal, codified, and can be readily explained to people and captured in a computer program.  But, tacit knowledge, a concept first introduced in the 1950s by scientist and philosopher Michael Polanyi, is the kind of knowledge we’re often not aware we have, and is therefore difficult to transfer to another person, let alone capture in a computer program.

“We can know more than we can tell,” said Polanyi in what’s become known as Polanyi’s paradox.  This common sense phrase succinctly captures the fact that we tacitly know a lot about the way the world works, yet aren’t able to explicitly describe this knowledge.  Tacit knowledge is best transmitted through personal interactions and practical experiences.  Everyday examples include speaking a language, riding a bike, and easily recognizing many different people, animals and objects.

Machine learning, and related advances like deep learning, have enabled computers to acquire tacit knowledge by being trained with lots and lots of sample inputs, thus learning by analyzing large amounts of data instead of being explicitly programmed.  Machine learning methods are now being applied to vision, speech recognition, language translation, and other capabilities that not long ago seemed impossible but are now approaching or surpassing human levels of performance in a number of domains.

The Transformative Power of Analytics - The Houston Astros: a Case Study

In 2013, the Houston Astros finished the season with a 51-111 record, - the worst record in the history of the franchise, having also lost over 100 games in the previous two years.  In 2017, the Astros won 101 games, finished first in the American League West Division, and went on to win their first ever World Series.  How did the Astros go from worst team record to first World Series victory in only four years?

The answer, in a nutshell, is analytics.  In December, 2011 the Astros hired Jeff Luhnow from the St Louis Cardinals as their General Manager.  As VP of scouting and player development in St Louis, Luhnow built a strong analytics department and enjoyed great success as one of the top talent producers in baseball, helping the Cardinals win the 2011 World Series as well as reach the playoffs from 2012 to 2015.  Luhnow then brought his analytics and management skills to Houston and proceeded to transform the team into a top contender in a short few years.

Today, just every major sports team makes extensive use of sports analytics, but none more so than Major League Baseball (MLB) teams.  Baseball has long been a game of statistics, collecting and analyzing vast amounts of data going back to the founding of the MLB in the years following the Civil War. 

Top MLB players are constantly compared not just to other contemporaneous players, but to other top players across the long history of the game, - a history that was very nicely depicted in Baseball, Ken Burns excellent documentary miniseries. Babe Ruth, Willie Mays, Mickey Mantle, Hank Aaron and countless other players are never forgotten in the world of baseball statistics.  For those of us who are fans, - as I am, not surprisingly given my Cuban roots, - baseball’s rich history is a major part of our enjoyment of the game. 

Getting on the AI Learning Curve: A Pragmatic, Incremental Approach

After decades of promise and hype, AI is now seemingly everywhere.  Over the past several years, the necessary ingredients have come together to propel AI beyond the research labs into the marketplace: powerful, inexpensive computer technologies; huge amounts of data; and advanced algorithms including machine learning.

While AI is likely to become one of the most important technologies of our era, we’re still in the early stages of deployment, especially outside leading-edge technology companies.  But, as AI continues its rapid progress, it’s not too early to ask a few important questions: What is AI’s overall value to the economy? What are the biggest application opportunities?; and, what are AI’s most serious challenges and limitations?

To address these questions, McKinsey recently published a discussion paper on the marketplace potential of AI.  The paper is particularly focused on machine learning and related technologies, and is based on a detail analysis of more that 400 use cases across 19 industries and 9 business functions.

“Two-thirds of the opportunities to use AI are in improving the performance of existing analytical use cases,” is the paper’s overriding finding.  This is a very interesting insight.  AI is now being successfully applied to tasks that not long ago were viewed as the exclusive domain of humans, - machine translation, natural language processing, defeating the world’s top Go players, - but only 16% of the use cases studied by McKinsey are greenfield cases, where only machine learning techniques can be used.  For the vast majority of the use cases, 69%, the key value of machine learning is to improve performance beyond that provided by traditional analytic techniques.  And, in the remaining 15% of cases, machine learning provided limited additional performance over existing analytical methods. 

Education, Work and US Leadership in the 21st Century

A few weeks ago, the Council on Foreign Relations (CFR) released The Work Ahead - Machines, Skills, and U.S. Leadership in the Twenty-First Century, a report by a CFR-sponsored Task Force to assess the future of work and its implications for the US economy and national security.  The Task Force was led by Penny Pritzker, - former Secretary of Commerce and founder and chair of PSP Partners, and by John Engler, - former Governor of Michigan and interim President of Michigan State University.

Their comprehensive report, - over 150 pages and 276 citations, - explored in great depth the changing nature of work as global competition and advanced technologies like AI, robotics, and smart IoT devices transform jobs and careers.  It states right up front that “The most important challenge facing the United States - given the seismic forces of innovation, automation, and globalization that are changing the nature of work - is to create better pathways for all Americans to adapt and thrive.  The country’s future as a stable, strong nation willing and able to devote the necessary resources and attention to meeting international challenges depends on rebuilding the links among work, opportunity, and economic security.”

For much of the 20th century, technology and globalization helped lead America to its leadership position in the world stage.  Underpinning this US leadership was the American Dream, - the promise that anyone can get ahead and achieve success and prosperity through talent and hard work.  But this promise has been eroding for decades.  For many of its citizens, the American Dream has been steadily disappearing.

Technology advances have led to massive innovations and efficiencies, enabling companies to get the same work done with significantly fewer people.  At the same time, many US-headquartered companies are truly global, doing business all over the world.  Beyond outsourcing jobs to countries with lower labor costs they’ve been investing and adding jobs where the demand is highest, - which has often been in faster growing emerging markets.

The US thus faces two major challenges: “creating new work opportunities, better career paths, and higher incomes for its people, while developing a workforce that will ensure U.S. competitive success in a global economy that will continue to be reshaped by technology and trade.” 

The Impact of AI on R&D and Innovation

For the past couple of centuries, general-purpose technologies (GPTs) have been the key drivers of productivity and economic growth, - thanks to their rapid improvements in price and performance, pervasive applications, and ability to spur complementary innovations across a wide variety of industries.  The steam engine, electricity and the internal combustion engine are prominent examples of GPTs in the 18th and 19th century.  More recently, semiconductors, computers and the Internet have led to the digital revolution of the past several decades.

Beyond innovations in existing sectors, the rapidly improving price/performance of GPTs have led over time to the creation of whole new applications and industries.  For example, the steady declines in the price of electricity-generated power and the improvement in the efficiency of electric motors led to the radical transformation of manufacturing in the early part of the 20th century with the advent of the assembly line.  It also led to the creation of the consumer appliance industry, e.g., refrigerators, dishwashers, washing machines.  Similarly, as the semiconductor industry took off, it led to the historical transition from the industrial economy of the past two centuries to our ongoing digital economy.

How about artificial intelligence?  Beyond its use by leading edge technology companies, we’re still in the early stages of AI deployment.  It’s only been in the last few years that major advances in machine learning have taken AI from the lab to early adopters in the marketplace.  While considerable innovations and investments are required for its wider deployment, AI is likely to become one of the most important GPTs in the 21st century. 

Harnessing the Collective Intelligence of Humans and Machines

For over a decade, MIT professor Thomas Malone has been conducting pioneering research on the collective intelligence of groups.  Malone is the founding director of the MIT Center for Collective Intelligence, and has authored and co-authored a number of seminal articles on the subject.  A few weeks ago, Malone published Superminds: The Surprising Power of People and Computers Thinking Together, - a book that nicely explains the concept of collective intelligence and how our increasingly capable machines are now complementing the intelligence of groups of people.  The book illustrates these concepts with a wide variety of examples and case studies.

He defines a supermind as “a group of individuals acting in ways that seem intelligent.”  A supermind has measurable properties just as individual do.  It can have a life of its own beyond those of its individuals constituents.  And, as Malone points out, throughout history, groups of people working together as superminds have been responsible for the vast majority of human achievements. 

From Horseless Carriages to Driverless Vehicles

Autonomous vehicles (AVs) may well be the quintessential symbol of our AI/robotics age.  Cars are a major part of our daily lives.  A self-driven car is a concept that requires little explanation, something we can all quickly grasp.  It wasn’t that long ago that the notion on an AV driving us around while we read or sleep would have felt like the stuff of science fiction.  Having experimental AVs coursing through public roads in Silicon Valley, Pittsburgh and Phoenix is concrete evidence that our smart machines are achieving human-like intelligence, raising a number of important questions:  how long before AVs are all around us?; how will they impact our lives?; what unintended consequences might we have to deal with?; and what should be done to ensure that they arrive as safely and smoothly as possible?

These are among the questions addressed in the March 1 issue of The Economist, which includes a comprehensive special report on autonomous vehicles with seven articles on the subject.  The special report starts out with the assumption that whatever technological hurdles lie in their way will be eventually overcome.  But there are wider economic, social and public policy issues to be explored, starting with: what can we learn from the transition to horseless carriages in the 20th century that can now be applied to the transition to driverless cars?

Human-Machine Work Teams

Will there be enough work in the future?  What’s the likely impact of our continuing technology advances on jobs?  How will they impact productivity?  These are all-important questions to reflect on as our increasingly smart technologies are now being applied to activities that not long ago were viewed as the exclusive domain of humans.

While no one really knows the answer to these questions, most studies of the subject have concluded that relatively few occupations, 10% or less, will be entirely automated and disappear over the next 10 - 15 years.  Instead, a growing percentage of occupations will significantly change as technologies automate the more routines tasks within those occupations.  People will still be involved, but their jobs will be transformed by the advanced tools they now have to master.  Moreover, a growing technology-based economy will likely create all kinds of new occupations, which will more than offset declines in occupations displaced by automation, - as has been the case over the past couple of centuries.

One would also expect that technology advances will increase the productivity of the workers involved in these new or transformed occupations.  But, if we look at the past 10-15 for guidance, we’ll find that productivity growth has significantly declined over this timeframe, notwithstanding huge technology advances like smartphones, cloud computing, big data and artificial intelligence.  Economist have proposed competing explanations for the declining productivity growth, but have so far failed to reach consensus.  Understanding this productivity puzzle may well hold the key to future productivity improvements and long-term economic growth.

Human + Machine: The Impact of AI on Business Transformation

Last week I wrote about the impact of blockchain technologies on business transformation.  I now want to turn my attention to the more immediate impact of AI on business, based on Human + Machine: Reimagining Work in the Age of AI, a recently published book by Accenture executives Paul Daugherty and Jim Wilson.

The book reminds us that technology has long been transforming business.  Electricity and standardized processes ushered the first wave of business transformations about 100 years ago.  Manufacturing was decomposed into a set of processes and reassembled into the assembly line, leading to the age of mass production.

Several decades later, IT and process automation brought us the second wave of business transformation.  Mainframes and data processing software enabled the automation of back-office processes in the 1960s and 1970s.  Front-office automation followed in the 1980s with the advent of PCs and client-server systems.  A decade later, advances in networking and ERP software led to  business process reengineering and management. 

“Now, the third wave involves adaptive processes,” write Daugherty and Wilson.  “This third era, which builds on the previous two, will be more dramatic than the earlier revolutions enabled by assembly lines and digital computers, and will usher in entirely new, innovative ways of doing business… This adaptive capability is being driven by real-time data rather than by an a priori sequence of steps.   The paradox is that although these processes are not standardized or routine, they can repeatedly deliver better outcomes.”

The MIT Intelligence Quest

How does human intelligence work, in biological as well as in engineering terms? And how can we use such an understanding of human intelligence to build wiser and more useful machines?  On February 1, MIT launched the Intelligence Quest (MIT IQ), - an initiative aimed at addressing these big questions by advancing the science and engineering of both human and machine intelligence.  MIT IQ aims “to discover the foundations of human intelligence and drive the development of technological tools that can positively influence virtually every aspect of society.”

MIT has been deeply involved in artificial intelligence since the field’s inception in the 1950s.  MIT professors  John McCarthy and Marvin Minsky were among the founders and most prominent leaders of the new discipline.  AI was one of the most exciting areas in computer sciences in the 1960s and 1970s.  Many of the AI leaders in those days were convinced that a machine as intelligent as a human being would be developed within a couple of decades.  They were trying to do so by somehow programming the machines to exhibit intelligent behavior, even though to this day we have little idea what intelligence is about, let alone how to translate intelligence into a set of instructions to be executed by a machine.  Eventually, all these early AI approaches met with disappointment and were abandoned in the 1980s.  After years of unfulfilled promises, a so called AI winter of reduced interest and funding set in that nearly killed the field.

AI was reborn in the 1990s when it adopted a more applied, engineering-oriented paradigm.  The new AI paradigm enabled computers to acquire intelligent capabilities by ingesting and analyzing large amounts of data using powerful computers and sophisticated algorithms.  Instead of trying to explicitly program intelligence, this new approach was based on feeding lots and lots of data to the machine, and then letting the algorithms discover patterns and extract insights from all that data.

Such a data-driven, machine learning approach produced something akin to intelligence or knowledge.  Moreover, unlike the explicit programming-based approaches, the statistical-based ones scaled very nicely.  The more data you had, the more powerful the supercomputers, the more sophisticated the algorithms, the better the results.  Machine learning and related advances like deep learning, have played a major role in AI’s recent achievements. 

A Framework for Building AI Capabilities

After decades of promise and hype, artificial intelligence has finally reached a tipping point of market acceptance.  AI is seemingly everywhere.  Every day we can read about the latest AI advances and applications from startups and  large companies.  AI was the star of the 2018 Consumer Electronic Show earlier this year in Las Vegas.

But, despite its market acceptance, a recent McKinsey report found that AI adoption is still at an early, experimental stage, especially outside the tech sector.  Based on a survey of over 3,000 AI-aware C-level executives across 10 countries and 14 sectors, the report found that 20 percent of respondents had adopted AI at scale in a core part of their business, 40 percent were partial adopters or experimenters, while another 40 percent were still waiting to take their first steps.

The report adds that the gap between the early AI adopters and everyone else is growing.  While many companies have yet to be convinced of AI’s benefits, leading edge firms are charging ahead.  Companies need to start experimenting with AI and get on the learning curve, or they risk falling further behind. 

AI will likely become the most important technology of our era as it’s improved upon over time, but we’re still in the early stages of deployment.  It’s only been in the last few years that complementary innovations, especially machine learning have taken AI from the lab to early marketplace adopters.  And, history shows that even after technologies start crossing over into mainstream markets, it takes considerable time, - often decades, - for the new technologies and business models to be widely embraced by companies and industries across the economy.

Human Beings Need a Story

It wasn’t that long ago that we didn’t have much use for probabilities in our daily life, - outside of weather reports, election predictions, baseball and financial markets.  But that’s all changed with the growing datafication of the economy and society.  Probabilities have been playing an increasing role in our work and personal life given our newfound ability to quantify just about anything.  In all kinds of everyday situations, - from medical diagnoses to financial decisions, -  we now have to accept the fact that it’s impossible to predict what will actually happen.  Instead, we have to get used to living in a complex world of uncertainties and probabilities.  We have to learn how to deal with the very messy world of big data, and how to best apply our learning to make good decisions and predictions.

Physics went through such a transition about 100 years ago.  In the deterministic world of classical mechanics, there’s always a real truth.  The same objects, subject to the same forces, will always yield the same results.  Elegant mathematical models can be used to make perfect predictions within the accuracy of their human-scale measurements.  Early in the 19th century, French mathematician and scientist Pierre-Simon Laplace observed that if we knew the precise state of the universe as represented by the position and speed of every one of its particles, classical mechanics would enable us to calculate all past and future states of the universe.

But, this predictable world began to fall apart in the early 20th century.  19th century classical physics was replaced by a kind of 20th century magical mystery tour, ruled by the new principles of quantum mechanics and relativity.   Classical mechanics could not explain the counter-intuitive and seemingly absurd behavior of energy and matter at atomic and cosmological scales.

At these scales, the world is intrinsically unpredictable.  Instead of a deterministic world, we now have a world based on probabilities.  You cannot predict all the future states of an electron, for example, based on its present state.  You can map out its behavior, but only as probability distributions of all the possible states it could be at.  Moreover, the Heisenberg uncertainty principle tells you that it’s impossible to know the exact state of a particle.  You cannot simultaneously determine its exact position and velocity with any great degree of accuracy no matter how good your measurement tools are.

The Impact of Automation on Everyday Life

This past October, the Pew Research Center released Automation in Everyday Life, a report on what Americans think about advanced technologies like artificial intelligence and robotics, and the impact they expect them to have on their everyday lives.  The report is based on a national survey of 4,135 randomly selected American adults.  To help gauge their opinion on such complex topics, the survey questions were framed around four specific scenarios related to these advanced technologies: driverless cars, workplace automation, robot caregivers, and computer algorithms that evaluate and hire job applicants.

“Americans anticipate significant impacts from various automation technologies in the course of their lifetime,” noted the report.  “Although they expect certain positive outcomes from these developments, their attitudes more frequently reflect worry and concern over the implications of these technologies for society as a whole.”

Let me discuss some of the report’s key findings.

AI and the Productivity Paradox

AI is being successfully applied to tasks that not long ago were viewed as the exclusive domain of humans, matching or surpassing human level performance in more and more domains.  But, at the same time, productivity growth has significantly declined over the past decade, and income has continued to stagnate for the majority of Americans.  This puzzling contradiction is addressed in Artificial Intelligences and the Modern Productivity Paradox, a working paper recently published  by the National Bureau of Economic Research (NBER).

The paper’s authors, - MIT professor Erik Brynjolfsson, MIT PhD candidate Daniel Rock, and University of Chicago professor Chad Syverson, - note that “aggregate labor productivity growth in the U.S. averaged only 1.3% per year from 2005 to 2016, less than half of the 2.8% annual growth rate sustained from 1995 to 2004…  What’s more, real median income has stagnated since the late 1990s and non-economic measures of well-being, like life expectancy, have fallen for some groups.”

This productivity puzzle isn’t confined to AI.  “Few topics in economics today in most large economies generate as much debate as the productivity puzzle,” said McKinsey in a March, 2017 report.  “In  the United States, productivity growth has declined sharply since 2004 yet digital technology has been widely apparent during this period…  The answer to this puzzle holds the key to future prosperity because now more than ever our economy depends on productivity improvements for long-term economic growth.  Economists have proposed competing explanations for declining productivity growth and so far have failed to reach a consensus.” 

AI and The Future of Work

Last month I attended AI and the Future of Work, a conference hosted by MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and its Initiative on the Digital Economy (IDE).  The two-day agenda included over 20 keynotes and panels on AI-related topics, and around 60 speakers and panelists from academia and business.

New technologies have been replacing workers and transforming economies over the past two centuries.  But, over time, these same technologies led to the creation of whole new industries and new jobs.  While the technologies of the industrial economy helped to make up for our physical limitations, the technologies of the digital economy are now enhancing our cognitive capabilities.  They’re being increasingly applied to activities that not long ago were viewed as the exclusive domain of humans.  Will the AI revolution play out like past technology revolutions, - short term disruptions followed by long-term benefits, - or will this time be different?

Conference participants generally agreed that AI will have a major impact on jobs and the very nature of work.  But, for the most part, they viewed AI as mostly augmenting rather than replacing human capabilities, automating the more routine parts of a job and increasing the productivity and quality of workers, so they can focus on those aspect of the job that most require human attention.  Overall, a small percentage of jobs will be fully automated, while many more will be significantly transformed.

Conference participants also generally agreed that the more advanced AI-based transformations will not happen rapidly, but are likely decades away.  Much progress has been recently made in the ability to extract features from all the data we now have access to, as well as in machine learning algorithms that give computers the ability to learn by ingesting large amounts of data instead of being explicitly programmed.  While such statistical pattern recognition approaches can be applied to many tasks, they’re no substitute for model formation, the main approach used by humans, - from toddlers to physicists, - to understand how the world works.  We’re a long way from the development of AIs that truly learn and reason like people. 

Human-AI Decision Systems

I recently read The Human Strategy, a very interesting article on Human-AI decision systems by MIT Media Lab professor Sandy Pentland.  Pentland is the faculty director of MIT Connection Science, - with which I’m associated as a Fellow, - as well as director of the Human Dynamics Lab.  In 2014 he published Social Physics: How Social Networks Can Make Us Smarter.

“Perhaps the most critical function of any organization or society is its decision systems,” wrote Pentland.  “In modern societies, decision systems provide a leader the ability to make informed and timely decisions, supported by a complex enterprise of distributed information and communication systems that provide situational awareness.  Traditionally, decision systems have been confined to individual physical domains, such as logistics, physical plant, and human resources, and more recently virtual domains such as cyber, by both policy and technology, resulting in challenges with the integration of information across disparate domains.”

But, despite the increasingly complex decisions that organizations are called upon to make, decision-making remains human-intensive and anecdotal.  Few organizations have applied social network analysis to help them scale the size and expertise of the decision-making group.  Nor have they integrated the large amounts of data, analytical tools and powerful AI systems now at our disposal into their decision making systems.