Irving Wladawsky-Berger

Large Language Models: A Cognitive and Neuroscience Perspective

Over the past few decades, powerful AI systems have matched or surpassed human levels of performance in a number of tasks such as image and speech recognition, skin cancer classification, breast cancer detection, and highly complex games like Go. These AI breakthroughs have been based on increasingly powerful and inexpensive computing technologies, innovative deep learning (DL) algorithms, and huge amounts of data on almost any subject. More recently, the advent of large language models (LLMs) are taking AI to the next level. And, for many technologists like me, LLMs and their associated chatbots have introduced us to the fascinating world of human language and cognition.

I recently learned the difference between form, communicative intent, meaning, and understanding from “Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data,” a 2020 paper by linguistic professors Emiliy Bender and Alexander Koller. These linguistic concepts helped me understand the authors’ argument that “in contrast to some current hype, meaning cannot be learned from form alone. This means that even large language models such as BERT do not learn meaning; they learn some reflection of meaning into the linguistic form which is very useful in applications.”

A few weeks ago, I came across another interesting paper, “Dissociating Language and Thought in Large Language Models: a Cognitive Perspective,” published in January, 2023 by principal authors linguist Kyle Mahowald and cognitive neuroscientist Anna Ivanova and four additional co-authors. The paper nicely explains how the study of human language, cognition and neuroscience sheds light on the potential capabilities of LLMs and chatbots. Let me briefly discuss what I learned.

“Today’s large language models (LLMs) routinely generate coherent, grammatical and seemingly meaningful paragraphs of text,” said the paper’s abstract. “This achievement has led to speculation that these networks are — or will soon become — thinking machines, capable of performing tasks that require abstract knowledge and reasoning. “Here, we review the capabilities of LLMs by considering their performance on two different aspects of language use: formal linguistic competence, which includes knowledge of rules and patterns of a given language, and functional linguistic competence, a host of cognitive abilities required for language understanding and use in the real world.

The Economic Value of Open Source

The supply side of open source software (OSS), — that is, the reasons for joining and contributing to open source communities, — has been extensively studied and quantified. For example, the Linux Foundation (LF) has over 1,500 members, and is home to nearly 1,000 open source projects from over 15,000 contributing organizations. The LF estimates that over 50 million lines of code are added weekly to its sponsored projects, which comprise a significant percentage of the world’s mission-critical infrastructures, a figure that’s significantly higher if we include the contributions of other open source communities like the Apache Software Foundation.

However, we know much less about the demand side of OSS. Despite the wide adoption of OSS by firms and industries around the world, we don’t really understand, nor are we able to quantify its overall economic value to the many organizations that rely on open source. What benefits do these organizations derive from OSS? What costs  do they incur from using or contributing to OSS? And, if not able to use OSS in a project, what would have been the next-best alternatives, and how much would they have cost?

To get answers to these and other important OSS questions, the LF sponsored a study led by UC Berkeley professor Henry Chesbrough, — whose fourth book on open innovation, Open Innovation Results, was published in 2020. The key findings and conclusions from the study are discussed in a White Paper, “Measuring the Economic Value of Open Source.”

Driving Value with the Metaverse in the Enterprise

A killer app is an IT application whose value to individuals and companies is relatively simple to explain, its use is fairly intuitive, and the application turns out to be so useful that it can singlehandedly propel the success of a new product or service. Word processing and spreadsheets, for example, were major factors in the widespread adoption of personal computers in the 1980s.

I got interested in 3D immersive environments in the mid-2000s when IBM launched an exploratory initiative on the topic. As part of these efforts, we conducted a number of experiments on Second Life, a platform that allowed its users to develop a variety of virtual world applications. We built a number of workplace applications, including virtual meetings, and online events, and even built a virtual replica of Beijing’s Forbidden City in partnership with the Palace Museum that was opened to the public. But, despite our high expectations, we couldn’t find any compelling enterprise killer apps, and after about three years, the project was dissolved. In the ensuing years, 3D immersive technologies and applications continued to advance, but mostly focused on video-games and related industries.

The past few years have seen a surge of interest in the metaverse and extended reality. “After desktop computing, the consumer internet and the smartphone boom, the consumer-computing industry is past due its Next Big Thing,” wrote The Economist in  “A reality check for the metaverse is coming,” an article in its 2022 year-end issue. “[A]n internet which is still largely flat — based on two-dimensional text, images and video — is ripe for replacement with one that is three-dimensional and immersive.”

Are GPT-3 and ChatGPT Truly Intelligent?

Large language models (LLMs) — e.g., BERT, GPT-3, — and chatbots like ChatGPT, are quite impressive. In the coming years, they will likely lead to a number of transformative tools and applications. They may well herald a coming AI revolution. But, how much closer are they getting us to the kind of general intelligence that, for the foreseeable future, only humans have? How can we best sort out their true significance and value from the accompanying hype? Let me discuss how two different articles addressed these questions.

“The success of the large neural language models on many NLP [Natural Language Processing] tasks is exciting” wrote linguistic professors Emiliy Bender and Alexander Koller in their 2020 paper “Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data.” “This has led to claims, in both academic and popular publications, that such models understand or comprehend natural language or learn its meaning. From our perspective, these are overclaims caused by a misunderstanding of the relationship between linguistic form and meaning.”

In their paper, Bender and Koller explain why LLMs like GPT-3 are likely to become innovative language tools, — kind of like highly advanced spell checkers or word processors, — while dismissing claims that they have the ability to reason and understand the meaning of the language they’re generating. Their explanations are couched in linguistic concepts which they carefully define: form, communicative intent, meaning, and understanding.

Management Priorities for Highly Turbulent Times

“The COVID-19 pandemic rewrote the rules, and now a new and potent disruption seems to arrive every other day,” said a recent McKinsey article, What matters most? Six priorities for CEOs in turbulent times. “You know the list of issues; we won’t go through them here. Suffice to say that managing complex organizations is much harder today than it was just a few years ago. And the hardest task of all for CEOs is to decide what needs to be done now and what can wait.”

To find out what matters most almost three years into the pandemic, McKinsey spoke with hundreds of leading CEOs around the world. Six priorities featured prominently on their agendas:

• Resilience. Not surprisingly, “resilience is emerging as a vital muscle for companies operating in a world of endless volatility and disruption.”
• Strategic Courage. “The best leaders and companies are ambidextrous: prudent about managing the downside while courageously pursuing the upside.”
• Building New Businesses. “Going for greatness within a company’s industry is one thing. Venturing into an entirely different sector is another, more complicated story.”
• Technology. “Building a new business inevitably means new and better technology.” 
• Net-zero emissions. “[T]he goals of sustainability, economic competitiveness, affordability, and national security dovetail as never before.”
• Rebuild the employee experience. “In the wake (we hope) of the COVID-19 pandemic, CEOs need to find a new plane of engagement.”

Can an AI System Exhibit Commonsense Intelligence?

“One of the fundamental limitations of AI can be characterized as its lack of commonsense intelligence: the ability to reason intuitively about everyday situations and events, which requires rich background knowledge about how the physical and social world works,” wrote University of Washington professor Yejin Choi in “The Curious Case of Commonsense Intelligence,” an essay published in the Spring 2022 issue of Dædalus. “Trivial for humans, acquiring commonsense intelligence has been considered a nearly impossible goal in AI, added Choi.”

What is commonsense intelligence? In a 2020 presentation, Choi defined common sense as “the basic level of practical knowledge and reasoning concerning everyday situations and events that are commonly shared among most people.” Common sense is essential for humans “to live and interact with each other in a reasonable and safe way,” and for an AI system, common sense is essential “to understand humans needs and actions better.” She further explained the difference between intuitive commonsense reasoning and analytical rational reasoning by discussing the pioneering research of Daniel Kahneman, — Princeton Professor Emeritus and recipient of the 2002 Nobel Prize in Economics, — and his long time collaborator Amos Tversky, who died in 1996.

In his excellent 2011 bestseller Thinking, Fast and Slow, Kahneman explained that our mind is composed of two very different systems of thinking, System 1 and System 2.  System 1 is the intuitive, unconscious, fast, and effortless part of the mind. Thoughts come automatically and very quickly to System 1 without us doing anything to make them happen.

Is AI (Finally) Coming of Age?

Artificial intelligence first came to light in the mid-1950s as a promising new academic discipline. AI became one of the most exciting areas in computer sciences over the next two decades. But, after years of unfulfilled promises and hype, a couple of so called AI winters of reduced interest and funding set in that nearly killed the field. AI was successfully reborn in the 1990s with a new statistical paradigm based on analyzing large amounts of data with powerful computers and sophisticated algorithms. Now, six decades after the field was founded, AI seems to be finally coming of age.

“2021 saw the globalization and industrialization of AI intensify, while the ethical and regulatory issues of these technologies multiplied,” said the 2022 AI Index report on the progress of AI, which was released in March of 2022 by Stanford’s Institute for Human-Centered Artificial Intelligence (HAI). “2021 was the year that AI went from an emerging technology to a mature technology - we’re no longer dealing with a speculative part of scientific research, but instead something that has real-world impact, both positive and negative,” wrote Jack Clark, co-chair of the AI Index. Multiple factors led to his conclusion, in particular the advent of foundation models like OpenAI’s GPT-3 and Google’s BERT.

In the past decade, increasingly powerful AI systems have matched or surpassed human levels of performance in a number of specific tasks like image and speech recognition. These task-specific deep learning (DL) systems have generally relied on supervised learning, a training method where the data must be carefully labelled, - e.g., cat, not-cat, - thus requiring a big investment in time and money to produce a model that’s narrowly focused on one task and can’t be easily repurposed.

Foundation models have gotten around these DL limitations based on two recent advances, transfer learning and huge scale. Transfer learning takes the knowledge learned from training a specific task and applies it to different but related tasks, - such as using the training in recognizing cars in images and applying it to recognizing trucks and buses. As a result, foundation models can be adapted to many different tasks with relatively small amounts of fine-tuning.

Why Digital Wallets Are Critical to the Evolving Digital Landscape

A few months ago, the Linux Foundation (LF) announced its intent to launch the Open Wallet Foundation (OWF) in early 2023. The OWF aims to develop open source software, key building blocks, and best practices that anyone can use to build multi-platform, interoperable, secure, and privacy-protecting digital wallets that will support a wide variety of applications from identity credential to payments.

I’ve been really intrigued by digital wallets, a fairly complex and poorly understood subject. To learn more, I read “The Current and Future State of Digital Wallets,” a comprehensive report on digital wallets published in April of 2019 by technologist and entrepreneur Darrell O’Donnell. The report was sponsored by a consortium of Canadian companies that recognized that digital wallets were a major requirement to be able to properly manage our digital identities, — arguably the toughest challenge in the continuing evolution of the internet and the digital economy.

“Technologies are converging and provide immense potential for shifting how we can improve every aspect of our lives,” wrote O’Donnell in the report’s Executive Summary. “Underneath all of that technology lies the internet, which has a flaw that could have been fatal – until recently it has lacked a built-in digital identity. Hacks to get around this missing piece have resulted in huge security gaps and abusive digital relationships.”

Why Are Enterprises Struggling with Blockchain?

“Blockchain, the technology underpinning bitcoin and other cryptocurrencies, for years has been viewed by some companies as a way to drive industry-transforming projects, among them the tracking of assets through complex supply chains,” according to a recent article in the Wall Street Journal, “Blockchain Fails to Gain Traction in the Enterprise.”

“So far that hasn’t happened.”

Blockchain technologies first came to light in 2008 as the architecture underpinning bitcoin, the best known, most valuable, and most widely held cryptocurrency. Over the years, blockchain has evolved in two major directions. One continues to focus on blockchain as the underlying platform for bitcoin, as well as for the large number of cryptocurrencies, digital tokens, and other crypto assets that have since been created. The other focuses on blockchain as a distributed data platform for collaborative applications involving multiple institutions, such as supply chains, financial services and healthcare systems. 

Blockchain made its first appearance in Gartner’s yearly hype cycles in 2016. Despite its being at an early and immature stage, many of its proponents believed that enterprise blockchain platforms were just around the corner, and so it quickly rose to the peak of inflated expectations. But as reality set in, expectations started sliding down in 2021, and by 2022 blockchain platforms had reached the bottom of the trough of disillusionment.

As years go, 2022 was a very tough one for blockchains in general. In the crypto space, bitcoin and most other crypto assets have lost a very large chunk of their value, and numerous crypto companies have collapsed, most prominently FTX. In the enterprise, a number of ventures went belly up. we.trade, a blockchain-based trade finance, shut its operations in June after running out of cash. A few months later, the Australian Securities Exchange canceled its much-delayed blockchain-based clearing system. Most prominently, in late November TradeLens announced that it was discontinuing its blockchain-based global trade platform. Other enterprise blockchain projects, like Walmart’s food traceability application, still carry on but their uptake and progress have been slower than anticipated.

Will the Metaverse and AR/VR Headsets Be IT’s Next Big Thing?

“Is it really the next big thing,” asked The Economist in A reality check for the metaverse is coming, another recent article in “The World Ahead 2023”, its year-end issue which I wrote about last week. “After desktop computing, the consumer internet and the smartphone boom, the consumer-computing industry is past due its Next Big Thing,” said the article. “The coming year will see big tech firms doubling down on two related, much-hyped possibilities. One is virtual- (vr) and augmented-reality (ar) headsets; the idea that, having shrunk computers into our pockets, the next step is to strap them to our faces. The other is the metaverse, which holds that an internet which is still largely flat - based on two-dimensional text, images and video - is ripe for replacement with one that is three-dimensional and immersive, experienced as a sort of globe-spanning video game.”

There is little question that the metaverse and AR/VR headsets are important trends to watch out for in the coming years. As The Economist wrote in a November, 2021 article, “as computers have become more capable, the experiences which they generate have become richer. The internet began its life displaying nothing more exciting than white text on a black background.”

The last major advance in user interfaces took place in the 1980s when text interfaces gave way to graphical user Interfaces (GUIS). GUIs were first developed at Xerox PARC in the late 1970s and later popularized by the Apple Macintosh in the ’80s. In the 1990s, GUIs were embraced by just about every PC and user device, and GUI-based Web browsers played a major role in the explosive growth of the internet.

The History and Current State of Human Progress

“For hundreds of thousands of years, human history played out without any rapid, marked advance in material living standards,” wrote NY Times journalist Ezra Klein in the introduction to his recent podcast, We Know Shockingly Little About What Makes Humanity Prosper. “And then, suddenly, in just the past few hundred years, everything changed: Humanity achieved a truly mind-boggling amount of progress in the evolutionary blink of an eye. In the early 21st century, we are all living in the world that progress bequeathed. And yet we understand shockingly little about what drives that progress in the first place. That’s important because, at least according to some metrics, progress seems to be slowing down.”

Klein’s podcast guest was Irish entrepreneur Patrick Collison, co-founder and CEO of the financial services company Stripe, one of the world’s most profitable and highest valued startups. Collison is very interested in the long history of human progress, and has been a leading advocate for the formation of a new discipline of Progress Studies.

In the podcast, Klein and Collison discussed a number of topics on the history and current state of human progress. But I’d like to focus my summary of the podcast on three key questions:What is progress?; why is progress slowing down?; and do we need a new discipline of Progress Studies?

Getting the Most Out of Cloud Investments

“Over the past decade, two major trends have emerged concurrently,” said Five Keys To Innovating Faster With Cloud, a recent report by Forrester Research. “First, there has been a growing need for rapid innovation and adaptivity as businesses respond to customer-driven trends. At the same time, cloud platforms have advanced rapidly in their capabilities and ease of deployment. Leveraging cloud has become the fastest way to modernize your tech stack and deliver the innovation and adaptivity needed in a customer-obsessed strategy.”

Cloud emerged in the late 2000s as a new model of computing, - the third in the history of the IT industry after the mainframe-based centralized computing and the PC-based client-server models. The internet is the defining technology of the cloud model.

Cloud has gone through three major stages over the past fifteen years. First came infrastructure-as-a-service, offering near unlimited scalability at very attractive prices. Then came software-as-a-service, offering a faster and less costly way of prototyping and deploying innovative applications with advanced tools like containers, Kubernetes, and microservices. Cloud computing has now entered the third stage, - a major engine of business transformation that’s helping companies adapt to the digitalization of the economy, - a digitalization which has significantly accelerated since the advent of Covid-19 in March of 2020.

The Current State of Service Science

On October 17, the National Academy of Engineering (NAE) conducted an online forum on Service Systems Engineering in the Era of Human-Centered AI. “With AI advances poised to drive service system productivity and quality - similar to the way previous generations of technology revolutionized agriculture and manufacturing productivity and quality - it is time to take stock for industry-academic-and-government stakeholders on this important topic,” wrote the NAE in its website.

The agenda included an opening keynote by retired IBM executive Nick Donofrio. It was followed by four panels on various aspects of service systems, and concluded with an open discussion of the way forward. I was a member of the panel on Evolving Engineering Education. In my prepared remarks, I reflected on the current state of service science and related sociotechnical systems. Let me share my remarks, slightly edited for clarity.

Service Science, Management and Engineering (SSME) is an initiative launched in IBM’s Almaden Research Lab in the early 2000s in partnership with a number of universities as an interdisciplinary field of study aimed at applying science, technology and innovation to the service sector of the economy.  The service sector is the largest in most economies around the world. At the time, services already accounted for over 70% of GDP and jobs in advanced economies, as well as an increasing portion of the revenues of many companies, including close to 60% of IBM’s revenues.

The Disruptive Economics of AI

In 2017, I attended a seminar by University of Toronto professor Avi Goldfarb on the economic value of AI. Goldfarb explained that the best way to assess the impact of a potentially transformative technology is to look at how the technology reduces the cost of a widely used function. Computers, for example, have dramatically reduced the cost of digital operations like arithmetic by several orders of magnitude. As a result, we’ve learned to define all kinds of tasks in terms of digital operations, e.g., financial transactions, inventory management, word processing, photography. Similarly, the internet has reduced the cost of communications and the Web has reduced the cost of access to information, which has led to a huge increase in applications based on communications and information, like music and video streaming, and digital media.

Viewed through this lens, AI is essentially a prediction technology, and its economic impact is to reduce the cost and expand the number and variety of applications that rely on predictions. A key finding of Stanford’s 2022 AI Index report was that AI is becoming much more affordable and higher performing, leading to the widespread commercial adoption of AI-based applications. “Since 2018, the cost to train an image classification system has decreased by 63.6%, while training times have improved by 94.4%.,” said the report.

Goldfarb and his U of T colleagues Ajay Agrawal, and Joshua Gans published their research in the 2018 book Prediction Machines: The Simple Economics of Artificial Intelligence. And this week, they just published Power and Prediction: The Disruptive Economics of Artificial Intelligence.

“When we published Prediction Machines in 2018, we thought we had said all we needed to on the economics of AI,” wrote the authors in the Preface of their new book. “We were wrong. We laid out a framework for the economics of AI in that book, which remains useful today. However, the Prediction Machines framework only told part of the story, the point solutions part. In the years since, we discovered that another key part of the AI story had yet to be told, the systems part. We tell that story here.”

The Changing Structure of Global Supply Chains

“The golden age of globalisation, in 1990-2010, was something to behold,” wrote The Economist in a January, 2019 article.  “Commerce soared as the cost of shifting goods in ships and planes fell, phone calls got cheaper, tariffs were cut and the financial system liberalised.” The very nature of the global firm was transformed during these two decades, with The Globally Integrated Enterprise destined to become the corporate model of the future, said IBM CEO Sam Palmisano in a 2006 Foreign Affairs article.

But, then global trade started slow down. “After the go-go 1990s and 2000s the pace of economic integration stalled in the 2010s, as firms grappled with the aftershocks of a financial crisis, a populist revolt against open borders and President Donald Trump’s trade war,” wrote The Economist in its June 18, 2022 lead article. “The flow of goods and capital stagnated. Many bosses postponed big decisions on investing abroad: just-in-time gave way to wait-and-see. No one knew if globalisation faced a blip or extinction.”

“Now the waiting is over, as the pandemic and war in Ukraine have triggered a once-in-a-generation reimagining of global capitalism in boardrooms and governments,” added the article. “Everywhere you look, supply chains are being transformed, from the $9trn in inventories, stockpiled as insurance against shortages and inflation, to the fight for workers as global firms shift from China into Vietnam. This new kind of globalisation is about security, not efficiency: it prioritises doing business with people you can rely on, in countries your government is friendly with.”

Most everyone agrees that global supply chains will be restructured over the next decade, accelerating the changes that were already in motion. “Decision-makers are increasingly concerned that supply chains should be robust, not just efficient,” said the briefing article in the same Economist issue. “As a result they are choosing to depend less on jurisdictions where they are exposed to risk. And countries are experimenting with industrial policies aimed at self-reliance or international pre-eminence in at least some ‘strategic’ technologies and businesses.” Let me discuss each of these potential changes.

The AI Maturity Framework

I recently attended a seminar, The Art of AI Maturity, by Accenture executives Philippe Roussiere and Praveen Tanguturi as part of MIT’s Initiative on the Digital Economy (IDE) lunch seminar series. The seminar was based on their recently published article The Art of AI Maturity: Advancing from Practice to Performance.

“Today, so much of what we take for granted in our daily lives stems from machine learning,” wrote the authors in the article’s executive summary. “Every time you use a wayfinding app to get from point A to point B, use dictation to convert speech to text, or unlock your phone using face ID ... you're relying on AI. And companies across industries are also relying on - and investing in - AI to drive logistics, improve customer service, increase efficiency, empower employees and so much more.”

To determine the true state of AI maturity in the marketplace, Accenture conducted a survey in August and September of 2021 of over 1,600 C-suite executives at nearly 2,000 of the world’s largest companies across 16 industries with headquarters in 15 countries. In addition, they interviewed 25 CEOs, Chief Data Officers and Chief Analytics Officers, as well as a number of Accenture AI experts, and developed over 40 case studies on AI transformation.

The Latest Trends in Work from Home (WFH)

For years, companies and governments found all kinds of reasons for not embracing work from home, virtual meetings, telemedicine, online learning, and other online applications. But, the pandemic forced us to accelerate the digital transformation of the economy and society to help us cope with the crisis. And, not only have these digital applications worked remarkably well, but they offer a number of important benefits, like not waiting for a straightforward doctor diagnosis in a room full of sick people, and not having to travel for hours to participate in a 60 minute meeting.

For example, about a year ago I participated in an online panel. To start with, the moderator asked each panelist to introduce ourselves and to briefly say something positive about our lives over the past year despite the obvious challenges of coping with Covid. I truly struggled to find something positive to say about the highly frustrating pandemic-induced limitations of the previous year.

But finally, I did find something positive to say. I’ve been able to attend a number of meetings, including weekly lunch seminars, that pre-pandemic required me to drive for a few hours or catch a plane. In response to the pandemic, these meeting first changed from physical to online, and more recently to hybrid. In other words, my ability to attend such meetings from home was my one pandemic-induced benefit, a major one the more I thought about it.

Work from home (WFH) has been around for decades, modestly growing in the 1990s with the rise of the internet. The share of WFH three or more days per week was under 1% in 1980, 2.4% in 2010, and 4.0% in 2018. Then came Covid-19, forcing tens of millions around the world to work from home and triggering a mass workplace experiment that broke through the technological and cultural barriers that had prevented its adoption in the past.

How Should Companies Prepare for the Coming Cybersecurity Regulations

“Cybersecurity has reached a tipping point,” wrote MIT professor Stuart Madnick in a recent Harvard Business Review article, New Cybersecurity Regulations Are Coming. Here’s How to Prepare. “After decades of private-sector organizations more or less being left to deal with cyber incidents on their own, the scale and impact of cyberattacks means that the fallout from these incidents can ripple across societies and borders.”

Given the growing threat of cyberattacks, there’s an urgent need to improve the security of IT systems. However, we still don’t know a lot about cyberattacks, including how many attacks have taken place and who’s been attacked. Until recently, cybersecurity regulation were mostly focused on data privacy, and the only attacks that had to be reported were those involving personal information, such as the theft of names and credit card numbers.

For example, when Colonial Pipeline suffered a serious ransomware attack in May of 2021 that shut down nearly 50% of fuel deliveries to the US East Coast, neither the company nor the pipeline operators were required to report the attack because personal information wasn’t stolen. “As a result, it’s almost impossible to know how many cyberattacks there really are, and what form they take,” said Madnick. “Some have suggested that only 25% of cybersecurity incidents are reported, others say only about 18%, others say that 10% or less are reported.” We need detailed information on who is being attacked, how are they getting attacked, what are the attackers after, and what have they’ve actually stolen.

The Recent Impact of AI on Jobs and Economies

“Recent developments in Artificial Intelligence (AI) have stoked new fears about large-scale job loss, stemming from its ability to automate a rapidly expanding set of tasks (including non-routine cognitive tasks), and its potential to affect every sector of the economy,” said The impact of Artificial Intelligence on the labour market: What do we know so far?, a recently published report by the Organization for Economic Co-operation and Development (OECD). Founded in 1961 to stimulate economic progress and world trade, the OECD now has 38 member countries around the world.

“Furthermore, there are concerns about employee well-being and the broader work environment, linked to the idea that AI may soon become pervasive in the workplace and threaten and undermine humans’ place in it,” noted the report. “However, AI also has the potential to complement and augment human capabilities, leading to higher productivity, greater demand for human labour and improved job quality.”

A number of recent studies have focused on the impact of AI on the future of work. For example, in the spring of 2018, MIT launched a task force on the Work of the Future to understand whether and how our current period of technological disruption differs from prior industrial epochs. The task force final report, The Work of the Future: Building Better Jobs in an Age of Intelligent Machines, released in November of 2020, concluded that:

 “No compelling historical or contemporary evidence suggests that technological advances are driving us toward a jobless future. On the contrary, we anticipate that in the next two decades, industrialized countries will have more job openings than workers to fill them, and that robotics and automation will play an increasingly crucial role in closing these gaps. Nevertheless, the impact of robotics and automation on workers will not be benign. These technologies, in concert with economic incentives, policy choices, and institutional forces, will alter the set of jobs available and the skills they demand.”

Human Capital: The Value of Work Experience and Continuous Learning

“The most important resource in any economy or organization is its human capital - that is, the collective knowledge, attributes, skills, experience, and health of the workforce,” said Human capital at work: The value of experience, a recent research report by the McKinsey Global Institute. Human capital development starts in early childhood, continues through formal educations, and grows as a person acquires and deploys new skills through their working life. The value of human capital can be approximated by an individual’s lifetime earnings, and represents roughly two-thirds of their total wealth.

“Human capital is much more than a macroeconomic abstraction,” notes the report. “Each person has a unique, living, breathing set of capabilities. They belong to the individual, who decides where to put them to work. The degree of choice is not limitless, of course. People are the products of geography, family, and education; their starting points matter. Having career options also depends on an individual’s abilities and attributes, their networks, their family obligations, the health of the broader labor market, and societal factors. While we recognize these constraints, career moves are nevertheless an important mechanism for expanding skills and increasing earnings.”

McKinsey’s study is focused on the contributions of work experience to a person’s human capital. The research analyzed a dataset containing the de-identified histories of all the job moves made by about a million workers in each of four countries: the US, UK, Germany, and India. The study traced the work history of each individual and their lifetime earnings starting with the first job after the last degree obtained and after all subsequent role moves. For each role move, the study estimated its skill distance, defined as the the opportunity to acquire new and additional skills in the new job. Finally, the study calculated both the share of lifetime earnings attributed to the entry-level skills acquired through formal education, and the share of lifetime earnings attributed to work experience.

The Importance of Open Source Digital Wallets to the Future of the Internet

On September 13, the Linux Foundation (LF) announced its intent to form the Open Wallet Foundation (OWF) at its Open Source Summit Europe in Dublin. The OWF aims to develop open source software, key building blocks, and best practices that anyone can use as a starting point to build multi-platform, highly secure, privacy-protecting digital wallets. In addition, the OWF will advocate for the wide adoption of open source wallets that can be used to support a wide variety of applications, from payments to digital identities.

The LF was originally founded in 2000 as the Open Source Development Labs to help set the standards for the Linux operating system and support its continued development and commercial adoption. It grew over the years and assumed its current name, Linux Foundation, in 2007. Over the past decade, the LF has undergone a major expansion beyond its original Linux mission. It now has over 1,260 company members and supports hundreds of open source projects. Some of the projects are focused on technology horizontals, - e.g., AI, blockchain, security, cloud, - and others on industry verticals, including energy, automative, government, and health care.

Let me attempt to explain the importance of this announcement by addressing three key questions: what is a digital wallet?; what are digital identities?; and why do we need standard-based, open source digital wallets?

The 2022 State of AI in the Enterprise

“Rapidly transforming, but not fully transformed - this is our overarching conclusion on the market, based on the fourth edition of our State of AI in the Enterprise global survey,” said Becoming an AI-fueled organization, the fourth survey conducted by Deloitte since 2017 to assess the adoption of AI across enterprises. “Very few organizations can claim to be completely AI-fueled, but a significant and growing percentage are starting to display the behaviors that can get them there.”

AI is increasingly viewed by workers as a trusted assistant. “Within just the last 18 months, AI capabilities have advanced considerably, maturing from what was often experienced as a bothersome critic - telling workers what to do or pointing out their mistakes - to more frequently serving as a copilot, independently executing on insights and trends surfaced through the power and speed of cloud- based data hosting and computation.”

These conclusions are similar to those of Stanford’s 2022 AI Index report, which found that AI was becoming more affordable and higher performing, with lower training costs and faster training times across a number of AI tasks including recommendation engines, image classification, object detection, and language processing. This has led to the widespread commercial adoption and increased real-world impact of AI systems.

The Carnegie Cloud Governance Project

Two years ago, the Carnegie Endowment launched the Cloud Governance Project, a multi-year study on the governance challenges associated with cloud computing. “This project recognizes that the cloud offers huge benefits for individuals, organizations, and national economies through greater IT convenience, flexibility, and cost savings,” said the project’s website. “However, the risks of a major disruption affecting cloud services will invite regulation by governments at the local, national, and international levels. Moreover, as the world grows increasingly dependent on the cloud, other aspects of the technology – related to consumer protection, sustainability, inclusiveness, and human rights – will also attract scrutiny and regulation to protect or advance public interests.”

I’ve been closely following cloud computing since it first emerged in the late 2000s. In 2008 I gave a talk at a conference on The Promise and Reality of Cloud Computing. Most everyone at the conference agreed that something profound and exciting was going on, but there was no real consensus on what cloud computing was. A major reasons for both the excitement and lack of consensus about cloud is that we were basically seeing the emergence of a new model of computing in the IT world, only the third such model in the history of computing. The mainframe-based centralized computing model first appeared in the 1960s. Then came the PC-based client-server model in the 1980s. The more recent internet-based cloud computing model emerged in the late 2000s.

Over the past fifteen years, cloud has gone through three major stages. First came infrastructure-as-a-service, offering near unlimited scalability at very attractive prices. Then came software-as-a-service, offering a faster and less costly way of prototyping and deploying innovative applications leveraging advanced tools like containers, Kubernetes, and microservices. Cloud computing has now become a major engine of business transformation, helping companies adapt to the accelerated digitalization of the economy, especially since the advent of Covid-19 in March of 2020.

A Framework for Understanding the Power of Platforms

This past July, MIT’s Initiative on the Digital Economy held its 10th annual Platform Strategy Summit. The hybrid Summit included a number of panels and keynotes throughout the day, a few of which I attended online. In particular, I’d like to discuss the very interesting presentation by Boston University professor Marshall Van Allstyne as part of a panel on emerging trends in which he raised and answered five fundamental questions about platforms:

• Structure: Why do platforms have such high market caps but so few assets or employees?
• Strategy: Why does traditional product strategy fail in platform markets?
• Regulation: Why does traditional antitrust fail in platform markets?
• Fake News: Why is misinformation such a hard problem?
• Can Decentralized Autonomous Organizations dethrone platforms?

In his presentation, Van Alstyne, aimed to articulate a single framework that will help answer all five questions, that is, a consistent way of thinking about platforms based on the research that he and his colleagues have been conducting over the past decade. Let me summarize Van Alstyne’s presentation by first describing his unifying framework, followed by how the framework applies to each of the five key questions.

Web3: The Promise of a Blockchain-Based Internet of Value

On June 30, MIT hosted the Imagination in Action Web3 Summit, an event jointly organized by MIT Connection Science, Forbes, and Link Ventures. The Summit brought together almost 600 developers, entrepreneurs, investors, and academics to discuss the present state and potential evolution of Web3. The agenda featured a number of panels and talks. Speakers included Alex (Sandy) Pentland, - MIT professor and Connection Science Director, Michael Federle, - CEO of Forbes, John Werner, - Managing Director at Link Ventures, and Esther Dyson, - investor, journalist, and philanthropist.

The MIT Summit is the first in a series of Imagination in Action conferences. The next one will take place on October 6 in San Francisco, followed by one on January 17 at Davos, and back to MIT in June of 2023. The summits aim “to take us from reflections on the last major social and economic transformation precipitated by the arrival of the Internet era, to the economics to build and the governance to sustain Web3,” as well as to envision the possibilities of a Web3 future and to try to discern hype from reality in these early days of a potential new internet era.

I wasn’t able to attend the Summit in person, but I watched the recordings of a number of the sessions, where I heard a variety of opinions on what Web3 is about. This is not surprising. In their early years, major new technologies are generally accompanied by a mixture of excitement, speculation and confusion, as people sort out what the technology might be about and how it’s likely to evolve. Something important is going on out there, but it takes time and marketplace experience to sort things out.

I’ve seen this play out a few times throughout my long career in the IT industry, especially in the early 1990s as the internet was transitioning from a DARPA-sponsored project in the research community to an initiative that was starting to take the broader marketplace by storm. A lot was happening around the internet, but it wasn’t at all clear where things were heading, and in particular what the implications would be to the world of business. Let me share some of the lessons learned in the 1990s while formulating IBM’s internet strategy in our newly formed Internet Division.

The Economic Potential of the Internet of Things

In June of 2015, the McKinsey Global Institute published a report on the Internet of Things (IoT). The report analyzed the long term economic potential of IoT by examining over one hundred use cases and the IoT-based solutions in which they were generally deployed such as factories, homes, offices, and cities. It then estimated that the economic potential of IoT by 2025 would be approximately $3.9 trillion to $11.1 trillion per year.

The estimate was based on the likely evolution of technology and the rate of adoption of IoT solutions between 2015 and 2025, as well as economic and demographic trends. The estimated range was so wide because there were so many unknowns at that early stage in IoT’s development, including the costs of technology, the rate of development and deployment of these highly complex solutions, their level of acceptance by consumers and workers, and the policies and regulations enacted by governments.

Six years later McKinsey published a new report, The Internet of Things: Catching up to an accelerating opportunity, which updated the 2015 analysis. The recent report evaluated how much of its projected value in the earlier report had been captured and estimated the IoT’s economic potential through 2030, as well as the key headwinds slowing its progress and the key tailwinds propelling it forward.

“While the potential economic value from the IoT is large and growing, capturing this value has proved challenging,” said the recent report. “Our latest research shows that the total value captured in 2020 ($1.6 trillion) is at the lower end of the range of the scenarios we mapped out in 2015. We have updated our estimates for 2025 and beyond by adjusting for current conditions, and we have developed scenarios that account for the range of various uncertainties. In aggregate, both the low- and high-end scenarios are lower than the original 2015 estimates: around $2.8 trillion to $6.3 trillion in potential IoT economic value in 2025, compared with around $3.9 trillion to $11.1 trillion from the 2015 work.”

The Open Source AI and Data Ecosystem

“The availability of enterprise-grade open source software (OSS) is changing how organizations develop, maintain, and deliver products,” wrote Ibrahim Haddad in a recent report, Artificial Intelligence and Data in Open Source. Haddad is VP of Strategic Programs at the Linux Foundation (LF) and Executive Director of the LF AI & Data initiative. “Using and adopting OSS can offer many benefits, including reduced development costs, faster product development, higher code quality standards, and more. The open source methodology offers key and unique benefits to the domains of AI and data, specifically in areas of fairness, robustness, explainability, lineage, availability of data, and governance.”

Earlier this year, Stanford released the 2022 AI Index report, its fifth annual study on the impact and progress of AI. The Stanford report noted that “2021 was the year that AI went from an emerging technology to a mature technology - we’re no longer dealing with a speculative part of scientific research, but instead something that has real-world impact, both positive and negative.” 

A few weeks ago I wrote about the impressive scope of the Linux Foundation. The LF supports a large and growing number of open source projects in a wide variety of areas. AI is no different from other technology domains, so it’s not surprising that open source now plays a major role as AI is being increasingly integrated into the economy.

Disruptive Forces Necessitate Bold Decisions

In January of 2021 I attended Predictions 21, an online event organized by Forrester Research.  “Faced with the pandemic, firms did things that once seemed impossible - sometimes overnight,” said Forrester last year, adding that “2021 will be the year that every company - not just the 15% of firms that were already digitally savvy - doubles down on technology-fueled experiences, operations, products, and ecosystems.”

Earlier this year I attended Predictions 22, and was particularly curious to see how things had changed in the intervening year. “Disruptive Forces Necessitate Bold Decisions,” was the overriding message in this year’s event guide.  “The old ways of working no longer work. The future is up for grabs. Leading firms will use the crucibles of 2020 and 2021 to forge a path to an agile, creative, and resilient tomorrow.”

Let me summarize some of Forrester’s key predictions in three areas: technology, customer experience, and industry trends.

Web3 - Safeguarding Our Identity and Personal Data in the Digital World

Transformative technologies are generally accompanied by a mixture of excitement and confusion in their early years. Something important is going on out there, although there’s no consensus on what it is yet. A major reason for the lack of consensus is that there’s no single dimension around which to define an emerging technology or business model. It’s like the fable of the blind men and the elephant. Each one touches a different part of the elephant. They then compare notes on what they felt, and learn that they are in complete disagreement.

This was the case with the advent of the commercial internet in the early 1990s. A lot was starting to happen around the internet, but we weren’t sure where things were heading. It was pretty clear that a communications revolution was underway: after all, the internet was fundamentally a network of networks, and email was one of its earliest and most popular applications. It was also an information revolution: anyone with a browser, a PC, and an internet connection could now access all kinds of content in the new World Wide Web. And, above all, it promised to be an economic revolution: the internet ushered a historical transition to a new kind of digital economy, including many innovative e-business applications.

Over the past few decades, the term internet has come to encompass a number of related technologies including broadband networks, mobile devices, social media, cloud computing, e-commerce platforms, big data, AI, and more. More recently, we’ve seen the emergence of a new set of technologies and business models that are once more generating excitement, confusion, and multiple opinions on what they’re all about: Web3.

Last week I wrote about web3, referencing two recent books: the just-published Digital Asset Revolution by Alex Tapscott, and the soon-to-be-published Think Blockchain by Jerry Cuomo.

I wrote that web3 aims to usher a more open, entrepreneurial internet and digital economy by replacing today’s corporate mega-platforms with blockchain-based decentralized networks. Web3 would thus give creators, developers and users a way to monetize their contributions, involve them in the governance and decision-making of the platforms supporting their work, and give individuals more privacy and control over their data.

I now want to discuss another perspective on web3 based on The Emerging New Economy: Causes and Consequences of Web 3.0, a recent Stanford seminar by Alex (Sandy) Pentland, MIT professor and faculty director of the MIT Connection Science Research initiative, - with which I’ve long been affiliated as a Connection Science Fellow.

The Potential Impact of Web3 on the Internet and the Digital Economy

A few weeks ago, the Blockchain Research Institute (BRI) announced that its annual global conference would be renamed Web3 Blockchain World, and that the Enterprise Blockchain Awards would now be called the Web3 & Blockchain Transformation Awards.

I’ve been affiliated with BRI for a number of years as a research contributor and as chair of its blockchain awards committee. I was somewhat concerned with the name change, because, as is generally true of new technologies in their early stages, there are multiple views of what web3 is all about. Critics view web3 as little more than hype, a rebranding effort to shed some of the cultural and political baggage of crypto, while supporters believe that web3 represents the future of the internet, upending traditional gatekeepers and ushering a middleman-free digital economy.

Over the following weeks I learned from BRI founder Don Tapscott his reasoning behind the decision that BRI should embrace web3 in its research and events. I read Think Blockchain, a book by IBM Fellow and VP of Blockchain Technologies Jerry Cuomo that will be published later this month. I read the section on Web3 and the New Internet of Value in Digital Asset Revolution, a book by BRI co-founder Alex Tapscott that was just published earlier this week. And I looked at a number of articles from both supporters and critics of web3.

In the end, I became convinced that web3 is now where the commercial internet was in the early 1990s, cloud computing in the late 2000s, and blockchain in the mid 2010s: a very promising set of technologies and applications in their early years. While, at this stage it’s not yet clear how it will evolve, it’s time to start paying serious attention to web3 and try to influence its directions. Let me briefly share my views of web3.

Foundation Models: AI’s Exciting New Frontier

Over the past decade, powerful AI systems have matched or surpassed human levels of performance in a number of specific tasks such as image and speech recognition, skin cancer classification and breast cancer detection, and highly complex games like Go. These AI breakthroughs have been based on deep learning (DL), a technique loosely based on the network structure of neurons in the human brain that now dominates the field. DL systems acquire knowledge by being trained with millions to billions of texts, images and other data instead of being explicitly programmed.

These task-specific, DL systems have generally relied on supervised learning, a training method where the data must be carefully labelled, - e.g., cat, not-cat, - thus requiring a big investment of time and money to produce a model that’s narrowly focused on a specific task and can’t be easily repurposed. The rising costs for training ever-larger, narrowly focused DL systems have prompted concerns that the technique was running out of steam.

Foundation models promise to get around these DL concerns by bringing to the world of AI the reusability and extensibility that have been so successful in IT software systems, from operating system like iOS and Android to the growing number and varieties of internet-based platforms.

The Impressive Scope of the Linux Foundation in the 21st Century Digital Economy

Last week, the Linux Foundation held its North America Open Source Summit in Austin. The week-long summit included a large number of breakout sessions as well as several keynotes. Open Source Summit Europe will take place in Dublin in September and Open Source Summit Japan in Yokohama in December.

I’ve been closely involved with open, collaborative innovation and open source communities since the 1990s. In particular, I was asked to lead a new Linux initiative that IBM launched in January of 2000 to embrace Linux across all the company’s products and services.

At the time, Linux had already been embraced by the research, internet, and supercomputing communities, but many in the commercial marketplace were perplexed by IBM’s decision. Over the next few years, we spent quite a bit of effort explaining to the business community why we were supporting Linux, which included a number of Linux commercials like this one with Muhammad Ali that ran in the 2006 Super Bowl. IBM also had to fight off a multi-billion dollar lawsuit for alleged intellectual property violations in its contributions to the development of Linux. Nevertheless, by the late 2000s, Linux had crossed the chasm to mainstream adoption, having been embraced by a large number of companies around the world.

In 2000, IBM, along with HP, Intel, and several other companies formed a consortium to support the continued development of Linux, and founded a new non-profit organization, the Open Source Development Labs (OSDL). In 2007, OSDL merged with the Free Standards Group (FSG) and became the Linux Foundation (LF). In 2011, the LF marked the 20th anniversary of Linux at its annual LinuxCon North America conference. I had the privilege of giving one of the keynotes at the conference in Vancouver, where I recounted my personal involvement with Linux and open source.

How to Build Organizational Resilience to Cyberattacks

I became quite interested in cybersecurity around a year ago, given the growing threats of cyberattacks by criminal groups and adversarial governments. I then joined CAMS, MIT’s interdisciplinary cybersecurity consortium and started attending its online weekly seminars. A few weeks ago I attended a CAMS seminar on cyber resilience by Manuel Hepfer, a research affiliate at Oxford University and a research analyst at ISTARI, a cyber risk management company.

What is cyber resilience? While cybersecurity is the practice of protecting critical systems and sensitive information from digital attacks, cyber resilience is the ability to prepare for, withstand, and recover rapidly from any major disruption, whether an intentional cyberattack or a natural disaster.

The seminar was based on a study by Hepfer and colleagues of the 2017 NotPetya ransomware attacks, a series of powerful cyberattacks which caused over $10 billion in global economic damage across a number of industries. Their study was primarily focused on three global companies that were the subject of NotPetya attacks: a logistics company with over 60,000 employees, a manufacturing one with over 20,000 employees, and a third in professional services with over 5,000 employees.

The study conducted around 50 interviews with senior executives across the three companies, including CEOs, CFOs, CIOs, and CISOs. In addition, it reviewed internal documents, presentations, press releases, newspaper coverage, and audio and video files related to events before, during, and after the attack. Their study was published in two articles in the MIT Sloan Management Review, the first in 2020, the second in 2022.

Can We Trust Our Impressive AI Language Models?

One of the key findings of the 2022 AI Index Report was that large language models (LLMs) are setting records on technical benchmarks thanks to advances in deep neural networks and computational power that allows them to be trained using huge amounts of data. LLMs are now surpassing human baselines in a number of complex language tasks, including English language understanding, text summarization, natural language inference, and machine translation.

A.I. Is Mastering Language. Should We Trust What It Says?, a recent NY Times Magazine article by science writer Steven Johnson, took a close look at one such LLM, the Generative Pre-Trained Transformer 3, generally referred to as GPT-3. GPT-3 was created by the AI research company OpenAI. It’s been trained with over 700 gigabytes of data from across the web, along with a large collection of text from digitized books. “Since GPT-3’s release, the internet has been awash with examples of the software’s eerie facility with language - along with its blind spots and foibles and other more sinister tendencies,” said Johnson.

“So far, the experiments with large language models have been mostly that: experiments probing the model for signs of true intelligence, exploring its creative uses, exposing its biases. But the ultimate commercial potential is enormous. If the existing trajectory continues, software like GPT-3 could revolutionize how we search for information in the next few years.” Instead of typing a few keywords into Google and getting back a long list of links that might have the answer, you’d ask GPT-3 what you’re looking for in English and get back a well-written, accurate answer. “Customer service could be utterly transformed: Any company with a product that currently requires a human tech-support team might be able to train an L.L.M. to replace them.”

A Human-Compatible Safe, Controllable AI

“Even in its present state, the technology of artificial intelligence raises many concerns as it transitions from research into widespread use,” wrote UC Berkeley professor Stuart Russell in a recent essay, If We Succeed. “These concerns include potential misuses such as cybercrime, surveillance, disinformation, and political manipulation; the exacerbation of inequality and of many forms of bias in society; the creation and deployment of lethal autonomous weapons; and the usurpation of human roles in the economy and in social relationships.”

“My concern here, however, is with the potential consequences of success in creating general-purpose AI: that is, systems capable of quickly learning to perform at a high level in any task environment where humans (or collections of humans) can perform well.” Decades ago, Alan Turing expressed a similar concern. While answering a question during a lecture in 1951, Turing said: “It seems probable that once the machine thinking method had started, it would not take long to outstrip our feeble powers. … At some stage therefore we should have to expect the machines to take control.” Over the past decade, a number of public figures like Stephen Hawking and Nick Bostrom have argued that a superintelligent AI could become difficult or impossible for humans to control, and might even pose an existential threat to humanity.

Until recently, we didn’t have to worry about the consequences of such superhuman AIs because AI applications weren’t all that smart and were mostly confined to research labs. But, powerful AI systems have now matched or surpassed human levels of performance in a number of application- specific tasks such as image and speech recognition, skin cancer classification, breast cancer detection, and complex games like Jeopardy and Go. While these application-specific breakthroughs, - generally referred to as soft, or narrow AI, - were once viewed as the exclusive domain of humans, AI still lacks the deeper, general-purpose human intelligence that’s long been measured in IQ tests.

The 2022 Work Trend Index: How to Get Hybrid Right

“The people returning to the office are not the same as those who left in March 2020,” wrote Jared Spataro, Microsoft Corporate VP for Modern Work, in a recent Harvard Business Review (HBR) article, 5 Key Trends Leaders Need to Understand to Get Hybrid Right. The article summarized the findings of Microsoft’s 2022 Work Trend Index, a report based on various sources of information, including a survey of 31,000 people in 31 countries, labor trends from LinkedIn, and trillions of signals from Microsoft 365 suite of productivity apps.

“The data shows that the past two years left an indelible mark on the psyche of workers, altering their expectations and fundamentally shifting the way work gets done,” said Spataro. “This is a pivotal moment, where leadership matters more than ever. Those who embrace a new mindset and shift cultural norms will best position their people and their business for long-term success. Adapting to the influx of change in this business environment is no easy feat.”

“After many false starts, organizations are finally making the shift into a true hybrid work model. Like every other turning point over the past two years, there’s no shortage of perspectives on what this next phase will look like.” However, there’s no going back to the way work once was. Adapting to the many changes in this business environment is a challenge now facing every leader. To navigate the transition to hybrid work, leaders need to understand five key trends.

How Can You Trust the Predictions of a Large Machine Learning Model?

Artificial intelligence has emerged as the defining technology of our era, as transformative over time as the steam engine, electricity, computers, and the Internet. AI technologies are approaching or surpassing human levels of performance in vision, speech recognition, language translation, and other human domains. Machine learning (ML) 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. 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. But such statistical methods are not equally suitable for all tasks. Tasks that are particularly suitable for machine learning, exhibit several key criteria, such as the availability of large data sets of well-defined input-output pairs for training ML classifiers, - e.g., carefully labeled cat, not-cat pictures for cat recognition classifiers, and english-french document pairs for machine translation algorithms.

The methods behind a machine learning prediction, - subtle adjustments to the numerical weights that interconnect its huge number of artificial neurons, - are very difficult to explain because they’re so different from the methods used by humans. The bigger the training data set, the more accurate the prediction, but the more difficult it will be to provide a detailed, understandable explanation to a human of how the prediction was made.

A few weeks ago I attended an online seminar, - How Can You Trust Machine Learning? by Stanford professor Carlos Guestrin, - on the difficulty of understanding machine learning predictions. Guestrin’s seminar was based on a 2016 article he co-authored with Marco Tulio Ribeiro and Sameer Singh , “Why Should I Trust You?” Explaining the Predictions of Any Classifier.

The 2022 AI Index Report: AI’s Emerging Real-World Impact

On March 16, Stanford’s Institute for Human-Centered Artificial Intelligence (HAI) released the 2022 AI Index report, its fifth annual study on the impact and progress of AI. The comprehensive study was led by an interdisciplinary group of experts from across academia and industry, and highlights the rapid rate of technical and marketplace AI advances, as well as the growing ethical and regulatory concerns associated with AI.

“2021 was the year that AI went from an emerging technology to a mature technology - we’re no longer dealing with a speculative part of scientific research, but instead something that has real-world impact, both positive and negative,” said Jack Clark, co-chair of the AI Index. “This year’s AI Index tells us that AI is being integrated into the economy and the effects of it are beginning to go global across research, deployment, and even funding.” Two weeks ago I attended an online HAI seminar where Clark gave this excellent presentation on the AI Index.

The report is organized into five chapters: Technical Performance, AI Ethics, R&D, Economy and Education, and Policy and Governance. Let me summarize the key findings of each chapter.

The Tech Trends that Will Most Reshape Companies Over the Next Decade

“In the next decade, we’ll experience more progress than in the past 100 years combined, as technology reshapes health and materials sciences, energy, transportation, and a wide range of other industries and domains,” notes a recent McKinsey study on the The top trends in tech. “The implications for corporations are broad.”

“These trends may not represent the coolest, most bleeding-edge technologies. But they’re the ones drawing the most venture money, producing the most patent filings, and generating the biggest implications for how and where to compete and the capabilities you need to accelerate performance.”

“Unifying and underlying them all is the combinatorial effect of massively faster computation propelling new convergences between technologies; startling breakthroughs in health and materials sciences; an array of new product and service functionalities; and a strong foundation for the reinvention of companies, markets, industries, and sectors.”

What Is Web3, and Could It Usher a New Economic System?

Bitcoin was introduced in 2008 with the release of Bitcoin: A Peer-to-Peer Electronic Cash System, which explained how to design a decentralized cryptocurrency and digital payment system without the need for central banks or trusted intermediaries. Blockchain, the digital ledger for managing and certifying the validity of bitcoin transactions, was introduced in the same paper.

Over the years blockchain has transcended its original objectives, and has evolved in two major directions. One continues to focus on blockchain as the underlying platform for bitcoin, but it’s also become the platform for the large number of cryptocurrencies, digital tokens, and other cryptoassets that have since been created. The other is focused on the use of blockchain as a trusted distributed data base for private and public sector applications involving multiple institutions, such as supply chains, financial services and healthcare. The cryptocurrency camp is based on public permissionless blockchains, which anyone can join and require some kind of proof-of-work or proof-of-stake systems. The multi-institution camp is based primarily on private permissioned blockchains where participation is restricted to the institutions transacting with each other.

Given my professional interests, I’m particularly interested in the use of blockchains in business and public sector applications for two major reasons. First, I consider blockchain as a major next step in the continuing evolution of the Internet, helping us enhance the security of Internet transactions and data by developing a layer with the standard encrypted services for secure communication, storage and data access. And, longer term, blockchain technologies can significantly improve the efficiency, resilience, and management of complex global applications involving multiple institutions.

But, while I’ve been somewhat skeptical of the crypto camp, a number of intriguing crypto-related topics have recently captured my attention, including Non-Fungible Tokens (NFTs), Decentralized Finance (DeFi), and especially Web3. These various topics were nicely explained by technology columnist Kevin Roose in The Latecomer’s Guide to Crypto, an article in the March 20 SundayBusiness section of the NY Times.

The Widening Gap Between Technological and Institutional Change

A few weeks ago I read The Exponential Age Will Transform Economics Forever, an article in Wired by Azeem Azhar based on his recently published book The Exponential Age. Azhar’s central thesis is that the chasm between what technology allows us to do and what our institutions are prepared to handle has been rapidly widening. New technologies are being invented and scaled at an ever-faster pace. But our institutions, - including our economic systems, political organizations and social norms, - are changing much more slowly. While technological advances follow an exponential curve, institutional adaptation follow a straight, incremental line.

The gap between technological and institutional change is nothing new. Ever since the advent of the Industrial Revolution, there’s been a significant time lag between the emergence of a transformative technology and its ensuing impact on economies and societies. Even after reaching a tipping point of market acceptance, it takes considerable time, - often decades, - for their benefits to be fully realized.

In The Productivity J-Curve, a 2018 NBER paper, Erik Brynjolfsson, Daniel Rock, and Chad Syverson  explained that general purpose, transformative technologies, - such as the steam engine, electricity, and semiconductors, - “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.”

The life cycle of historically transformative technologies consists of two district phases, investment and harvesting. Since the technologies are general purpose in nature, they require massive complementary investments, including business process redesign, co-invention of new products and business models, and the re-skilling of the workforce. Moreover, the more transformative the technologies, the longer it takes for them to reach the harvesting phase when they are widely embraced by individuals, companies and industries across the economy.

The Impact of Open Source on the EU Economy

In the 1990s, open source was generally viewed as experimental approach to software development mostly used by research and academic communities on emerging projects like the Internet and Linux. Open source has since evolved to become a well-accepted model of economic production across just about all private- and public-sector communities around the world. But, while widely used, what’s been the economic impact of open source on nations and firms? This question was addressed in a recent study on the impact of open source on the European economy.

“The increasing relevance of Open Source (OS) during the last two decades requires an update of an in-depth analysis of its current role, position and its potential for the European economy,” said the European Commission (EC) in its report on The impact of open source software and hardware on technological independence, competitiveness and innovation in the EU economy.

“Whereas Open Source Software (OSS) has become mainstream across all sectors of the software industry during the past 20 years, Open Source Hardware (OSH) is still in an emerging phase. However, the business ecosystem for OSH is developing.”

The EC conducted a comprehensive analysis of the commercial uses, costs, and benefits of OSS and OSH. “On the basis of this information, the study assesses the potential for the European Union (EU) to achieve its policy goals (including economic growth, greater competitiveness, innovation, and job creation) through the use, promotion and support of OSS and OSH.”

The Rise and Fall of Nation-States

Several weeks ago I attended The End of Nation-States, by technology executive and consultant Tomás Pueyo, - part of the Stanford Digital Economy Lab seminar series. In May of 2021, Pueyo launched Unchartered Territories, a newsletter he describes as aiming to explore the unchartered territories of a fast changing world “to know how we can prepare for them.”

His seminar discussed the role of information technologies in the rise of nation-states throughout history, and how information technologies are likely to lead to the end of nation-states in the coming decades. Let me summarize Pueyo’s key arguments based on his talk and two related newsletters.

The Rise of Nation-States

In Internet and Blockchain Will Kill Nation-States, published in August of 2021, Pueyo explained how the printing press led to the emergence of nation-states in the 16th century.

The feudal system was the basic structure of society between the 9th and 15th centuries in medieval Europe. Feudalism was based on the relationship between a landed-aristocracy, - e.g., kings, dukes, counts, - and the vassals under their protection who vowed them allegiance. Power in the feudal system was hyperlocal and widely distributed across different landowners. Communication was difficult across different localities, since they each generally spoke different dialects or altogether different languages.

The Catholic Church was the biggest power in medieval Europe. The root of its power was the Church's well organized hierarchy of priests, bishops, archbishops, cardinals, and the Pope, all of which were able to communicate with each other far better than everyone else, - aristocrats and vassals alike.

How Do Digital Infrastructures Increase a Firm’s Value?

“In the information age, the value of a firm rests fundamentally on how it stores, shares and processes information,” wrote Seth Benzell, Jonathan Hersh, and Marshall Van Alstyne in a recent paper, How APIs Create Growth by Inverting the Firm. “Digital infrastructure is therefore central to a firm’s success.”

But, how do digital infrastructures increase a firm’s value? One method is to make more efficient use of the firms internal capital by modularizing, reconfiguring and reusing the firms IT resources. A second method, used in platform businesses, is to leverage external capital by giving access to the firm’s resources to third parties, and capturing a share of the resulting surplus. Both methods require the implementation of Application Programming Interfaces (APIs), whether for only for internal use by the firm’s developers or exposed externally to the firm’s platform partners. Both methods also feature a modular architecture of remixable resources. “Modularity combines the advantages of standardization typically associated with high volume processes together with the advantages of customization typically associated with bespoke processes.”

“An API is a set of routines, protocols, and tools that standardizes building software applications compatible with an associated program or database. APIs fundamentally are code that control access to information. … They govern the type and format of calls or communications that any application can make of another associated program. The answering program is agnostic about the source of the call, yet can require access permission, and the calling program need not know anything about the internal workings of the answering program. … When designing an API, the architect decides how much of the computer system or data to expose to which users. Being accessible on the web, these API endpoints act as a constant conduit to business processes that the firm itself specifies.”

How Does AI Improve Human Decision-Making?

“Artificial Intelligence (AI) is the newest general-purpose technology (GPT),” wrote Sukwoong Choia, Namil Kim, Junsik Kim, and Hyo Kang in a recent paper, How Does AI Improve  Human Decision-Making? Evidence from the AI-Powered Go Program. “One remarkable characteristic of AI is its ability to provide humans with high-quality predictions at relatively low cost and to automate a wide range of predictions. AI has already outperformed human professionals in many domains - including strategic gameplay, medical diagnosis, and new drug development.” 

“Furthermore, as a GPT, the domains where AI outperforms humans are expanding at a fast pace. The rapid development and adoption of AI thus raise an interesting yet pressing question about how AI affects human tasks in these various domains. Studies of AI and human capital, for instance, have examined the performance gap between humans and AI, highlighting AI’s potential for replacing jobs.” However, we should keep in mind that new technologies have been replacing workers and transforming economies for over two centuries. But, over time, these same technologies led to the creation of whole new industries and new jobs, increasing productivity and economic output, raising earnings, and augmenting the demand for labor.

Workers can increase their productivity and performance when assisted in their jobs by AI-based tools. But beyond its assistant role, AI can play a major instructional role by training human professionals to make better decisions. Recent technology advances have significantly improved the quality and reduced the cost of AI-based predictions.

The paper raises a number of important questions about the use of AI to improve human decisions making. Can we show quantitative evidence that AI-based predictions actually improve the quality of human decisions? If so, by what mechanisms are AI-based human decisions improved? And, do the improvements vary between depending on ones attitude toward AI-based decisions?

The Promise & Peril of Human-Like Artificial Intelligence

In his 1950 seminal paper, Computing Machinery and Intelligence, Alan Turing proposed what’s famously known as the Turing test, - a test of a machine’s ability to exhibit intelligent behavior indistinguishable from that of a human. If a human at a keyboard couldn’t tell whether they were interacting with a machine or a human, the machine is considered to have passed the Turing test. “Ever since, creating intelligence that matches human intelligence has implicitly or explicitly been the goal of thousands of researchers, engineers and entrepreneurs,” wrote Erik Brynjolfsson, - Stanford professor and Director of the Stanford Digital Economy Lab, - in a recent article, The Turing Trap: The Promise & Peril of Human-Like Artificial Intelligence.

“The benefits of human-like artificial intelligence (HLAI) include soaring productivity, increased leisure, and perhaps most profoundly, a better understanding of our own minds. But not all types of AI are human-like - in fact, many of the most powerful systems are very different from humans - and an excessive focus on developing and deploying HLAI can lead us into a trap. … On the one hand, it is a path to unprecedented wealth, increased leisure, robust intelligence, and even a better understanding of ourselves. On the other hand, if HLAI leads machines to automate rather than augment human labor, it creates the risk of concentrating wealth and power. And with that concentration comes the peril of being trapped in an equilibrium where those without power have no way to improve their outcomes, a situation I call the Turing Trap.”

Over the past decade, powerful AI systems have matched or surpassed human levels of performance in a number of tasks such as image and speech recognition, applications like skin cancer classification and breast cancer detection, and complex games like Jeopardy and Go. These AI breakthroughs are generally referred to as soft, narrow or specialized AI, inspired by, but not aiming to mimic the human brain. They’ve been generally based on machine learning, that is, on the analysis of vast amounts of information using powerful computers and sophisticated algorithms, whose results exhibit qualities that we associate with human intelligence.

Is It Possible to Establish Norms for Responsible State Behavior in Cyberspace?

In mid-December, the Council on Foreign Relations sponsored a virtual roundtable with Joseph Nye, - former dean of Harvard’s Kennedy School of Government, - to discuss his recent Foreign Affairs article The End of Anarchy?: How to Build a New Digital Order. Professor Nye has long been regarded as one of America’s preeminent strategic thinkers and political scientists. In the 1970s he chaired the National Security Council Group on Nonproliferation of Nuclear Weapons, and over the past decade he’s brought his expertise to the study of conflict and deterrence in cyberspace.

Cybersecurity is an increasingly important aspect of the of US national security strategy, including global trade and the protection of our critical infrastructures. In June of 2021, FBI Director Christopher Wray compared the danger of ransomware attacks on US firms by Russian criminal groups to the September 11 terrorist attacks. And, in a July editorial, the NY times said that ransomware attacks have emerged as “a formidable potential threat to national security,” given “their ability to seriously disrupt economies and to breach strategically critical enterprises or agencies,” urging governments that “It is a war that needs to be fought, and won.”

At an MIT conference in February of 2019, former US Secretary of State Henry Kissinger was asked if we need cybersecurity control agreements with Russia, China and other nations similar to the nuclear arms control agreements that he spent so much time negotiating during the Cold War. Dr. Kissinger replied that for arms control to be effective, the two sides needed to share information and agree to inspections. But such mechanisms are harder to apply in the digital world, because the transparency that was essential for arms control would be very hard to establish for cyber threats. In addition, while controls of physical arms are relatively explicable and negotiable, the variety and speed of cyber attacks make it much harder to develop adequate control agreements.

A Blockchain-based Framework for Safeguarding the Integrity of Real-World Assets

Blockchains first came to light in 2008 as the architecture underpinning bitcoin, the best known and most widely held cryptocurrency. It’s a truly brilliant architecture built on decades-old fundamental research in cryptography, distributed data, game theory and other advanced technologies. The blockchain’s original vision was limited to enabling bitcoin users to transact directly with each other with no need for a financial institution or government agency to certify the validity of the transactions. But, like the Internet and other transformative technologies, blockchain has now transcended its original objectives.

Blockchains are a kind of distributed ledger technologies (DLT), which also include non-blockchain DLTs. Over the past decade, an increasing number of people, including me, consider blockchains and DLTs as major next steps in the evolution of the Internet. In 2016 the World Economic Forum (WEF) named The Blockchain in its annual list of Top Ten Emerging Technologies citing its potential to fundamentally change the way markets and governments work. “Like the Internet, the blockchain is an open, global infrastructure upon which other technologies and applications can be built,” said the WEF. “And like the Internet, it allows people to bypass traditional intermediaries in their dealings with each other, thereby lowering or even eliminating transaction costs.”

The Internet is a general purpose network platform that unlocked huge innovations and economic value by significantly lowering the cost of connections and supporting a large variety of applications. A major reason for its ability to support so many different kinds of applications is that the Internet’s foundations, its TCP-IP layer, has stuck to its basic data-transport mission, i.e., just moving bits around.  It has no idea what the bits mean or what they’re trying to accomplish.

The design decisions that shaped the Internet back in the mid-late 1980s didn’t optimize for security and privacy, or for the ability to authenticate transactions between two or more parties. That’s all the responsibility of the applications running on top of the TCP/IP layer, and they each generally do it their own way, sometimes not at all. Not surprisingly, the lack of standards for security, privacy and transactional integrity have been the biggest challenges facing the Internet in our fast growing digital economy.

What’s the State of Our Digital Technology Revolution?

I first met Carlota Perez in the mid-2000s when she gave a seminar at IBM based on her 2002 bestseller Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages. At the time, we had been living through a series of major changes, - the explosive growth of the Internet, the advent of the digital economy, the dot-com bubble, and the bursting of the bubble. In her excellent presentation Perez explained these turbulent times by positioning them within the historical perspective she wrote about in her book.

If you look at the historical big picture, patterns begin to emerge which serve as a good guide for understanding the past and thinking about the future. Since the onset of the Industrial Revolution we’ve had 5 major technological revolutions, each one lasting roughly 40 to 60 years. First was the age of machines, factories and canals starting in 1771. This was followed by the age of steam, railways, iron and coal, starting in 1829; steel, electricity and heavy engineering in 1875; oil, automobiles, and mass production in 1908, and our present information technology and telecommunications (ICT) digital age starting in 1971.

Technology revolutions are engines of growth, ushering new paradigms for innovation, rejuvenating and transforming the economy, and re-shaping social behavior and the institutions of society. To better understand the dynamics of a technology revolution, we should split it into two different periods, each lasting 20 to 30 years.

A New Measurement Framework for the Digital Economy

Several weeks ago I heard a very interesting keynote presentation by Cambridge professor Diane Coyle at the annual workshop of the Stanford Digital Economy Lab, - What Don’t We Know About Measuring the Digital Economy? Professor Coyle is also a research associate of the UK Economics Statistics Centre of Excellence. In 2017 she was a recipient of the Indigo Prize on how to measure economic activity in the digital economy for her essay Making the Future Count, co-authored with Benjamin Mitra-Kahn.

“GDP captures only market transactions at the price of exchange, and not the welfare gains, externalities, environment, distribution of wealth or innovation which occurs in an economy,” wrote the authors. “Hence almost since its creation in the 1940s it has been criticised for its inability to capture economic welfare. Now changes in the economy, being restructured by digital technology and paying the price for unsustainable growth, make the case for a new measurement framework more pressing than ever. GDP was never an ideal measure of economic welfare and its suitability has been decreasing.”

Gross Domestic Product (GDP) became the accepted international measure of economies in the1940s. While being a good measure for the 20th century industrial economy, GDP is a flawed measure for the 21st century economy. It was suitable when the economy was dominated by the production of physical goods, but GDP doesn’t adequately capture the growing share of services and other intangible assets that now characterize advanced economies. Nor does it reflect important economic activity beyond production, such as income, consumption and living standards.