Stanford ECON295/CS323 I 2024 I A World Without Work, Daniel Susskind

Daniel Susskind and AI

• Daniel Susskind is a research professor of Economics at King's College in London and a senior research associate at The Institute for ethics and AI at Oxford University. (12s)
• Susskind is also the author of the books A World Without Work and The Growth Reckoning. (22s)

AI and Technological Unemployment

• Susskind believes that artificial intelligence has the potential to cause technological unemployment, similar to how the combustion engine replaced horses in transportation. (4m27s)
• Machines are steadily encroaching upon an increasing number of tasks and activities that were once exclusive to humans, encompassing manual, cognitive, and emotional capabilities. (17m9s)

Economic Growth and Automation Anxiety

• Economic growth is a recent phenomenon, with most of human history experiencing relatively stagnant economic activity. (6m49s)
• The Industrial Revolution, starting around the 1760s, marked a period of significant economic growth led by Britain, driven by new machines that improved production efficiency. (7m34s)
• Automation anxiety, the fear that machines replacing human labor would lead to job losses, has been present since the Industrial Revolution and continues today. (8m1s)

AI's Capabilities and Task Automation

• One way to analyze the capabilities of machines is to examine the tasks themselves and determine if they possess characteristics that make them suitable for automation. (14m8s)
• Between 1980 and 2012, jobs in the US that require a high level of human interaction grew by about 12% as a share of the workforce. (13m51s)

AI's Evolution: From Replicating Humans to Performing Tasks

• In the 1980s, the speaker's father and Philip Capper, an expert in a complex law called the Latent Damage Act, collaborated to create the first commercially available AI system for law. (20m0s)
• The AI system, designed as a decision tree, was based on Capper's expertise and aimed to make understanding the law more accessible. (20m37s)
• The prevailing approach to AI in the 1980s involved observing human experts and replicating their thought processes, but this method faced limitations when experts struggled to articulate their expertise, as exemplified by chess champion Garry Kasparov's inability to fully explain his chess prowess. (21m57s)
• Experts did not anticipate the exponential growth of processing power that began in the 1950s. (25m29s)
• Deep Blue, a chess-playing computer, defeated Garry Kasparov in 1997 by using its superior processing power to calculate millions of moves per second, demonstrating that machines do not need to replicate human thinking to outperform them. (25m50s)
• The success of Deep Blue marked a shift in artificial intelligence from trying to replicate human thought processes to focusing on building machines that can perform tasks effectively, regardless of how they achieve the outcome. (27m12s)

AI and Uncertainty

• People often seek expert judgment from professionals like doctors, lawyers, and accountants in situations of uncertainty, where facts are unclear, information is ambiguous, or they are unsure about the best course of action. (30m51s)
• Machines, particularly those using artificial intelligence, excel at handling large datasets and identifying patterns, enabling them to outperform humans in dealing with uncertainty, as exemplified by medical diagnostic systems like the one developed at Stanford that can diagnose skin cancer with accuracy comparable to leading dermatologists. (31m13s)
• The increasing capability of non-thinking machines, as demonstrated by IBM's Watson winning Jeopardy without possessing human-like thought processes, highlights a significant trend in artificial intelligence where machines perform complex tasks in fundamentally different ways than humans, expanding the realm of automatable activities. (33m30s)

Mismatches and Displaced Workers

• There are three types of mismatches that can lead to displaced workers: skills mismatch (36m1s), place mismatch (36m1s), and identity mismatch (36m35s).
• Identity mismatch occurs when displaced workers are unwilling to take on jobs that do not align with their existing work identity, such as men refusing "pink-collar" jobs traditionally held by women (36m54s).

Avoiding Excessive Automation

• Daron Acemoglu argues that "excessive automation" can be avoided if society consciously chooses a different technological path, suggesting that technological progress is not predetermined (40m8s).
• There are two paths for technological progress: one where technology substitutes for workers and one where technology complements workers. (42m5s)
• Taxes, subsidies, laws, regulations, social norms, and customs can be used to incentivize the development of technologies that complement rather than substitute for workers. (42m54s)

Limitations of Redirecting Technological Progress

• While attempting to redirect technological progress is worthwhile, there are limitations to this approach, as seen in the context of climate change and the development of clean technologies. (46m54s)

Challenges of a Future with Less Work

• There are three major problems to consider in a future with less work: economic inequality, the power of technology companies, and the challenge of finding meaning and purpose in life. (48m1s)
• With potentially less paid work available, the traditional method of income distribution through wages might become less effective, necessitating a larger role for the state in distributing wealth. (48m20s)
• The political power of large technology companies, and their influence on liberty, democracy, and social justice, is a growing concern. (50m11s)
• As work is a significant source of meaning and purpose for many, a future with less work could lead to a sense of emptiness, potentially requiring new societal structures and support systems. (52m3s)

Historical Context of Economic Growth

• For most of human history, global GDP per capita has hovered around a few hundred dollars. (53m39s)
• Technological progress has caused global GDP to increase to approximately $11,000-$12,000 per capita. (54m3s)

Government's Role in a Future with Less Work

• Daniel Susskind believes that as technology continues to increase productivity, governments will need to take a larger role in distributing wealth and may need to tax assets or technology to fund social programs like a basic income. (57m59s)
• There are differing opinions on the amount of a basic income, with those on the right supporting a minimal amount to simplify the tax system and those on the left advocating for a more substantial amount that allows individuals to thrive. (59m31s)

Distributive and Contributive Justice in a Basic Income System

• A universal basic income, while addressing distributive justice by promoting fair income distribution, raises concerns about contributive justice as it may undermine the social solidarity derived from individuals actively contributing to society through work. (1h0m34s)
• The implementation of a basic income could lead to the centralization of power, potentially diminishing the economic and political bargaining power of individuals who rely on it, making them dependent on the goodwill of the state. (1h1m51s)

Meaning and Purpose in a Future with Less Work

• Many people do not find meaning and purpose in work and would prefer to receive income from other sources. (1h5m46s)
• There is a possibility that people may use their increased leisure time for unproductive activities, raising questions about the role of the state in such a scenario. (1h6m33s)
• The pandemic provided a glimpse into a world with less work, as it led to a decrease in demand for many jobs, prompting individuals to seek meaning and purpose in activities like DIY projects and baking. (1h8m37s)
• Society understands the concept of gainful employment but lacks a clear understanding of gainful unemployment. (1h9m35s)

Crime and a Future with Less Work

• A significant concern is how a future without work would impact crime rates, given the correlation between legal employment and reduced crime. (1h10m23s)

The Current Challenge: Access to Work

• The current challenge is not a complete absence of work, but rather the existence of work that is inaccessible to many people for various reasons. (1h14m25s)

Interventions and Regulations for a Future with Insufficient Work

• The possibility of a future with insufficient work for everyone requires different interventions and regulations than those currently considered appropriate. (1h15m42s)

AI's Impact on Emotional Labor

• Economists may be underestimating the future capabilities of AI, particularly in roles requiring emotional capacity, such as kindergarten teachers, nurses, social workers, or therapists. (1h16m28s)

Regulation of AI Beyond Unemployment Concerns

• Regulation of AI is necessary now due to its impact on liberty, social justice, democracy, and worker surveillance, rather than solely due to potential unemployment concerns. (1h18m19s)

Rethinking Traditional Problem-Solving

• People often mistake the traditional way of solving problems (face-to-face interaction) with the problem itself. Technology is demonstrating that many problems can be solved without this traditional approach. (1h21m22s)
• While some services, like end-of-life care, are valued for human interaction, many tasks are not. Clients often prioritize efficiency and effectiveness over personal interaction. (1h21m14s)

Market Forces and Technological Development

• Market systems, prices, and incentives are powerful tools for encouraging technological development. These market forces can be leveraged to address significant challenges, including the impact of AI on work, climate change, and societal inequalities. (1h22m41s)