Nobel Laureate Daron Acemoglu's Economic Forecast: Why AI Will Automate Only 5% of Tasks and Add 1% to Global GDP

MIT Sloan Management Review

Introduction: AI’s economic impact predictions [0:00]

• The video introduces the widespread hype around AI and its predicted transformative impact on the economy.
• It then presents MIT economist and Nobel Laureate Daron Acemoglu's surprising, data-driven forecast which projects a much more modest impact.

Acemoglu’s 5% automation prediction [0:34]

• Acemoglu's research, published in a paper titled "The Simple Macroeconomics of AI," estimates that over the next 10 years:
  • Only about 5% of all tasks will be profitably automated by AI technology.
  • AI is likely to contribute just 1% to global GDP.

    

Why Acemoglu’s estimates differ from others [1:15]

• Acemoglu acknowledges the inherent uncertainty in these predictions due to the rapidly changing nature of AI technology.
• He states that despite recent advances, the industry has not yet produced AI applications that are "critical for the production process" or capable of generating "hugely valuable" new goods and services.

Comparing AI’s impact with the internet’s [2:27]

• Acemoglu contrasts AI's current state with the early days of the internet.
  • Internet: From its inception, it was clear how the internet would transform communication, introduce new goods and services, and provide platforms for various human interactions (production, recreation).
  • AI: These clear transformative pathways are not yet apparent for AI.
• Belief in Artificial General Intelligence (AGI) being "around the corner" might suggest otherwise, but even then, the integration of AGI tools into the production process remains unclear.

  

Which tasks AI can and cannot automate [3:42]

• Acemoglu identifies specific types of cognitive tasks that current AI approaches are well-suited for:
  • Tasks performed in predictable office environments.
  • Tasks that do not require much social interaction.
  • Tasks that do not require very high levels of judgment.
  • Examples: Basic software routines, IT security, accounting.
• Conversely, AI is not yet capable of performing or significantly contributing to:
  • Roles requiring high levels of judgment and social intelligence.
  • Examples: CEOs, CFOs, entertainers, professors, construction workers, custodial workers, or general blue-collar roles. These often involve tacit knowledge and complex real-world interactions that current AI and robotics cannot replicate.

How Acemoglu arrived at the 5% prediction [4:47]

• His calculation considers that approximately 20% of the economy is "at the cross hairs of AI to be automated" or could be significantly boosted by AI.
• However, many of these feasible tasks are:
  • Time-consuming to implement.
  • Performed in small companies where automation might not be profitable.
• These factors lead to the more conservative 5% automation figure.

The challenge of tacit knowledge in occupations [5:54]

• Every occupation contains complex parts rooted in "tacit knowledge" that require a lot of checking and different types of intelligence.
• AI's impressive achievements (like AlphaGo or AlphaFold) occur in domains with a "ground truth" that AI can learn if present in its training data.
• However, real-world tasks are far more complex:
  • They involve interactions and tacit knowledge that are hard to codify.
  • They require matching contextual understanding with specific problems (e.g., diagnosing difficult ailments, finding appropriate financial products).
• Current AI, particularly generative AI, can imitate human decision-makers by learning from vast datasets of their decisions.
  • This means AI can only be as good as the human data it learns from, and often cannot surpass the best human performance, especially when the "best" human decision-makers are not clearly identifiable.

AI’s effect on jobs in the next decade [8:35]

• Based on these limitations, Acemoglu predicts that no current occupation will be eliminated in the next 5 or 10 years.
• He argues that if one truly believes in the imminent "complete transformation" of the economy by AI, they should be able to list occupations that will entirely disappear, which he does not foresee.

A more pro-human approach to AI [9:29]

• Acemoglu argues that AI is not currently being developed in the "best possible way."
• A better, "pro-human" approach would focus on AI augmenting human decision-makers, recognizing where AI excels and where humans remain superior.

AI’s potential to create new services [9:56]

• The greatest promise of AI lies in its potential to create new goods and services and new ways of doing things for humans.
• Acemoglu highlights several major societal transformations where AI could play a role:
  • Aging Society: Meeting the growing demands for new goods, services, and accommodations for an older population.
  • Financial Inclusion: Providing better financial services for underserved populations.
  • Climate Change: Addressing the complexities of climate change mitigation and adaptation, which will require new tasks and innovations in production processes.
• These opportunities represent areas where AI could make a significant difference, moving beyond mere cost-saving.

Advice for business leaders: beyond the hype [10:53]

• Acemoglu advises business leaders: "Don't be taken by the hype." He views hype as an "enemy of business success."
• Instead of focusing on cost-cutting or blindly following competitors, leaders should:
  • Think about how their most important resource (human capital) can be better deployed.
  • Leverage human resources together with technology and data to increase efficiency and enable the creation of better and newer goods and services.
  • Focus on doing "new things that are so important in this changing world."

Avoiding blind AI investments [12:47]

• Acemoglu expresses concern that many business executives are investing in AI "blindly."
• This often happens without a clear understanding of how AI can be synergistically deployed with their workforce.
• Executives are often under "tremendous pressure" from consultants and media to invest heavily in AI to avoid falling behind competitors, which is not a recipe for sustained business success.

Working with employees to identify AI value [14:11]

• The recommended approach is to identify opportunities for new services with the help of the most skilled employees.
• This involves understanding market demand for these new services and how they can be made possible.
• AI then becomes a "great tool to augment the capabilities of your workforce and yourself" in achieving these innovations.