Making AI Work: Productivity, Diffusion, and Policy

Long hallway with computers lining both sides and overhead lighting.

Posted in Events News  |  Tagged

Georgetown on the Hill Tackles AI’s Structural Bottlenecks

Video link: https://www.youtube.com/live/egWkiioxKko?si=2AxVy_wPP_8p4mTr

Experts from academia and industry gathered in the Rayburn House Office Building to address the empirical realities of Artificial Intelligence (AI). Hosted by Georgetown McDonough’s Center for Business and Public Policy, the discussion revealed that while AI is already driving massive productivity gains, its long-term success is threatened by aging infrastructure, a shifting skills landscape, and significant measurement gaps.

The Productivity “Innovation”

Carol Corrado, senior policy scholar at the center, presented her research showing that AI is not just a tool, but an innovation in the process of innovation.

  • Historical Context: AI’s current impact is already larger than steam or electricity at comparable stages of diffusion.
  • The “Use” Channel: Two-thirds of AI’s productivity effect comes from firms reorganizing their R&D, marketing, and logistics, rather than just purchasing technology.

Infrastructure: The Grid as a Growth Barrier

Olivia Igbokwe-Curry (AWS) shared about Amazon’s staggering $25 billion investment in Mississippi alone, underscoring the exponential demand for compute. However, Rachel Mural (Harvard Kennedy School) warned that the U.S. energy grid is a primary bottleneck to AI diffusion, noting that “a data center can come online in 18 to 24 months, but transmission development can take 7 to 10 years.”

The Human Element: Skills Over Displacement

Josh Connley of LinkedIn challenged the popular narrative of mass job displacement. Instead, LinkedIn data suggests a radical shift in skill composition.

  • Skill Velocity: Since 2015, 25% of the skills required for the same job titles have changed. By 2030, that number is projected to hit 70%.
  • Technical Demand: AI technical roles now make up 7% of all tech jobs on the platform, with a 70% year-over-year increase in jobs requiring AI literacy.
  • Economic Stakes: Closing the AI talent gap could unlock $4.1 trillion in productivity for the U.S. economy.

Policy Solutions and “Myth-Busting”

The panel concluded by addressing the “myths” hindering effective policy. Panelists urged lawmakers to focus on organizational readiness rather than just hardware supply.

  • Measurement Gap: Corrado argued that current GDP statistics fail to capture the full scope of AI capital and R&D investment, leading to “consequential decisions about a technology we can’t fully see.”
  • Diffusion Lag: Despite leading in frontier models, Connley noted the U.S. ranks No. 24 in AI diffusion (actual uptake by companies), suggesting the “race” is further from won than many believe.
  • Public-Private Partnership: Industry leaders emphasized that grid modernization and workforce training must be collaborative, with Igbokwe-Curry highlighting Amazon’s Ratepayer Protection Pledge to ensure local consumers don’t foot the bill for infrastructure buildouts.

Georgetown on the Hill will return in the fall for discussions on Cloud Computing and AI followed by a session on AI and Antitrust.