The Lawfare Podcast: Scaling Laws – The AI Economy and You

How AI Is, Will, and May Alter the Nature of Work and Economic Growth
Guests: Anton Korinek (Univ. of Virginia; Anthropic Economic Advisory), Nathan Goldschlag (Economic Innovation Group), Bharat Chander (Stanford Digital Economy Lab)
Host: Kevin Frazier (AI Innovation and Law Fellow, Texas Law; Lawfare Institute)
Date: November 14, 2025

Overview

This episode tackles a timely and crucial question: How is AI reshaping the workforce, now and in the future? Economic experts Anton Korinek, Nathan Goldschlag, and Bharat Chander join host Kevin Frazier to analyze what current data does and doesn’t tell us, identify where consensus among economists (if any) lies, and explore how labor, productivity, policy, and society may change as AI advances from complementary chatbot tools to potentially autonomous work agents.

Key Discussion Points & Insights

1. Economic Consensus: How Do Economists View AI’s Impact?

[06:21-08:45]

• Productivity vs. Labor Angle:
  • Median economists tend to be more skeptical than technologists about AI's near-term productivity boosts.
  • “The median economist is probably more skeptical of productivity impacts than the median technologist, as you might expect.” – Bharat Chander [06:59]
• Early Impacts:
  • No large, economy-wide labor disruptions yet, but young workers in AI-exposed sectors (e.g. software dev, customer service) are already affected.
  • Reference: Chander’s “Canaries in the Coal Mine” paper on declining employment for entry-level, AI-exposed workers.
  • “Employment in these occupations has been declining...there seems to be kind of a robust relationship between these employment changes for entry level workers and AI exposure.” – Bharat Chander [08:18]

2. The Amazon Layoff Lens: How To Interpret Big Tech Layoff Headlines?

[08:45-11:08]

• Contextualizing Layoffs:
  • Citing AI as a cause for layoffs is more palatable to shareholders than alternatives (over-hiring, sluggish sales).
  • The labor market is vast; tens of thousands of layoffs at one firm are “a drop in a bucket” given millions of jobs turn over quarterly.
• No Clear Displacement Evidence Yet:
  • “When you look at these for evidence of displacement effects for the economy overall, we just aren't seeing it yet.” — Nathan Goldschlag [10:43]

3. Augmentation vs. Automation: How Is AI Changing Work?

[12:02-13:55]

• Chatbots → AI Agents:
  • Chatbots are complementary (helping humans do tasks better), but autonomous agents could substitute for human workers.
  • “We are moving from chatbot interactions to AI agents...that is going to fundamentally change the nature in which AI will affect labor markets.” — Anton Korinek [13:27]
• Definition of ‘Agent’: “An AI agent that can do any task you can do on a computer...for a wide swath of jobs...this isn’t just augmentation, this is complete wholesale adoption of the key tasks of that job.” — Kevin Frazier [13:55]

4. Is AI “Different” From Past Technological Change?

[15:28-17:23]

• Historical Parallel:
  • Innovation always creates and destroys jobs, but so far, societies have adjusted.
• What’s New About AI?:
  • AI may be unique if the jobs created to replace automation are also filled by AI, reducing human labor demand.
  • “As models are improving and becoming more agentic...you could imagine a world in which the new work that’s created...is also being done by AI.” — Bharat Chander [16:05]
• Enormous Uncertainty:
  • Uncertainty about technological direction means scenario planning is crucial.

5. Data & Measurement: What Do We (Not) Know?

[17:23-21:26]

• Adoption Rates:
  • Firm-level rates of AI use are still low: ~9–10% overall; higher in information sectors (25%+). But individual exposure/use is higher.
• Need For Better Data:
  • Economists want richer data: more granular tracking of what work is being automated, how workers/firms use AI, and how impacts play out over time.
  • “We need to be able to see adoption and deadoption over time to get a sense of how the adopters look different...and how their trajectories change.” — Nathan Goldschlag [19:59]

6. Planning for Uncertainty: Modeling, Scenario Planning & Policy

[21:26-32:17]

• Frontier Advances:
  • Rapid progress at AI’s “frontier” could mean bigger, sooner labor effects than current backward-looking data suggests.
• Economists are focusing on scenario-based approaches to policy modeling, so society is not caught “flat-footed,” even if the predictions are uncertain.
  • “Given that the technology is evolving much more rapidly than any prior technology, it is crucial to kind of embrace that uncertainty and...be prepared for radical scenarios.” – Anton Korinek [22:37]

7. Policy Questions and Societal Wellbeing

[28:20-32:17]

• Tech Progress Correlates with Prosperity...But That’s Hard to Sell:
  • “It’s not exactly a compelling political message...to say, look, Amazon employees, you’ll be fine in 10 years.” – Kevin Frazier [29:41]
• Modeling Policy Responses:
  • Simulations can help anticipate sector-by-sector worker shifts, informing policies beyond UBI — such as retraining or creative labor market interventions.
• Early Careers & Education:
  • There’s an urgent research need on how AI is changing education choices and skill demand. For now, young people should take advantage of AI as a learning tool and focus on adaptability and social skills.

8. Economic Dynamism & Reallocation

[35:09-38:00]

• Declining U.S. Business Dynamism:
  • Last 40 years saw a slowdown in new business launches and job switching, though Covid temporarily reversed this.
  • “New businesses are usually more likely to introduce new ideas...they play a disproportionate role in job creation.” – Nathan Goldschlag [36:07]
• Key Statistic to Watch:
  • Are people and firms reallocating in response to AI? Healthy economies feature lots of movement.

9. Bad & Good Policy Ideas

[39:32-46:34]

• Worst Ideas:
  • Just prescribe more “dynamism” if AI fully displaces labor may not make sense (Korinek).
  • Degrowth — intentionally stalling tech or economic growth out of AI fears — is a “terrible idea.” (Chander, Goldschlag)
  • Freezing the job distribution “in amber” and resisting occupation shifts is misguided; reallocation is normal and essential.
• What matters:
  • Support for individuals, not just occupations; maintain long-term focus on societal wellbeing, not just GDP.

10. Who’s Doing It Right?

[46:34-50:54]

• Census Bureau’s Early AI Survey:
  • U.S. Census started measuring AI adoption pre-ChatGPT, giving researchers precious longitudinal “before/after” data.
• Transparency from AI Labs & Academia:
  • OpenAI’s GDP Val and Anthropic’s Economic Index provide valuable, public task-level data, but economists want even more robust, open data sets.

11. Closing Reflections: Embrace Scenario Planning

[50:54-51:36]

• Economists hate to predict, but scenario planning is now imperative, even with uncertainty.
  • “Most of our predictions are going to turn out to be false, [but] I think it is crucial to engage in scenario planning to head into this rapidly changing future.” – Anton Korinek [50:56]

Notable Quotes & Memorable Moments

• “AI only works if society lets it work.” – Kevin Frazier [04:32]
• “The main mode that workers interact with AI is the chatbot format...we are moving from chatbots to AI agents...that is going to fundamentally change...AI will become a technology that actually displaces work.” – Anton Korinek [13:03]
• “We value work for reasons other than just money.” – Bharat Chander [43:32]
• “There’s this idea...can we freeze the economy in amber?...That’s a mistake.” – Nathan Goldschlag [45:20]
• “New businesses are usually more likely to introduce new ideas...they play a disproportionate role in job creation.” – Nathan Goldschlag [36:07]
• “If we only studied the diffusion with a rear view mirror...it doesn’t tell you what we should prepare for six or twelve months from now.” – Anton Korinek [21:56]

Timeline of Key Segments

| Timestamp | Segment/Topic | |---|---| | 06:21–08:45 | Economic consensus, early impacts (“Canaries in the Coal Mine”) | | 08:45–11:08 | Amazon layoffs – is it really AI? Labor market context | | 12:02–14:35 | Augmentation vs. automation, emergence of AI agents | | 15:28–17:23 | Historical parallels, is AI fundamentally different? | | 18:07–21:26 | Data needs: adoption rates, surveys, frontier use vs. diffusion | | 21:26–23:49 | Frontier model advances, scenario planning, future prep | | 28:20–32:17 | Policy modeling, logic of scenario planning for policymakers | | 32:17–34:17 | Early career advice, education impacts, future skills | | 35:09–38:00 | Dynamism, reallocation of labor, business formation | | 39:32–46:34 | The “worst ideas”: degrowth, freezing jobs, limits of dynamism | | 46:34–50:54 | Good practices: Census Bureau’s stats, lab/academic transparency | | 50:54–51:36 | Importance of scenario planning |

Tone & Style

The conversation is substantive, occasionally humorous, and forthright about uncertainty: the guests emphasize evidence, humility in forecasting, and continuity with (as well as potential breaks from) past technological disruptions. The economists are practical and direct, noting data limitations and admitting what’s simply unknown.

Summary Takeaways

• So far, AI’s labor disruptions are narrow and localized, not broad-based.
• Augmentation can easily tip to automation as AI agents become more autonomous and capable.
• Reallocation and economic dynamism—not occupational stasis—are key to healthy adaptation.
• Best policies will support workforce transitions, not cling to status quo jobs or halt progress.
• Scenario planning is essential, given both the speed of AI advances and profound future uncertainty.

This episode provides a frank, data-driven look at how AI is—and crucially, is not yet—reshaping work, and what both policymakers and individuals should consider as the future unfolds.