1:52 What’s the EPI historical view on technology and the economy? And why is it important to be clear about whether we think technology itself is the culprit for the problems workers face in our economy?

• EPI was essentially created to give a critical perspective on the conventional wisdom about technology and the economy
• Our work emphasized that policies and political institutions were far more important to the distribution of income and wages than technology – this was the exact opposite of lots of conventional wisdom. Since then, the world has moved our way on this question – until AI.
• Unbalanced labor market power is what makes technology—including AI—threatening to workers. The best “AI policy” to protect workers is boosting their bargaining position

5:51 What’s your alarm level on AI and its potential labor market effects and generally why is that? To set some parameters for this conversation: we are talking today really about AI and the economy, workers, and the labor market and those concerns specifically.

• Alarm level is pretty low – until proven otherwise, it’s one more technological shock hitting the US economy – but we’re hit by those all the time
• Technology tends to determine pace of overall growth and be good for that – it mostly does not determine the distribution of this growth – that’s determined by policies and institutions
• American policies and institutes for distributing growth equitably and making sure jobs have decent pay and working conditions are weak and bad – but they were weak and bad before AI.
• Not a lot of reason why AI will make this worse. If alarm about AI somehow leverages some political momentum to make these broader changes, great. It seems that more of the policy debate is not doing this but is trying to come up with tailored and AI-specific interventions.

11:44 How is AI affecting the U.S. economy?

• One is short-run macroeconomic effects based on spending flows associated with AI. We can be pretty precise and confident about this
• Longer-run effects on productivity and the labor market implications as AI either augments or replaces labor are highly uncertain – but today there is likely too much hyperbole about those.

13:50 What is the short-run macroeconomic effect of AI?

• There are two effects – capex (building data centers) and consumption fueled by stock bubble
• Capex effect smaller – not what you might guess from some commentary
• Together they’re about ½ of growth we saw in 2025

16:35 AI-related spending accounted for ½ of growth in 2025 – that means it’s good? Or not?

• It was a lucky macro card drawn by the Trump administration, for sure. Fallout of other policy choices have likely been masked by this AI spending surge
• This inflated spending is a narrow and fragile base of growth to rely on going forward
• It’s not too early to think through the recession playbook

19:30 There has been a lot of news coverage/attention given to AI-related spending. Does this wave of AI-related spending already make this the most economically significant burst of technology the US economy has ever seen? Where does this stack up vs. Technological changes in recent history?

• Both the late 1990s stock market bubble and capex associated with internet buildout were much bigger
• Real effects of that earlier period of tech-related spending turned out to be really significant in terms of productivity and other things; remains to be seen now

21:53 How should we organize our thoughts about AI and jobs?

• It is worth being very specific:
  • are we worried about overall, net job-loss in the economy (ie, a much higher unemployment rate)?
  • Or are we worried about lots of short-run churn (some occupations contract and others expand and there are transitional issues)?
  • Or are we worried that the job shifts of AI will leave some workers hard-pressed to find as-good jobs elsewhere in the economy without taking big wage cuts?
• We’re not very worried about overall net job-loss
• We are ALWAYS worried about churn and job-shifts – our country is unique among rich ones in hanging people out to dry when the labor market shifts beneath their feet

26:30 So the real story about jobs is how much AI might replace workers in specific occupations and the scale of the resulting flux in the labor market – what do we know about this scale of gross job destruction so far? In particular, there’s a lot of attention given to what’s happening to young workers, specifically young college graduates seeking entry-level white-collar jobs, and what’s happening to tech jobs in programming, etc.

• Direct studies of this are fairly neutral
• There is a general tendency to overread very ancillary data points
• It’s likely there is also a bit of “AI-washing” right now
• Class of 2026: Young college graduates face a more mixed labor market than headlines suggest

34:10 Given what we know about AI and our best guesses about how it will effect the economy, what should policymakers be doing now? There have been a lot of different proposals put forward in Congress at the federal level – some pretty sweeping, some that might not actually change that much. And there’s also a lot of activity at the state and local level, which is where most actual AI policy or regulation has actually passed.

• Action you take depends on what angle of rise from AI is concerning you
• Most of what should be done is focusing on all the broad policy areas where we’ve lagged behind and hurt workers
• Also, ensure that AI is not used as an excuse to further gut the capacity of the public sector

36:22 How do you put a check on AI companies or processes of implementation? How would that work?

• Implement a “token tax” to slow usage of and demand for AI
• Federal legislation that mandates data center companies make a net positive contribution, say to the electrical grid or water processing, to the locality they build in.

39:52 Say you’re worried that the next decade is going to be one of rapid AI implementation and poor labor market performance for many workers – what policies should you focus on then?

• Workers need policies that level the fundamental imbalances in power that afflict them when they try to bargain for higher wages against employers who are using every tool at their disposal – including technology and AI – to disempower workers
• From the University of California at Berkeley Labor Center, Negotiating Tech: An Inventory of U.S. Union Contract Provisions for the Digital Age

44:05 AI and the public sector: what policy can head off the harms of AI implementation?

• The US public sector has been cut extensively in recent decades
• The dysfunction associated with public services is not a sign of a powerful bureaucracy that is unresponsive to the public. It’s a hyper-responsive civil service utterly overwhelmed and falling back on hoping that process can substitute for active decision-making
• The public sector has no market test to drive hiring or service—it is entirely politics-driven. A choice.
• It’s never too early to be recession-planning

48:06 Say you’re wrong about the “normality” of AI as a technology and say that it does lead to unprecedented job-loss and devaluation of labor. What could we do to prepare for that?

• Whoever owns the AI or the businesses whose capital has become much more productive due to AI gets the benefits while everybody else loses
• Ownership has to be redistributed

51:07 What about the issue of AI exposure on the job? For instance, people having tools or software pushed on them, that may make their jobs less safe?

• Biggest solution is having a functional labor law where workers can organize and represent themselves at the workplace to ensure they are compensated appropriately for the difficult work they endure

54:10 There seems to be a lot of federal inaction on this. What is being done or can be done on the state level?

• Pre-emption is the biggest concern here.