Google DeepMind Director Calls for Manhattan Project Level Data Collection on Economy
"We don't have any data. I've been kind of saying we need a Manhattan Project for data. We don't have data on basically consumer demand elasticities. We don't know what they are. We're not really tracking what jobs are getting created or destroyed, the O*NET database with all of the tasks and different jobs that's been rarely updated. It's super low quality."
About this episode
In this episode, Dwarkesh Patel interviews Alex Imas, Director of AGI Economics at Google DeepMind and Professor of Economics at University of Chicago, alongside Phil Trammell, Head of Economics at Epoch and research scholar at Stanford. The conversation centers on what economic theory predicts about automation, wages, and wealth distribution in an AGI-dominated world. Imas challenges common predictions of labor market collapse, noting that labor share has remained remarkably stable at over 60% of GDP despite centuries of automation, and argues this could continue if demand patterns and capital variety expansion prevent satiation. He reveals critical data gaps in tracking consumer elasticities and job transformations, calling for a Manhattan Project level effort to collect economic data on AI's impact. Surprisingly, current evidence shows no significant white-collar job losses from AI, with even software engineering showing continued growth. The discussion explores whether a 'relational sector' where human involvement is intrinsically valued could sustain employment, or whether evolutionary selection for wealth-maximizing agents like Elon Musk will drive labor share toward zero through compound capital accumulation. On redistribution, they debate the feasibility of universal basic capital versus negative income tax, noting the political economy risks of government-dependent populations. For developing countries, they recommend indexing AGI supply chains through sovereign wealth funds rather than retraining programs, given AI's rapid advancement. The conversation concludes with concerns about concentration versus commoditization, noting that widespread AI access may be necessary both for broad prosperity and to prevent dangerous government control over a few powerful labs.
Key takeaways
- Labor share has held constant at over 60% through centuries of automation and could remain high through AGI depending on demand elasticities and capital variety expansion.
- Imas calls for Manhattan Project level data collection on consumer demand elasticities and job transitions, noting current economic data is insufficient for AGI forecasting.
- Yale Budget Lab analysis finds no evidence of white-collar job losses from current AI, even in exposed sectors like software engineering.
- Phil Trammell argues evolutionary selection for unsatiating capital accumulators like Musk could drive labor share to zero through compound wealth growth regardless of technology.
- For developing countries excluded from AI production, economists recommend sovereign wealth investment in AGI supply chains over retraining programs given speed of change.
- Redistribution mechanisms like universal basic capital face indexing challenges while UBI creates dangerous political dependencies on whoever controls payments.
- Whether AI becomes commoditized like electricity or concentrated like social media determines both broad prosperity prospects and feasibility of capturing returns through index funds.