The Turing Trap: AI, Automation, and Power
Explore how human-like AI risks a political trap where automation concentrates wealth and strips workers of bargaining power.
Read the analysisResearch area
Research on how AI changes productivity, work, investment, measurement, distribution, and economic adaptation
Explore how human-like AI risks a political trap where automation concentrates wealth and strips workers of bargaining power.
Read the analysisExplore how unmeasured intangibles cause two-sided productivity mismeasurement, forming a J-curve explaining recurring technological paradoxes.
Read the analysisExplore how online choice experiments measure consumer surplus from zero-price digital goods, revealing private welfare gains absent from GDP.
Read the analysisExplore how economists can prepare for transformative AI through nine grand challenges, new indicators, and scenario planning.
Read the analysisExplore how social pressure and peer adoption scenarios drive parental demand for educational AI, even when cognitive risks are known.
Read the analysisExplore how a GPT-3 assistant compresses productivity, accelerating novice learning and leveling the experience curve in customer support.
Read the analysisExplore how stated preference experiments reveal generative AI's massive consumer surplus, highlighting critical gaps in traditional GDP measurement.
Read the analysisExplore how ChatGPT access reshapes writing tasks, compressing performance inequality via effort substitution rather than skill complementarity.
Read the analysisExplore how automation displaces labor, new tasks reinstate it, and U.S. labor-demand growth shifted after 1987.
Read the analysis
Explore how AI compute concentration could reshape infrastructure, credit markets, regulation, fiscal systems, and economic power.
Read the analysisExplore how a task-based model constrains AI productivity forecasts to modest ten-year total factor productivity gains, highlighting GDP and welfare divergence.
Read the analysisExplore how GPT-4 access improved consulting output quality and speed on some tasks while reducing correctness on others in a field experiment.
Read the analysis