What Happened

MIT Technology Review spoke with University of Chicago economist Alex Imas, who argues that current tools for predicting AI's impact on employment are fundamentally inadequate. The US government's O*NET task catalogue — first launched in 1998 — remains the primary dataset researchers use to map AI capabilities to job tasks, but it was never designed for this purpose. Imas is calling on economists to begin collecting granular, task-level labor data before displacement effects become impossible to untangle.

Why It Matters

Anthropic CEO Dario Amodei has publicly described AI as a potential "general labor substitute" capable of performing all jobs within five years. A researcher at the same company flagged a likely "breakdown of the early-career ladder." Without reliable task-level data, policymakers and businesses cannot distinguish which roles face near-term automation risk from those that are merely augmented. For indie developers and SMEs, this uncertainty makes workforce planning and product positioning genuinely difficult.

  • Early-career roles in writing, coding review, and data entry show the highest task-overlap with current LLM capabilities
  • No government has yet produced a coherent policy response to potential structural unemployment from AI
  • Economists who previously dismissed AI job risk are revising their positions

Asia-Pacific Angle

Chinese and Southeast Asian developers building productivity or HR-tech tools for global markets face a specific opportunity here. The O*NET catalogue covers US job definitions only — there is no equivalent structured dataset for job tasks in China, Vietnam, Indonesia, or the Philippines. Developers who instrument their own SaaS tools to collect anonymized task-completion data could build a proprietary signal that Western competitors lack. Alibaba's Tongyi and Baidu's ERNIE teams are already exploring workforce analytics; independent developers who move first on task-level data collection in ASEAN markets have a real first-mover window.

Action Item This Week

If you run any productivity, project management, or HR tool, add a lightweight task-logging schema this week — even just category, duration, and completion status per work item. Aggregated and anonymized, this is exactly the dataset economists say is missing, and it compounds in value over time.