The one piece of data that could actually shed light on your job and AI
Why it matters: The lack of accurate data hinders any coherent plan for addressing AI's unprecedented impact on the global workforce.
- Anthropic CEO Dario Amodei believes AI could perform all human jobs in under five years, contributing to widespread panic among workers and calls for pausing data center construction.
- Economist Alex Imas (University of Chicago) criticizes current methods, including those used by OpenAI and Anthropic, for assessing AI job risk, stating that task "exposure" alone is insufficient.
- Imas advocates for a "Manhattan Project" to gather new, specific data on AI's actual impact on the workforce, highlighting the inadequacy of existing government task catalogs from 1998.
- OpenAI and Anthropic have used a 1998 US government catalog of job tasks to assess AI exposure, with Anthropic analyzing Claude conversations to see which tasks AI is actually completing.
While Silicon Valley leaders like Anthropic CEO Dario Amodei predict an AI-fueled jobs apocalypse within five years, economists like Alex Imas caution that current tools for predicting AI's impact are "abysmal." Imas argues that simply measuring a job's "exposure" to AI tasks, as done by OpenAI and Anthropic, is a "meaningless tool for predicting displacement" and calls for a "Manhattan Project" to collect new, more relevant data.
