TLDR: A recent seminar, ‘New Approaches to Characterize Industries: AI as a Framework and a Use Case,’ convened by the American Enterprise Institute, Stanford University’s Digital Economy Lab, and New York University, highlighted the critical need for new data and measurement frameworks to understand AI’s profound impact on industries, jobs, and skills. Traditional classification systems are proving inadequate, prompting calls for a ‘follow-the-people’ approach to track talent flows and the establishment of a new, independent institution to develop demand-driven tools and insights.
The rapid evolution of artificial intelligence, particularly generative AI systems, is fundamentally reshaping global economies and labor markets, yet traditional methods for classifying industries and measuring economic impact are falling short. This pressing issue was the central focus of a day-long seminar titled ‘New Approaches to Characterize Industries: AI as a Framework and a Use Case,’ held on March 18, 2024. Organized by the American Enterprise Institute, Stanford University’s Digital Economy Lab, and New York University, the event brought together leading economists, data scientists, and policy experts to explore innovative solutions.
The seminar underscored a critical deficiency in current data and evidence necessary to comprehend how AI is transforming business operations and the nature of work. Erik Brynjolfsson of the Stanford Digital Economy Lab, in his keynote, drew a parallel between the advent of generative AI and Anton van Leeuwenhoek’s microscope, suggesting that AI is at the nascent stage of a decades-long revolution. He emphasized that new digital tools will necessitate substantial investments in complementary intangible capital, offering an opportunity to better understand technology implementation within firms. This could lead to more granular skill and task taxonomies, surpassing the current O*NET system, and aid in identifying emerging ‘superstar’ AI firms.
Susan Athey, Chief Economist at the Federal Trade Commission, provided a complementary perspective, highlighting the practical challenges in data collection. She noted that even advanced Silicon Valley companies struggle with quantifying their processes and customer interactions, advocating for researchers to look beyond conventional survey tools to overcome mismeasurement.
Key Takeaways and Proposed Solutions:
Three primary conclusions emerged from the discussions:
1. Establish a New Institution: An independent, non-partisan institution is needed to develop ‘bottom-up, demand-driven tools and insights’ for businesses, workers, and governments. This entity would connect AI advancements with changes in jobs, skills, and economic opportunities.
2. Foster Partnerships: This new institution should build partnerships to understand AI’s impact on local and regional economies. Initial efforts would include providing data to firms transitioning AI from experimental to enterprise use, prototyping tools for workforce reskilling, and producing analyses for government agencies to allocate education and training resources.
3. Integrate with Existing Systems: The new framework and classifications must complement the Federal Statistical System (FSS), program providers, and scientific research, achieved through collaboration and focused initiatives.
The ‘Industries of Ideas’ Approach:
A recurring theme was the need to shift from document-based classifications to a ‘follow-the-people’ or ‘industries of ideas’ approach, pioneered by Julia Lane of New York University. This method involves tracing the career paths of grant-funded engineers and researchers from academia to industry. By linking administrative and wage records, it aims to map the flow of technological innovation and talent, providing a dynamic view of how ideas translate into economic impact. The IRIS UMETRICS dataset, which tracks research awards and personnel at over 100 U.S. universities, is a key tool in operationalizing this approach, with a pilot project already underway in Ohio.
Challenges in AI Measurement:
Defining AI: Identifying what constitutes AI and its varied applications across sectors remains a significant hurdle for economic accounts. David Wasshausen of the BEA compared this challenge to measuring the internet’s impact.
Data Granularity and Timeliness: Traditional data sources and classification systems are too slow and broad to capture the rapid, granular changes driven by AI. Diane Coyle emphasized that timely, even if less precise, data is often more valuable than delayed, highly accurate data.
Infrastructure and Talent: Prasanna ‘Sonny’ Tambe highlighted that AI investment is highly concentrated, with infrastructure (data, software, computing) often proprietary and rapidly evolving. There’s a global shortage of high-skilled AI talent, and tracking individual AI professionals is difficult due to a lack of reliable, up-to-date data.
Labor Market Impact: Significant uncertainty persists regarding AI’s effects on low-skilled workers and which skills will become obsolete or remain relevant. Harry J. Holzer of Georgetown University noted that while AI could displace jobs, it might also augment skills and create new roles, similar to past technological shifts.
Regulatory Landscape: Nestor Maslej of Stanford HAI pointed to a ‘tidal wave’ of new AI regulations, stressing the need for better measurement of government spending on AI R&D and more granular tracking of state and local AI policies.
Innovative Data Collection and Analysis:
Wage Records: Adam Leonard of the Texas Workforce Commission demonstrated how administrative wage records can provide deeper insights into labor market dynamics and firm-level AI investments through hiring trends. He cited a proof-of-concept using University of Texas at Austin data to track former employees entering AI-related industries.
Online Labor Market Data: Layla O’Kane of Lightcast discussed using online job postings and professional profiles to define AI jobs based on a ‘basket-of-skills’ approach. This real-time data can track demand for AI skills, salary premiums, and educational requirements, offering a flexible and dynamic measurement system.
AI-Powered Career Guidance: Lesley Hirsch of the New Jersey Department of Labor and Workforce Development showcased the NJ Career Navigator, an AI-powered tool that uses machine learning and natural language processing to provide personalized career, training, and job recommendations to jobseekers.
LinkedIn Data: Karin Kimbrough of LinkedIn shared data indicating a 70% increase in AI discussions and 22% growth in AI adoption in 2024. She categorized AI’s impact on tasks as augmented, disrupted, or insulated, noting that women and Gen Z workers are disproportionately exposed to AI-related changes. However, Josh Hawley cautioned that LinkedIn data may not always be reliable for comprehensive workforce trends.
Also Read:
- Over 19 Million U.S. Jobs Face High Automation Risk, SHRM Report Reveals
- AI Revolutionizes Construction and Engineering: Industry Leaders Convene to Chart Future Innovations
Conclusion:
The overwhelming consensus from the seminar is the urgent need for an authoritative, operational definition of AI to guide classification systems for engineers, companies, products, and skills. Without a collaborative effort between government, industry, and researchers to establish these standards, the economy risks ‘data and analytical chaos,’ hindering informed decision-making for education, training, and workforce development investments.
