Comprehensive Summary

The Threshold 2030 conference brought together 30 leading economists, AI policy experts, and professional forecasters to evaluate the potential economic impacts of artificial intelligence by the year 2030. Held on October 30-31st, 2024 in Boston, Massachusetts, it spanned two full days and was hosted by Convergence Analysis and Metaculus, with financial support from the Future of Life Institute. 

Participants included representatives from the following organizations: Google, OpenPhil, OpenAI, the UN, MIT, DeepMind, Stanford, OECD, Partnership on AI, Metaculus, FLI, CARMA, SGH Warsaw School of Economics, Convergence Analysis, ICFG, AOI, and FHI.

During the conference, attendees were given three different scenarios of AI capabilities advancement by 2030, spanning modest improvements on today’s AI systems to powerful, general AI agents that outperform humans at all cognitive labor. Attendees engaged in a series of exercises to explore these scenarios, conducting in-depth worldbuilding, economic causal modeling, forecasting exercises, and extensive debates. 

  • By bringing together experts to rapidly evaluate potential economic impacts through structured scenario modeling, Threshold 2030 aimed to:

  • Develop a clearer understanding of how economists view outcomes under extremely rapid AI advancement scenarios.

  • Create better frameworks & metrics to measure AI's economic impacts.

  • Generate concrete research questions to address uncertainties around economic impacts & policies for a post-AI economy.

  • Build stronger connections & consensus between AI policy experts and leading economists.

Three Scenarios of AI Capabilities in 2030

At the outset of the conference, attendees were presented with scenarios describing three potential levels of frontier AI capabilities in 2030. Attendees considered many conference exercises in the context of these three scenarios, withholding their prior expectations around AI capabilities development.

Scenario 1: Current AI systems, but with improved capabilities

Overall, AI systems in 2030 are more powerful versions of today’s LLMs, but with similar structural limitations. LLMs still primarily function in response to direction from humans, and do not take the initiative or act independently.

Scenario 2: Powerful, narrow AI systems that outperform humans on 95% of well-scoped tasks

AI systems achieve better results than people in most constrained or well-scoped tasks. However, they fail to outperform humans in task integration, handling multifaceted responsibilities, and communication with other humans. They still require oversight.

Scenario 3: Powerful, general AI systems that outperform humans on all forms of cognitive labor

Powerful AI systems can meet and surpass the performance of humans in all dimensions of cognitive labor, and can function as “drop-in” replacements for nearly all human jobs.

The rest of the report is divided into three sections, corresponding to the three overarching exercises the participants performed: (1) Worldbuilding, (2) Economic Causal Models, and (3) Forecasting.