AI Researchers' Views on Automating AI R&D and Intelligence Explosions

TL;DR

An epistemic divide emerged between frontier lab researchers and academic researchers, the latter of which expressed more skepticism about explosive growth scenarios, and 17/25 participants expected AI systems with advanced coding or R\&D capabilities to be increasingly reserved for internal use at AI companies or governments, unseen by the public.

Abstract

Many leading AI researchers expect AI development to exceed the transformative impact of all previous technological revolutions. This belief is based on the idea that AI will be able to automate the process of AI research itself, leading to a positive feedback loop. In August and September of 2025, we interviewed 25 leading researchers from frontier AI labs and academia, including participants from Google DeepMind, OpenAI, Anthropic, Meta, UC Berkeley, Princeton, and Stanford to understand researcher perspectives on these scenarios. Though AI systems have not yet been able to recursively improve, 20 of the 25 researchers interviewed identified automating AI research as one of the most severe and urgent AI risks. Participants converged on predictions that AI agents will become more capable at coding, math and eventually AI development, gradually transitioning from `assistants' or `tools' to `autonomous AI developers,' after which point, predictions diverge. While researchers agreed upon the possibility of recursive improvement, they disagreed on basic questions of timelines or appropriate governance mechanisms. For example, an epistemic divide emerged between frontier lab researchers and academic researchers, the latter of which expressed more skepticism about explosive growth scenarios. Additionally, 17/25 participants expected AI systems with advanced coding or R\&D capabilities to be increasingly reserved for internal use at AI companies or governments, unseen by the public. Participants were split as to whether setting regulatory ``red lines" was a good idea, though almost all favored transparency-based mitigations.

Figures (4)

• Figure 1: Participant predictions on whether AI companies will publicly deploy models capable of meaningfully accelerating AI research, grouped by affiliation type. The figure shows a categorical dot plot with participants grouped into "Expects Internal," "Nuanced," and "Expects Public" categories. Each dot is a participant; a full participant table can be found in Table \ref{['tab:participants']}. Participants with an unclear opinion are omitted from the figure. Participants who expect public deployment cite economic pressures, funding needs, competitive dynamics, and responsible decision-making. Those expecting internal deployment believe ASARA is too valuable to share and offers higher returns when used for internal R&D rather than API deployment. The "Nuanced" category includes participants who believe deployment decisions depend on government intervention and inter-lab differences in policies and culture.
• Figure 2: Participant views on the clarity of the trajectory toward ASARA, grouped by affiliation. The figure shows participants distributed along a spectrum from "Clear Path" through "Major Obstacles" to "Unknown/Unknowns." Frontier lab researchers predominantly cluster toward viewing the path as clear, while academic researchers are more distributed across the spectrum, with several identifying major obstacles. Ex-frontier lab researchers show mixed views, and nonprofit/industry participants fall primarily in the middle range (major obstacles; no one fell into the final category 'unknown unknown' using the method in Appendix \ref{['app:ai_prompts']}).
• Figure 3: Participant risk perception levels for ASARA, grouped by affiliation type.
• Figure 4: Participant views on red lines for autonomous research grouped by affiliation type. Transcripts that did not express a clear perspective were not included.