AI Progress 'Very Fast,' Says Miles Brundage, Dismissing Slowdown Reports

Miles Brundage, a prominent AI policy researcher and former OpenAI Senior Advisor for AGI Readiness, recently asserted that "Reports of the slowing of AI progress are greatly exaggerated." His statement, shared on social media, directly challenges growing narratives suggesting a deceleration in artificial intelligence development. Brundage emphasized his perspective that AI advancement is currently "very fast," drawing on his extensive experience within the field.

Brundage elaborated on his Substack, noting that while perceptions of AI speed can vary based on metrics like GDP or revenue, a deeper look at task time horizons, evaluations, and insights from within leading AI companies reveals a different picture. He highlighted the rapid saturation of new AI benchmarks and the emergence of systems that outperform human experts on graduate-level tasks, such as Google-Proof Question Answering (GPQA), as evidence of rapid advancement. These breakthroughs, he explained, are often driven by innovations like "chain of thought" reasoning in AI models.

With a career spanning significant roles at OpenAI from 2018 to 2024, including Head of Policy Research and Senior Advisor for AGI Readiness, Brundage brings considerable authority to the discussion. His recent departure from OpenAI, coinciding with the dissolution of the AGI Readiness team, positions him as an independent voice on AI policy and development. This allows him to openly address critical industry observations, including the pace of technological evolution.

The perception of a slowdown sometimes arises from specific product launches, such as the initial reception of GPT-5. However, Brundage has previously indicated that such assessments might be based on a "false belief that it is based on a much larger base model," suggesting that not all product iterations represent fundamental scaling leaps. He argues that while some metrics might appear stagnant, underlying algorithmic and capability advancements continue at a significant pace.

Despite his conviction in rapid AI progress, Brundage has consistently advocated for societal preparedness and the need for "brakes" on AI development to ensure responsible integration. He maintains that neither AI labs nor the world are fully ready for advanced artificial general intelligence, underscoring the critical need for policy research and governance. His ongoing work aims to foster a shared understanding of AI's upsides and downsides to align collective interests and ensure safety.