Rudolph Angler Artificial general intelligence (AGI)—a still-theoretical AI capable of reasoning as well as or better than humans—remains an uncertain frontier. But top researchers believe we’re approaching a point where AGI may become a reality within the next few years.
Miles Brundage, formerly OpenAI’s head of policy research and AGI readiness, recently discussed the industry’s rapid advancements on the Hard Fork tech podcast. In the coming years, he suggests, we’ll see systems capable of performing essentially any computer-based task a human could do remotely, including operating a mouse and keyboard or even mimicking a “human in a video chat.” According to Brundage, this shift calls for governments to rethink sectors for taxation and the educational investments needed for future labor markets.
The timeline for achieving AGI is a topic of intense debate, but many influential voices in the field believe it’s within reach. John Schulman, OpenAI cofounder and former research scientist, and Dario Amodei, CEO of OpenAI competitor Anthropic, both estimate AGI could be here in just a few years, with Amodei projecting as early as 2026.
Brundage, who left OpenAI after six years, likely has one of the best insights into OpenAI’s AGI roadmap. However, he clarified that safety concerns did not prompt his departure. In his exit announcement on X, Brundage expressed a desire to make a broader impact in policy or nonprofit advocacy. On Hard Fork, he explained further: he left OpenAI partly because he wanted to work on cross-industry issues, like regulatory frameworks, rather than solely on internal projects. Another key motivation was a desire for independence, avoiding the potential dismissal of his views as merely corporate promotion.