OpenAI Chief Research Officer Mark Chen Warns AGI Window is Narrow: Models Shift Toward Autonomous Research, Evaluation Failure and Continual Learning Remain Bottlenecks

OpenAI Chief Research Officer Mark Chen Warns AGI Window is Narrow: Models Shift Toward Autonomous Research, Evaluation Failure and Continual Learning Remain Bottlenecks

N
News Editor
2026-06-30 19:01:42
OpenAI's Chief Research Officer Mark Chen has issued a significant statement on AGI timelines, asserting that the window for human intervention is 'very small'. He highlights that models are evolving toward autonomous research capabilities, defends the continued potential of scaling laws against the 'pretraining is dead' narrative, and predicts future researchers will adopt 'vibes-based research' focused on problem formulation and value judgment. At the same time, he warns that evaluation benchmarks are becoming obsolete and key bottlenecks like continual learning remain unresolved.
AGIOpenAIMark ChenScaling LawsPretrainingContinual LearningEvaluation Failure

AGI Countdown: A 'Very Small' Window for Humanity

OpenAI Chief Research Officer Mark Chen has delivered a stark warning on the arrival of Artificial General Intelligence (AGI), stating that the window for human adaptation is "very small." Chen points out that current models are transitioning from passive question-answering to autonomous research — attaining self-sustaining scientific capabilities where AI can independently design experiments, propose hypotheses, and iteratively evolve.

Chen emphasizes that Scaling Laws still hold significant potential, pushing back against the industry narrative that "pretraining is dead." He argues that greater compute and data continue to unlock emergent abilities. Looking ahead, Chen predicts that AI researchers will shift toward "vibes-based research" — moving away from parameter tuning to focus on problem formulation, value judgment, and high-level direction setting.

However, Chen also flags unresolved bottlenecks: existing evaluation frameworks are rapidly losing relevance, making it difficult to truthfully measure model performance on complex tasks; continual learning remains unachieved, as models still suffer catastrophic forgetting; and alignment and safety issues require sustained investment. These challenges imply that AGI won't arrive overnight, but the closing window demands urgent action.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
100

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.