Google Just Revealed What Comes After AGI And It’s Shocking

Value of the Information & Strength of the Argument

The video provides a comprehensive overview of the DeepMind paper, accurately conveying its key concepts and definitions. It effectively explains the four pathways to ASI and the six frictions, using clear examples and analogies. However, the argumentation is largely descriptive rather than critical, and the video tends to present speculative scenarios as plausible without thoroughly examining counterarguments or alternative viewpoints. The value lies in its clear synthesis of complex ideas, but it lacks depth in evaluating the feasibility and implications of the proposed pathways.

Scientific Rigor, Source Quality, Title Accuracy

The video cites the primary source (the DeepMind paper) and provides links to related publications, which is commendable. However, it does not critically assess the paper’s methodology or potential biases. The title is somewhat sensationalist but aligns with the content’s focus on the post-AGI transition. The video’s presentation is engaging but could benefit from more rigorous analysis and discussion of uncertainties.

The title is somewhat sensationalist ('shocking') but accurately reflects the content's focus on the post-AGI transition.

7/10

The video provides a detailed and faithful summary of a specific DeepMind paper, with references to primary sources. However, it lacks critical analysis and presents speculative scenarios as plausible without sufficient nuance.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:30Introduction to the DeepMind paper 'From AGI to ASI' and its authors.
• 02:00Definition of AGI as median human-level performance.
• 03:00Definition of ASI and the concept of universal AI (AXI).
• 05:00Pathway 1: Scaling and the thought experiment of 100 million AGIs.
• 07:00Pathway 2: Algorithmic paradigm shifts.
• 09:00Pathway 3: Recursive self-improvement.
• 11:00Pathway 4: Multi-agent collectives.
• 13:00Discussion of frictions and limitations of ASI.

• From AGI to ASI (DeepMind paper) — The main paper discussed in the video.
• From AGI to ASI (arXiv version) — Full technical paper on arXiv.
• DeepMind's framework for measuring AGI progress — Earlier framework for measuring AGI progress.
• Google's 2026 framework for tracking AGI — Google's framework for tracking progress toward AGI.
• DeepMind's approach to AGI safety — DeepMind's approach to AGI safety and security.
• Demis Hassabis on AI agents and the road to AGI — Interview with Demis Hassabis.
• Legg and Hutter paper on machine intelligence — Foundational paper on formal machine intelligence.

• DeepMind's AGI safety blog — Aligns with the paper's discussion of safety and responsible development.
• Legg and Hutter paper — Provides theoretical foundation for measuring intelligence, consistent with the paper's definitions.

• Comment on model collapse — A commenter points out that training on synthetic data can lead to model collapse, which the video mentions but does not deeply explore.
• Comment on energy constraints — Another commenter argues that energy and hardware constraints are more fundamental than the video suggests, potentially slowing progress more than anticipated.