Google quiere cambiar la IA para siempre (Nested Learning)

Item: Google quiere cambiar la IA para siempre (Nested Learning)
Rating: 4
Author: PostSingularity Institute

by EDteam

Formal & Physical Sciences AI — Models & Research UYQArtificial intelligence (AI)UYQNNeural networks and fuzzy systems

🎙 EDteam 👥 1.0M 📅 December 2, 2025 ⏱ 41 min 👁 54K 📄 science communication 🧭 2026-08-02

Available in: English (current) Français

Keywords

deep learningnested learningamnesiacatastrophic forgettingAGI

Summary

The video analyzes Google’s Nested Learning paper, which claims that deep learning is an illusion and that current models suffer from amnesia. It explains the fundamentals of deep learning, including layers, loss functions, gradients, and backpropagation, using analogies like a pizza chef and a reggaeton singer. The video discusses the vanishing gradient problem and the history of AI milestones from 1986 to Transformers. It then introduces Nested Learning as a new architecture that mimics the human brain, with fast and slow layers, allowing continuous learning without forgetting. The presenter evaluates the paper’s credibility, noting it has been peer-reviewed, and speculates on its potential to lead to AGI. The video also mentions RAG and context windows as partial solutions to catastrophic forgetting. Overall, it provides a comprehensive overview for a general audience, but lacks direct references to the paper itself.

139 words

Critical Evaluation

The video excels in making complex AI concepts accessible through effective analogies and a clear narrative structure. The explanation of deep learning, including loss functions and gradients, is accurate and well-illustrated. The presenter correctly identifies the vanishing gradient problem and the historical context of AI developments. However, the video’s main weakness is its reliance on the Nested Learning paper without providing direct citations or links to the original research. While the paper is peer-reviewed, the video does not offer evidence or external sources to support its claims, which limits its scientific rigor. The speculative discussion about AGI is engaging but not grounded in empirical evidence. The video also includes promotional segments for EDteam courses, which are clearly separated but may distract from the content. The title accurately reflects the content, and the video does not mislead viewers. Overall, the video is informative and well-produced, but its lack of direct references and speculative tone prevent it from being a fully rigorous scientific analysis.

162 words

Title / Content Match

The title accurately reflects the content, which focuses on Google's Nested Learning paper and its potential impact on AI.

Quality & Reliability

7/10

The video provides a clear and accessible explanation of deep learning and the Nested Learning paper, with accurate technical descriptions. However, it lacks direct citations to the paper or other sources, and some claims are presented without verification. The presenter's enthusiasm and speculative tone slightly reduce the overall reliability.

Chapters

Cited Sources

EDteam - Claude Code course — Mentioned as a course offering, not directly related to the paper.
EDteam - Full Stack with AI course — Mentioned as a course offering, not directly related to the paper.
EDteam - AI for everyone course — Mentioned as a course offering, not directly related to the paper.
EDteam - RAG course — Mentioned as a course offering, related to RAG as a solution discussed in the video.
EDteam - Free courses — Promotional link for free courses.
EDteam - All courses — Promotional link for all courses.
EDteam - Student scholarships — Promotional link for student scholarships.
EDteam - Instagram — Social media link.
EDteam - LinkedIn — Social media link.
EDteam - Premium — Promotional link for premium membership.
EDteam - Teachers — Promotional link for becoming a teacher.
EDteam - TikTok — Social media link.

Concurring Sources

Attention Is All You Need — The video references this paper as the foundation of modern AI, and it is consistent with the discussion of Transformers.

Dissenting Sources

On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?

Contribution & Novelties

The video provides a clear and engaging explanation of the Nested Learning paper, making complex concepts accessible to a broad audience. It highlights the paper’s claim that deep learning is an illusion and proposes a new architecture for continuous learning. The video’s originality lies in its pedagogical approach, using analogies to explain technical details. However, it does not offer new scientific insights beyond summarizing the paper.

Pour aller plus loin :

Nested Learning: The Illusion of Deep Learning — Note: This is a placeholder; the actual paper may be found on arXiv. The video does not provide the exact URL.
Catastrophic Interference in Connectionist Networks — Note: This is a classic paper on catastrophic forgetting, relevant to the video’s discussion.
Attention Is All You Need — Note: The original Transformer paper, mentioned in the video as a milestone.
RAG: Retrieval-Augmented Generation — Note: The paper introducing RAG, discussed as a partial solution.

152 words

Radar Profile

The radar profile shows high scores in quantity of information and technical level, but lower scores in reliability and quality of information, reflecting the video's strong explanatory content but lack of direct citations and speculative elements.

Reliability 6/10

💬 Très positif. Sur les 30 commentaires analysés, la grande majorité exprime une forte appréciation de la clarté et de la pédagogie de l'explication, avec quelques commentaires critiques ou spéculatifs sur le contenu.