La Superinteligencia nos extinguirá y no hay forma de detenerla 🤖 | 🎙️ Roman Yampolskiy

La Superinteligencia nos extinguirá y no hay forma de detenerla 🤖 | 🎙️ Roman Yampolskiy

by Jon Hernández

Formal & Physical Sciences AI & Society UBDigital and Information technology: general topicsUBJDigital and information technologies: social and ethical aspects
🎙 Jon Hernández 👥 725K 📅 December 11, 2025 ⏱ 93 min 👁 108K 📄 expert opinion 🧭 2026-08-03
Available in: English (current) Français

Keywords

AI safetyexistential risksuperintelligencecontrol problemAGI

Summary

In this podcast episode, host Jon Hernández interviews Roman Yampolskiy, a computer scientist and AI safety researcher known for coining the term ‘AI safety.’ Yampolskiy argues that superintelligent AI poses an existential threat to humanity, with a 99% probability of catastrophic outcomes. He explains the difference between tools and agents, emphasizing that AI agents can make independent decisions, making them unpredictable and uncontrollable. He discusses the rapid progress in AI capabilities, the lack of progress in control methods, and the impossibility of predicting or understanding a superintelligent entity. He also touches on the urgency of addressing existential risk over other issues like job displacement, the black box problem, and the potential for accidental creation of superintelligence. The conversation covers the need for global coordination to halt development, the role of governments, and the ethical implications. Yampolskiy remains pessimistic about our ability to prevent this outcome, suggesting that the only solution is to not build AGI at all.

157 words

Critical Evaluation

The interview provides a compelling and articulate presentation of the existential risks associated with superintelligent AI. Yampolskiy’s arguments are logically structured, drawing on his extensive research in AI safety. He effectively uses analogies (e.g., tools vs. agents, humans vs. ants) to make complex concepts accessible. The discussion is rigorous in its exploration of the control problem, highlighting the theoretical limits of our ability to align or constrain superintelligent systems. However, the content is largely based on speculative reasoning and personal probability estimates rather than empirical evidence, which is inherent to the topic. The host does a good job of guiding the conversation and asking pertinent questions, but the interview remains one-sided, with little critical pushback on Yampolskiy’s more extreme claims. The sources cited are not explicitly referenced during the conversation, but the guest’s credibility and prior publications lend weight to his statements. The title accurately reflects the content, and the overall production quality is high. The video does not include any sponsored segments, despite the presence of promotional links in the description. The public comments are overwhelmingly positive, praising the clarity and depth of the discussion, though some express concern about the pessimistic outlook. Overall, this is a valuable contribution to the public discourse on AI safety, though it should be viewed as an expert opinion rather than a balanced scientific review.

222 words

Title / Content Match

The title accurately reflects the core theme of the interview: the existential threat of superintelligence and the perceived impossibility of control.

Quality & Reliability

8/10

The video features a recognized expert in AI safety, Roman Yampolskiy, who provides a well-argued perspective on existential risks. The discussion is grounded in his research and publications, though it is primarily an opinion-based interview rather than a peer-reviewed presentation. The host adds context but the content is largely anecdotal and theoretical.

Chapters

Cited Sources

Concurring Sources

Dissenting Sources

Contribution & Novelties

The interview provides a comprehensive and accessible overview of the existential risks posed by superintelligent AI, articulated by a leading expert in the field. Yampolskiy’s perspective, while pessimistic, is grounded in his research on the limits of AI control, offering a unique angle that is often underrepresented in mainstream discussions. The conversation clarifies key concepts such as the difference between tools and agents, the unpredictability of superintelligence, and the inadequacy of current safety measures. It also addresses the socio-political dynamics that hinder effective regulation, such as the competitive race among AI labs.

Pour aller plus loin :

161 words

Radar Profile

The radar profile shows high scores in information quantity, quality, and reliability, with a slightly lower technical level, reflecting the accessible yet expert-driven nature of the content. The overall balance indicates a well-produced, informative interview that is both rigorous and engaging.

Reliability 8/10

💬 Très positif : Sur les 30 commentaires analysés, l'écrasante majorité exprime une admiration pour la qualité de l'entrevue et la clarté des explications, certains la qualifiant de meilleure vidéo du canal, avec quelques réserves sur le pessimisme du propos.