La IA está acelerando y existe un peligro real - Connor Leahy

Item: La IA está acelerando y existe un peligro real - Connor Leahy
Rating: 4
Author: PostSingularity Institute

by Inteligencia Artificial (Jon Hernández)

Humanities, Social Sciences & Thought AI & Society UBDigital and Information technology: general topics UBJDigital and information technologies: social and ethical aspects

🎙 Inteligencia Artificial (Jon Hernández) 👥 726K 📅 February 26, 2026 ⏱ 131 min 👁 110K 📄 expert opinion 🧭 2026-08-06

Available in: English (current) Français

Keywords

AGIASIalignmentAI riskopen source

Summary

In this long-form interview, Connor Leahy, founder of EleutherAI and CEO of Conjecture, discusses the rapid acceleration of AI development and the existential risks it poses. He argues that we are heading towards superintelligence (ASI) without adequate control mechanisms, and that current AI systems are already causing societal harm. Leahy distinguishes between AI, AGI, and ASI, and criticizes the ‘AI effect’ where people continually redefine AI to downplay progress. He highlights the dangers of open-sourcing powerful AI models, the opacity of neural networks, and the lack of understanding among policymakers. He proposes that we must slow down and implement robust governance and regulation before it’s too late. The conversation also touches on the role of corporations like OpenAI and Meta, geopolitical competition, and the concept of P(Doom). Leahy emphasizes that the window for action is closing and that we need to take proactive measures to ensure a safe future.

149 words

Critical Evaluation

The interview provides a compelling and thought-provoking perspective on AI risks, grounded in Leahy’s extensive experience in the field. His arguments are logically structured and he effectively communicates complex concepts to a broad audience. However, the discussion is largely based on personal opinions and speculative projections rather than empirical evidence. While Leahy references specific examples like GPT-2 and Claude Code, he does not provide detailed technical explanations or cite specific studies to support his claims. The conversation is balanced, with the host challenging some points, but it remains one-sided in its focus on catastrophic risks. The lack of counterarguments from other experts limits the depth of the analysis. Nevertheless, the interview serves as a valuable introduction to AI safety concerns and encourages critical thinking about the trajectory of AI development. The production quality is high, and the host’s bilingual approach makes it accessible to a wider audience.

147 words

Title / Content Match

The title accurately reflects the core theme of the interview: the acceleration of AI and the associated dangers, as discussed by Connor Leahy.

Quality & Reliability

8/10

The video features Connor Leahy, a recognized AI safety researcher and founder of EleutherAI, providing expert opinions based on his experience. The discussion covers technical aspects of AI development and risks, but relies heavily on personal assessments and projections rather than peer-reviewed data. The production quality is high, and the host facilitates a balanced conversation, though the content is inherently subjective and forward-looking.

Chapters

Cited Sources

Club de la IA — Community and AI news group mentioned in the description.
Hostinger — Sponsor link for web hosting services, mentioned in the description.
IA Perfecta — Website with best AIs, mentioned in the description.
InfoJobs offers — Sponsor link for job offers, mentioned in the description.
Jace AI — AI email assistant, mentioned in the description.
Guide to use ChatGPT — Guide mentioned in the description.
Newsletter — Weekly AI news newsletter, mentioned in the description.
WhatsApp channel — WhatsApp channel about AI, mentioned in the description.

Concurring Sources

AI alignment — Supports the discussion on alignment challenges.
Superintelligence — Provides background on the concept of superintelligence.

Dissenting Sources

OpenAI's approach to AI safety

Contribution & Novelties

The interview provides a comprehensive overview of AI safety concerns from a prominent figure in the field, offering a clear framework for understanding the distinctions between AI, AGI, and ASI. It emphasizes the urgency of addressing alignment and control issues before superintelligence is achieved. The discussion on open source risks and the ‘AI effect’ adds nuance to the public debate.

Pour aller plus loin :

AI alignment — Overview of the alignment problem in AI.
Superintelligence — Definition and discussion of superintelligent systems.
EleutherAI — Background on the organization founded by Connor Leahy.
P(Doom) — Discussion of the probability of AI catastrophe, related to the concept mentioned in the interview.

110 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, reflecting the depth of the discussion. The technical level is moderate, suitable for a general audience. Reliability is high due to the expert guest, though the subjective nature of the content limits the overall score.

Reliability 8/10

💬 Positive: The comments are overwhelmingly positive, with many viewers praising the interview as one of the best on AI, appreciating the depth of the discussion and Connor Leahy's insights. Some express concern about the future but are grateful for the information.