Artificial intelligence is rapidly transforming the world, raising urgent questions about its impact, governance, and the future of human-machine collaboration. As AI systems become more capable, society faces challenges around safety and the balance of power. What does it mean to build and deploy technology that can reason, create, and potentially automate research itself? How do leading researchers navigate the technical and ethical frontiers of this new era?Jakub Pachocki, Chief Scientist at OpenAI, and Szymon Sidor, Technical Fellow at OpenAI, share their journeys from early programming competitions in Poland to shaping some of the most advanced AI systems in the world. They discuss the evolution of AI research, the technical and emotional challenges of building breakthrough models, and the profound societal questions that come with unprecedented progress.2:42 - Origin story: high school to OpenAI6:31 - “AI enlightenment” and AlphaGo moment17:12 - Early OpenAI culture and impostor syndrome23:30 - Power duo dynamic and collaboration27:25 - Shift to reasoning models36:23 - Possibilities of AGI42:12 - OpenAI’s pandemic efforts showed AI’s immaturity51:15 - Governance lessons from crisis55:39 - AI safety and optimism for the future
--------
54:33
--------
54:33
Jonathan Zittrain: AI Agents and Trust
AI systems are beginning to act on our behalf — executing tasks, making decisions, and influencing daily life in ways that often escape direct human oversight. What are the consequences of delegating decision-making to systems we don’t fully understand? How do we design AI to enhance human capabilities without sidelining human judgment?In this episode, host Aleksander Mądry welcomes Jonathan Zittrain, Professor at Harvard Law School and Co-Founder of the Berkman Klein Center for Internet & Society. They explore the legal, ethical, and societal challenges of living alongside increasingly autonomous AI. Drawing on lessons from the internet’s evolution, Zittrain examines how we can structure responsibility, foster innovation, and build systems that truly serve the public good — all while navigating the profound opportunities and risks of this technological transformation.09:20 - Three eras of internet evolution14:56 - Generative tech and equilibrium challenges20:59 - Generativity vs. closed systems26:36 - The promise and pitfalls of AI assistants32:56 - Privacy, privilege, and AI user rights39:01 - Regulation, self-regulation, and global policy48:28 - Defining AI agents and their real-world impact52:26 - AI’s potential to empower the marginalized1:15:41 - Acceleration, agency, and the future
--------
1:21:12
--------
1:21:12
Reid Hoffman: AI and Superagency
As AI becomes more integrated into daily life, it raises urgent questions about trust, accountability, and human agency. What happens when people over-rely on AI or when helpful tools are held back by perfectionism? Can we deploy imperfect systems while still earning public confidence?Reid Hoffman, Co-Author of “Superagency: What Could Possibly Go Right with Our AI Future,” joins this episode to explore how AI can enhance, not replace, human judgment. He discusses democratic leadership, global competition, and the design choices that shape user agency. Reid also shares his perspective on long-term AI investment beyond short-term hype.03:00 - Superagency and human-centered design17:27 - Risk, regulation, and collective action21:42 - Rethinking agency in the age of agents31:18 - Trust, regulation, and system design55:09 - Building trust through real-world AI benefits1:06:07 - China’s ambition and multipolar competition1:09:00 - Investing in applied AI with real impact1:14:00 - Reid’s hope for AI engagement
--------
1:13:52
--------
1:13:52
Sendhil Mullainathan: AI and Algorithmic Bias
As AI continues to permeate various aspects of society, its impact on decision-making, bias, and future technological developments is complex. How can we navigate the challenges posed by AI, particularly when it comes to fairness and bias in algorithms? What insights can be drawn from the intersection of economics, computer science, and behavioral studies to guide the responsible development and use of AI?In this episode, Sendhil Mullainathan, a prominent economist and professor, delves into these pressing issues. He shares his journey from computer science to behavioral economics and discusses the role of AI in shaping the future of decision-making and societal structures. Sendhil provides a nuanced view of algorithmic bias, its origins, and the challenges in mitigating it. He also explores the potential and pitfalls of AI in healthcare and policymaking, offering insights into how we can harness AI for the greater good while being mindful of its limitations.0:00 - Start1:51 - Introducing Sendhil14:20 - Algorithmic bias29:20 - Handling Bias41:57 - AI and Decision Making57:01 - AI in our Future1:02:29 - Conclusion and the last question
--------
1:03:48
--------
1:03:48
Luis Videgaray: AI Policy Around the World
AI technology is transforming the world, raising crucial questions about governance, regulation, and societal impact. What measures should governments adopt to regulate AI effectively? How can academic institutions contribute to this development? Luis Videgaray is a professor at MIT and the director of the MIT AI Policy for the World Project. Previously, he also served as the Minister of Finance and the Minister of Foreign Affairs in the Mexican Government. In this episode, Luis shares his journey and insights on the role of governments in regulating and utilizing AI. He also covers the evolution of AI policy, the geopolitical landscape, and the future of AI regulation. 3:47 - Luis’ time in the government 7:12 - AI in the past 20:19 - Regulation of AI 43:13 - Governments using AI 51:03 - Geopolitics of AI 1:11:31 - Luis at MIT 1:16:50 - Luis’ perspectives on AI 1:25:28 - Conclusion and the last question
Artificial General Intelligence — the type of AI that reaches or even surpasses human capabilities — is an exciting topic. But what is not explored as much is an equally important question: what happens before AGI is here? That is, what should we be doing to prepare?
In Before AGI, I hope to have honest conversations about what are the goals, what are the problems, and what lies ahead as we develop AI.
España