Inside the minds building the first superintelligence

The CEOs of OpenAI, Google DeepMind, and Anthropic have all predicted that AGI will arrive within the next 5 years. In September 2024, Sam Altman stated we could have superintelligence in just a few thousand days. What exactly do we mean by superintelligence? How could we get there so quickly? And what might a world with superintelligence look like? Sana invited two leading minds to unpack these questions at Sana AI Summit 2025. What’s in this video The distinction between artificial intelligence and superintelligence Predictions on the timeline to achieving AGI and superintelligence The role of AI in scientific advancements, like nuclear fusion Philosophical musings on AI's ability to generate new knowledge The ethical and social implications of AI alignment and safety Global perspectives on AI development and geopolitical considerations Who you're hearing from Boaz Barak is a professor of Computer Science at Harvard, where he co-directs the university’s undergraduate Computer Science program. While originally a theoretical computer scientist, Boaz’s focus has increasingly turned to the safety of artificial intelligence systems. At the time of the Summit, he was on leave from Harvard to work on safety and alignment at OpenAI. David Pfau is a staff research scientist at Google DeepMind and visiting professor at Imperial College London’s Department of Physics. David’s work focuses on applying AI to scientific domains such as nuclear fusion and quantum mechanics. Moderating this discussion is Joel Hellermark, founder and CEO of Sana, an AI company which has raised over $130 million from NEA, EQT Ventures, and Menlo Ventures to date and been listed by Forbes as one of the top 50 AI companies in the world. An enterprising child, Joel taught himself programming in C at age 13 and founded his first company, a video recommendation engine, at 16. This fireside was recorded live at Sana AI Summit 2025, Stockholm, May 7th, 2025. Subscribe for more insights on AI and the future of technology. -- 00:42 - Starting with a lightning round 00:57 - Round 1: Tom Cruise or ChatGPT running the world? 01:37 - Round 2: Who should define the morals of AI systems? 02:42 - Round 3: Short one stock, long one stock? 03:19 - Round 4: Elon, Zuck, Demis, or Sam controlling AGI? 04:15 - Applying AI to accelerate nuclear fusion 06:25 - Why isn't AI better at creating new knowledge? 08:05 - The limits of next token prediction on knowledge generation 09:29 - Getting to a fully self-improving AI system 11:25 - Do we need to understand the models to ensure safety? 12:52 - Where we could go wrong with AI 13:50 - Geopolitical considerations of superintelligence 16:32 - Why we might not ever see true superintelligence