Into the Bytecode #52 – Michael Nielsen on being a wise optimist about science and technology
Mar 27, 2025
Michael Nielsen, a scientist and research fellow at the Astera Institute, shares his insights on maintaining optimism amid existential risks in science and technology. He discusses the concept of asymmetric leverage and questions the feasibility of unbiased models in AI. The conversation highlights the evolving role of AI in scientific methods and the ethical implications of military applications. Nielsen also explores the intricacies of human-AI interaction and the importance of crafting a moral framework for a future with advanced AI.
AI Snips
Chapters
Books
Transcript
Episode notes
Human-AI Interaction
- Interfacing with large language models (LLMs) feels strange because they possess vast knowledge, while we slowly process information.
- This interaction mirrors the potential future relationship between humans and ASI, where AI complexity far surpasses human comprehension.
Improving AI Interfaces
- Focus on improving interfaces with existing AI models.
- Better interfaces can unlock more of the models' potential without requiring changes to the models themselves.
Agentic AI
- We will likely have highly agentic AI systems due to economic incentives and research interests.
- These systems, similar to algorithms in finance, will have increasing autonomy to act independently, even in limited ways.
