Episode #449: The Strange Loop: How Biology and Computation Shape Each Other
Episode description
In this episode of Crazy Wisdom, Stewart Alsop speaks with German Jurado about the strange loop between computation and biology, the emergence of reasoning in AI models, and what it means to "stand on the shoulders" of evolutionary systems. They talk about CRISPR not just as a gene-editing tool, but as a memory architecture encoded in bacterial immunity; they question whether LLMs are reasoning or just mimicking it; and they explore how scientists navigate the unknown with a kind of embodied intuition. For more about German’s work, you can connect with him through email at [email protected].
Check out this GPT we trained on the conversation!
Timestamps
00:00 - Stewart introduces German Jurado and opens with a reflection on how biology intersects with multiple disciplines—physics, chemistry, computation.
05:00 - They explore the nature of life’s interaction with matter, touching on how biology is about the interface between organic systems and the material world.
10:00 - German explains how bioinformatics emerged to handle the complexity of modern biology, especially in genomics, and how it spans structural biology, systems biology, and more.
15:00 - Introduction of AI into the scientific process—how models are being used in drug discovery and to represent biological processes with increasing fidelity.
20:00 - Stewart and German talk about using LLMs like GPT to read and interpret dense scientific literature, changing the pace and style of research.
25:00 - The conversation turns to societal implications—how these tools might influence institutions, and the decentralization of expertise.
30:00 - Competitive dynamics between AI labs, the scaling of context windows, and speculation on where the frontier is heading.
35:00 - Stewart reflects on English as the dominant language of science and the implications for access and translation of knowledge.
40:00 - Historical thread: they discuss the Republic of Letters, how the structure of knowledge-sharing has evolved, and what AI might do to that structure.
45:00 - Wrap-up thoughts on reasoning, intuition, and the idea of scientists as co-evolving participants in both natural and artificial systems.
50:00 - Final reflections and thank-yous, German shares where to find more of his thinking, and Stewart closes the loop on the conversation.
Key Insights
- CRISPR as a memory system – Rather than viewing CRISPR solely as a gene-editing tool, German Jurado frames it as a memory architecture—an evolved mechanism through which bacteria store fragments of viral DNA as a kind of immune memory. This perspective shifts CRISPR into a broader conceptual space, where memory is not just cognitive but deeply biological.
- AI models as pattern recognizers, not yet reasoners – While large language models can mimic reasoning impressively, Jurado suggests they primarily excel at statistical pattern matching. The distinction between reasoning and simulation becomes central, raising the question: are these systems truly thinking, or just very good at appearing to?
- The loop between computation and biology – One of the core themes is the strange feedback loop where biology inspires computational models (like neural networks), and those models in turn are used to probe and understand biological systems. It's a recursive relationship that’s accelerating scientific insight but also complicating our definitions of intelligence and understanding.
- Scientific discovery as embodied and intuitive – Jurado highlights that real science often begins in the gut, in a kind of embodied intuition before it becomes formalized. This challenges the myth of science as purely rational or step-by-step and instead suggests that hunches, sensory experience, and emotional resonance play a crucial role.
- Proteins as computational objects – Proteins aren’t just biochemical entities—they’re shaped by information. Their structure, function, and folding dynamics can be seen as computations, and tools like AlphaFold are beginning to unpack that informational complexity in ways that blur the line between physics and code.
- Human alignment is messier than AI alignment – While AI alignment gets a lot of attention, Jurado points out that human alignment—between scientists, institutions, and across cultures—is historically chaotic. This reframes the AI alignment debate in a broader evolutionary and historical context, questioning whether we're holding machines to stricter standards than ourselves.
- Standing on the shoulders of evolutionary processes – Evolution is not just a backdrop but an active epistemic force. Jurado sees scientists as participants in a much older system of experimentation and iteration—evolution itself. In this view, we’re not just designing models; we’re being shaped by them, in a co-evolution of tools and understanding.