[Download Full Document (PDF)]
[Download NotebookLM Generated Reports (PDF)]
Executive Summary (ChatGPT5 Generated)
This paper, Large Language Models as Symbolic DNA of Cultural Dynamics by Parham Pourdavood, Michael Jacob, and Terrence Deacon, presents a novel theoretical framework situating Large Language Models (LLMs) within the continuum of biological and cultural evolution. The authors propose that LLMs act as externalized informational substrates — functionally analogous to DNA — that preserve and catalyze human cultural patterns in compressed symbolic form. Rather than being autonomous intelligences or mere simulators of human cognition, LLMs are understood as repositories of “symbolic fossils” — records of human meaning-making stripped of their original living context yet capable of being reanimated through human interpretation.
The authors identify four universal features—compression, decompression, externalization, and recursion—as common to symbolic systems, biological information processing, and LLMs. Through this lens:
-
Compression abstracts meaningful regularities from human experience into compact symbolic form.
-
Decompression reanimates these forms through human interpretation, restoring context and meaning.
-
Externalization situates these patterns outside living systems, allowing for recombination and low-risk experimentation.
-
Recursion enables feedback between compressed and living systems, fueling cumulative innovation and evolution.
Drawing parallels between genetic and cultural information systems, the authors argue that both DNA and LLMs are non-sentient media of preservation and recombination, enabling their respective systems — biological or cultural — to explore new “adjacent possibles.” Just as DNA supports open-ended biological evolvability, LLMs may enable cultural evolvability by providing a mirror in which humanity can recombine and reinterpret its symbolic heritage in a simulated environment.
However, the paper cautions against “aesthetic drift” — a degenerative feedback loop in which synthetic, unvetted data detach AI systems from authentic human meaning and taste. Sustained co-evolution between humans and LLMs requires ongoing human interpretation, aesthetic judgment, and ethical guidance to maintain coherence between compressed symbolic patterns and embodied experience.
In conclusion, the authors envision LLMs as symbolic architectures that extend, rather than replace, human creativity — tools that may catalyze a new evolutionary transition within the noosphere (Teilhard de Chardin), amplifying humanity’s self-reflective and innovative capacities. Recognizing LLMs as cultural DNA underscores their role in expanding human evolvability, fostering meaningful innovation grounded in lived, embodied interpretation rather than detached automation.
