Thoughts on Historical Language Models and Talkie-1930
2 days ago
- #AI in Humanities
- #Digital History
- #Historical Language Models
- Historical Language Models (Vintage LLMs) are emerging tools that simulate collective consciousness from historical texts, with Talkie-1930 as a recent example.
- Talkie-1930 is not strictly grounded in 1930 but reflects a floating index of ideas from the 19th and early 20th centuries, revealing period-specific epistemologies and biases.
- These models do not replace historical sources or enable authentic chats with historical figures, but they offer new ways to explore 'mental furniture'—assumptions and concepts of an era.
- Potential research applications include probing counterfactuals, analyzing genre and rhetoric through varied post-training data, and creating multi-agent simulations based on archival records.
- The field is open and interdisciplinary, with potential for novel humanistic research through open-source collaboration between STEM and humanities.