Article Summary

Res Obscura explores the advancing capabilities of AI models in historical research and analysis. Obscura, initially skeptical due to AI's impact on student cheating, now sees significant potential in AI tools for research and self-directed learning.

The article presents three case studies demonstrating AI's proficiency in historical tasks:

1. Transcribing and translating 16th-century Italian handwriting from Urbano Monte's world map, with impressive accuracy.

2. Analyzing an 18th-century Mexican medical text, providing detailed iconographic analysis and historical context.

3. Generating new historical interpretations and research directions based on correspondence between William James and Francis Galton, scientists from the late 1800s.

Obscura highlights OpenAI's models (GPT-4o and o1) as particularly adept at core historical skills, including paleography, translation, and contextual analysis. These models can aggregate expert-level knowledge, recommend relevant academic sources, and offer novel perspectives on historical topics.

Obscura says that AI's potential to generate PhD-level analysis isn’t there yet, and he maintains that human consciousness and creativity remain irreplaceable. He argues that AI tends to converge on median viewpoints, lacking the truly original insights that human researchers can provide.

The article concludes by suggesting that while AI has reached first-year PhD-student-level capabilities in historical research, there may be an inherent upper limit to what these models can achieve. Obscura emphasizes the continuing value of human researchers in producing genuinely innovative historical interpretations and analyses.