The use of AI in historical research has evolved significantly since 2024-25, particularly with the recent releases of GPT-6 Sol and Opus 5.5. These models are being utilized not only for tasks like transcribing documents but also for addressing historical problems. Collaborative efforts between historians and AI models may lead to advancements in historical knowledge and interpretation.
Key factors for successful AI application in historical research include the identification of solvable problems by experts, the availability of digitized data, and the ability of AI to conduct autonomous research across various datasets. Historical open problems that AI can assist with include cryptography, tracing texts across translations, and linking findings from niche subfields.
For example, GPT-6 Astra has been used to decrypt messages and identify passages from historical texts, such as recognizing a Latin translation by Isaac Newton from a French alchemical text. Additionally, the model has been involved in analyzing John Dee's coded manuscript, revealing patterns in the text and suggesting that some passages may encode meaning.
The collaboration between AI models and historical researchers has the potential to uncover new insights. For instance, GPT-6 is currently analyzing Charles Darwin's writings to find undiscovered links in the chain of knowledge between him and his informants. Similarly, Opus 5.5 is working with the papers of Samuel Hartlib, aiming to identify anonymous sources of scientific information.
Recent findings include the identification of anagrams used by both Newton and Hartlib for a key alchemical ingredient, suggesting a connection between their works. This discovery may warrant further investigation and potential publication due to its historical significance.
As AI continues to develop, its integration into historical research may yield unexpected results and enhance our understanding of the past.