Anthropic's Claude Fable 5.1 decodes a 370-year-old cryptogram by Thomas Erkhart, revealing a message in support of Charles II.
Anthropic's latest AI model, Claude Fable 5.1, has made headlines by successfully decoding a cryptogram known as Cyphral Distich, created by Scottish writer Thomas Erkhart. This cryptogram, which has puzzled experts for over 370 years, was recently decoded, as reported by the Vals AI project.
The cryptogram consists of two lines, each featuring 32 numbers, and it was included at the end of Erkhart's 1653 treatise, Logopandecteision. Initially brought to public attention in 1899, the cipher has since been a part of cryptographic literature and was ranked 28th among the most famous unsolved ciphers.
Decoding Methodology
Claude Fable 5.1 approached the challenge by focusing on two pivotal elements. The first was the presence of 32 "procurations"—numbered desires expressed by Erkhart, which concluded with phrases like "this wish". The second element was a prefatory poem suggesting that readers would uncover both their own desires and the author's thoughts within the cipher. By correlating the numbers to the procurations and the order of words, the model revealed a couplet supporting Charles II:
“God save King Charles the Second and make him the sovereign of this land.”
Vals AI confirmed the accuracy of the decoded message, noting the lines contain 32 letters and rhyme, aligning with Erkhart's intentions as a supporter of the Stuarts.
Decoding Another Cipher
Using a similar decoding strategy, Fable 5.1 also tackled another cipher, Cyphral Octastich, found in Erkhart's 1652 work, The Jewel. This eight-line verse comprises 285 numbers referencing page numbers in the book, although nine letters from the fifth line remain undecoded due to missing pages in digitized copies.
Decoding Process and Challenges
The decoding of Cyphral Distich took Fable 5.1 approximately 44 minutes, utilizing 176,000 tokens with minimal input from its operator, Gebi Jaff. Jaff had previously struggled for months with other models to solve historical ciphers, emphasizing that the hint in Erkhart's text should have been apparent to other AI systems as well.
Controversy Surrounding the Findings
Despite the excitement, the results have sparked skepticism among experts. A report from Reticuli Labs on GitHub attempted to replicate the findings and raised concerns about the algorithm used for decoding. They noted discrepancies, such as the letter "K" appearing in a position where no corresponding word began with that letter, suggesting randomness instead of a coherent solution. Additionally, they questioned the authenticity of the cipher's origin, pointing out that the digitized version of the 1653 edition does not include any numerical lines following the procurations.
The debate continues regarding the validity of the decoding, reflecting the complexities and challenges inherent in historical cryptography.
