Archive — US — techcrunch
Astra and Opus just passed Turing’s other test
Cryptanalysts have used AI language models Astra and Claude Opus to decode long-unsolved Enigma messages from WWII. These breakthroughs mirror Turing's legacy in cryptography, demonstrating significant advancements in AI's capabilities in research and problem-solving, potentially unveiling solutions to the last remaining unbroken messages.
Key points
- Carter Leffen used OpenAI's Astra to decode an Enigma message that remained unsolved since 2005, effectively researching and simulating the Enigma machine.
- Frode Weierud validated Leffen's findings and was impressed by Astra's capabilities, noting its professional-level performance in cryptanalysis.
- Jack Willis utilized Anthropic’s Claude Opus 5 to decode a different message using a known signature, showcasing AI's collaborative potential in cryptography.
- Both models demonstrated the ability to perform tasks quickly that would typically take human researchers weeks or months.
Why it matters
- This development aligns with the increasing integration of AI in complex problem-solving tasks, reflecting the growing trend of employing AI technologies in data analysis and cybersecurity by 2025.
- Historically, Turing's work laid the foundational principles for computer science and cryptography, underscoring the profound impact AI innovations have on revisiting historical challenges.
The successful decoding of Enigma messages by AI models highlights a significant evolution in cryptanalysis, reaffirming the importance of AI in historical research and reinforcing Alan Turing’s legacy in the age of advanced technologies.